You are on page 1of 147

ORF Capacity Monitoring Study

- Highlights
Wireless Network Engineering
CTF/W20

Nortel Confidential Information

Agenda
> Introduction
Rational, Traffic Evolution, Call profile, Benchmark

> Capacity Monitoring method

Blocking & Resources Load Characterization


Congestion and Resources Load at network level
Top N most congested cells identification
Bottlenecks identification
Bottlenecks troubleshooting & analysis

> Capacity Monitoring Use case Method application


> Conclusions
3

Nortel Confidential Information

UTRAN capacity analysis introduction


> Objective:
Networks have been deployed focusing on the service availability at the expense
of capacity
Traffic is growing constantly :
Traffic evolution during last 6 months
200000000
180000000
160000000
140000000
120000000
100000000
80000000
60000000

CS DL traffic KBytes

PS DL Traffic KBytes

15 A pr 06

8 A pr 06

1 A pr 06

2 5 -m a rs -0 6

1 8 -m a rs -0 6

1 1 -m a rs -0 6

0 4 -m a rs -0 6

25 Feb 06

18 F eb 06

11 F eb 06

4 Feb 06

2 8 -ja n v -0 6

2 1 -ja n v -0 6

1 4 -ja n v -0 6

31 Dec 05

24 D ec 05

17 Dec 05

10 Dec 05

3 Dec 05

2 6 -n o v -0 5

1 9 -n o v -0 5

1 2 -n o v -0 5

0 5 -n o v -0 5

2 9 -o c t -0 5

2 2 -o c t -0 5

1 5 -o c t -0 5

0 8 -o c t -0 5

0 1 -o c t -0 5

20000000

0 7 -ja n v -0 6

40000000

total PS + CS

Operators start to experience some congestion in their networks and foresee


further capacity issues.
They expect capacity analysis guidelines in order to be ready for capacity
planning.

> Scope:
This presentation has been prepared in the scope of UA4.1 pointing out also
possible UA4.2 evolutions (when the case)
4

Nortel Confidential Information

UTRAN capacity analysis introduction


> Study coverage
The capacity assessment is done for a period of two weeks: w10 &
w11
All the cells of the networks were analyzed (5495 cells)
All the RNCs (19 RNCs)

> Software release


Nortel OAM Release OAM5.0
Nortel UTRAN Release UA4.1 (17 RNCs) and UA4.2 (Lyonecul1 &
Lyontrio1)

> Elements of NodeB configuration:


Multi-carrier BTS : only some of the VO sites

X=1

X=2

X=3

X=4

BTS with X CEM

2058

28

BTS with X PCM

2060

29

Nortel Confidential Information

Call Profile
ORF
Metric name
BHCA
Mean Call Duration
Active Set Update per second
Mean Sector per user
Percentage of softer HO
Inter-RAT cell reselection
SMS
Nb of AO Downsize per PS call
Nb of AO Upsize per call
Nb of AO Release per call
Nb of IRM Sched Downgrade per
call
% of 3G-2G HO per CS call
% of 3G-2G HO per PS call
Number of 2G-3G CS HHO

PS
CS V CS D 384
1,05
95

0.01
108

0,01
636

Operator 1
PS
128

PS
Ps 64 CS V CS D 384

0,06 0,001 0,93


110 452,0 55

0,01
48

0,19
34

PS
128
0,02
34,9

Operator 3
PS
Ps 64 CS V CS D 384
0,01
23

2,04
82

0,06
103

0,26
50

0,12

0,18

0,145

1,71

1,95

1,6

27%
1,36
0,33
2,36
1,88
0,25

35%
1,81
0,27
1,57
0,79
0,68

24%
1,23
0,42
1,75
1,2
0,44

0,07

0,01

0,08

11%
13%
0

23%
14%
0

5%
10%
0

PS
128
0,03
119

Ps 64
N/A
N/A

Significant differences between ORF profile and other operators:


Low number of PS calls/Sub but very long calls (impact on UPlane) >
AON timers settings consequence.
Mean sector per user reasonable value (OK from capacity point of view).
Very low number of AON release per call unnecessary resources usage.
6

Nortel Confidential Information

Benchmark
Traffic Type & Volume
DedicatedDownlinkKbytesRlc (All traffic & PS only)

RBSetupRequest PS Repartition (%)


60000000

100%
90%

50000000

80%

40000000

70%
KByts

60%
50%
40%

30000000
20000000

30%

10000000

20%
10%

0
Day 1

0%
Day 1

Day 2

ORF PS384

ORF PS128

Day 3
OP1 PS384

Day 4
Op1 PS128

Op2 PS384

Day 5
Op2 PS128

Day 2

Day 3

Day 4

RNC: LYON1 (PS)

RNC: Op1 (PS)

RNC: Op2 (PS)

RNC: LYON1 (All)

RNC: Op1 (All)

RNC: Op2 (All)

Day 5

> Three different operators with three different traffic allocation strategies:

Orange mix of PS128 & PS384 RBs


Operator 1 mainly PS 384 (more than 95%)
Operator 2 mainly PS 128 (more than 90%)

> In terms of Volume: The most loaded is Operator 2 followed by Operator 1


which is also the most loaded in terms of PS traffic.

For this benchmark, ORF RNC (Lyonecul1) is the less loaded RNC
in terms of total traffic and PS traffic, but it has an homogeneous
PS traffic mix.
7

Nortel Confidential Information

Benchmark
High Level Blocking analysis
Blocking rate comparison (RNC Level)
0,80%

Lyn1 Blocking cause = Iub


Op1 Blocking cause = Codes
Op2 Blocking cause = RF Power

0,70%
0,60%
0,50%
0,40%
0,30%
0,20%
0,10%
0,00%

Blocking rate Light % Lyn1

Blocking rate Light % Op1

Blocking rate Light % Op2

> A Blocking rate Light taking into account all causes except unspecified is
calculated:
The Unspecified cause is not considered as it concerns only the CEM which is
not managed by the iRM in UA4.1 & UA4.2.
All other causes (Iub, Codes, RF Power) are covered by this metric.

> Operators 1 & 2 are almost not blocked at all (< 0,1%), blocking coming
from Codes (Op1) & RF Power (Op2)

Lyonecul1 is the most blocked RNC, the entire blocking coming


from Iub. For detailed iRM Benchmark information click Here
8

Nortel Confidential Information

Agenda
> Introduction
Rational, Traffic Evolution, Call profile, Benchmark

> Capacity Monitoring method

Blocking & Resources Load Characterization


Congestion and Resources Load at network level
Top N most congested cells identification
Bottlenecks identification
Bottlenecks troubleshooting & analysis

> Capacity Monitoring Use case Method application


> Conclusions
9

Nortel Confidential Information

Blocking Characterization
> Blocking situation definition : each time the UE requests an UTRAN
resource and is not getting it because the resource is not available.
Call Phase

Blocking Cause

Effect

Monitoring detection
method

Call Admission

Lack of resources at
call setup

Call admission failure

- RB Establishment Unsuccess

Lack of resources to
perform iRM
transitions (AON
Upsize, iRM Sched
Upgrade)

Call is not reconfigured - RL Reconfiguration Unsuccess


(iRM Reconf)
(impact on user
throughput)
- RL Setup Unsuccess (AON)

Call
Reconfiguration

Mobility

- RL Reconfiguration Unsuccess
and Cancel

- AON/iRM Scheduling failure


counters

- RL Setup Unsuccess
No resources
Additional RL not
available for additional added in the Active Set - RL Addition Unsuccess
RL
(risk of call drop)

- RL & RB Reconfiguration
Unssuccess for HSDPA

This study is mainly focused on Blocking during Call Admission phase as it


is the most impacting for call success.
In general, if the call admission blocking is solved, the blocking on other call
phases is also solved (same bottlenecks involved).
10

Nortel Confidential Information

Resources Load
Main possible bottleneck resources monitored at UTRAN Level:
> BTS Channel Elements - BTS Resource

The load can be monitored at BTS level


Blocking of this resource RB rejection or RL Setup/Addition/Reconfiguration
failures

> Iub ATM

The load can be monitored at BTS level (AAL2&AAL5 traffic)


alternative solution monitor the Traffic load per BTS at RNC level
Blocking of this resource RB rejection

> RF power - BTS Resource

The load can be monitored per cell via BTS or RNC counters
Blocking of this resource RB rejection

> OVSF Codes - RNC Resource

The load can be monitored per cell via RNC counters


Blocking of this resource RB rejection

> RNC CPU

11

The load can be monitored at card level through RNC counters


Blocking of this resource Overload mechanism => RB rejection
Nortel Confidential Information

Congested cells @ Network level


Global Blocking
Blocking Rate at cell BH Distribution
4989

100%

4500

90%

4000

80%

3500

70%

2807

3000
2500

60%
50%

2102

2000

40%

1500

% of cells

Nb of Cells concerned

5000

30%

1083

1000

20%

273

500

37

>7%

>10%

>20%

>30%

0
>=0%

>0%

>0,5%

>2%

>5%

10%

104

0%

Blocking rate > x %


Blocked cells Count

Blocked cells %

> Blocking Rate per cell is computed as a ratio between the RB Blocking at BH and
RB Setup Request at BH for the entire observation period (BH = Busy Hour).
> Main contributor to this high blocking level was a Core CS dimensioning issue
(MSC Trion) which affected almost all Nortel RNCs. The rest is coming from Iub
congestion.

12

Significant amount of blocking (> 2%) observed during w10 &


w11 at Cell BH for 22% of the cells
Nortel Confidential Information

CEM load @ Network level


CEM Load Distribution (Avg & Max @ BH)

4979
4757

4757

4500

100%
90%

4144

Nb of Cells concerned

4000

80%

3482

3500

70%

3000

60%

2717

2500

50%

2000

40%

1500

30%

1060
765

1000
500

20%

116

15

279
4

0
>10%

>20%

>30%

% of cells

5000

>40%

>50%

>60%

64

>70%

17

10%

>80%

>90%

0%

>95%

CEMUsed > x %
Max Load Count

Avg Load Count

Max Load (%)

Avg Load (%)

> During the Observation period, most part of the CEMs were loaded in average
less than 50% (the attention limit)
> When investigating the Max CEMUsed for the same period, we can see that 64
cells (and the associated CEMs) exceeded the 70% load which represents the
upper blocking limit of PS384.

13

37 BTS (64 cells) are reaching the dangerous zone of the CEM
Nortel Confidential Information

Iub load @ Network level


Iub real Load Distribution (Max@BH)
100%

4484

4500

90%

4000

80%

3500

70%

3016

3000

60%

2500

50%

2000

40%

1500

% of cells

Nb of Cells concerned

5000

30%

1074

1000

20%

311

500

112

48

18

>20%

>25%

>30%

>35%

40%

10%
0%

>2%

>5%

>10%

>15%

Iub Load > x %


Max Load Count

Max Load (%)

> As it is based on a BTS counter, the monitored Iub load (@ BH) has to be seen
as the average load per hour. In the graph, the Max of these averages (for the
observation period) is captured.

Iub utilization is very low, not exceeding 35% of the Iub bandwidth.
This is not preventing from having Iub blocking as the admission
criteria is based on the estimated Iub load (at the RNC)
14

Nortel Confidential Information

DL Power load @ Network level


DL RF Power Distribution (Avg & Max)
49574954

4780
4390

4500

100%

4326

90%

Nb of Cells concerned

4000

80%

3500

70%

3080

3000

60%

2503

2500

50%

2000

40%

1486

1500

30%

1000

283

500

% of cells

5000

20%

514
83

26

151

12

45

10%
0%

>10%

>20%

>25%

>30%

>35%

>40%

>45%

>50%

>60%

>70%

>80%

DL Pow er Load > x %


Max Load Count

Avg Load Count

Max Load (%)

Avg Load (%)

> During the Observation period, most part of the cells were loaded in average less
than 50%
> When investigating the Max load for the same period, we can see that only 1 cell
punctually exceeded 80% (Yellow2Red iRM power threshold).

DL power load is relatively low in the network. The loaded sites


being mostly sites from VO zone (JONAGE_COTTAGE, MEYZIEU_CARREAU,
MEYZIEU_GD_LARGE, MEYZIEU_OUEST) which use PA/2 (due to STSR2)
15

Nortel Confidential Information

Codes load @ Network level


OVSF Codes Load Distribution (Avg & Max @ BH)
4933

4797

4873

100%

4500

90%

3746

Nb of Cells concerned

4000

80%

3377

3500

70%

3000

60%

2500

50%

2000

40%

1352

1500

30%

745

1000

20%

459

500

% of cells

5000

93

54

10

12

10%

0%
>5%

>10%

>20%

>30%

>40%

>50%

>60%

>70%

>80%

Codes Load > x %


Max Load Count

Avg Load Count

Max Load (%)

Avg Load (%)

> During the Observation period, all the cells were loaded in average less than 50%
> When investigating the Max load for the same period, we can see that only 2 cells
punctually exceeded 70% (Yellow2Red iRM codes threshold).

16

Low codes load => only two cells being able to reach a max load
leading to a red iRM cell color.
Nortel Confidential Information

RNC CPU load @ Network level


TMU load Avg & Max
100
90
80

TM U Load (%)

70
GRENOBLE1 Avg

60

GRENOBLE1 Max
LYONECUL1 Avg

50

LYONECUL1 Max
MONTPELL1 Avg

40

MONTPELL1 Max

30
20
10
0
13/3/ 13/3/ 13/3/ 13/3/ 14/3/ 14/3/ 14/3/ 14/3/ 15/3/ 15/3/ 15/3/ 15/3/ 16/3/ 16/3/ 16/3/ 16/3/ 17/3/ 17/3/ 17/3/ 17/3/ 18/3/
06
06
06
06
06
06
06
06
06
06
06
06
06
06
06
06
06
06
06
06
06
0:00 6:00 12:00 18:00 0:00 6:00 12:00 18:00 0:00 6:00 12:00 18:00 0:00 6:00 12:00 18:00 0:00 6:00 12:00 18:00 0:00

> Only 3 RNCs are captured in this graph (3 most loaded from TMU point of view)
> For these 3 RNCs, the TMU load is remaining in acceptable limits:

Average < 20%


Max < 35% (the only exception is Montpell1 with some local issues on 15-03-06)

TMU load is far from being a concern for ORF network


17

Nortel Confidential Information

Agenda
> Introduction
Rational, Traffic Evolution, Call profile, Benchmark

> Capacity Monitoring method

Blocking & Resources Load Characterization


Congestion and Resources Load at network level
Top N most congested cells identification
Bottlenecks identification
Bottlenecks troubleshooting & analysis

> Capacity Monitoring Use case Method application


> Conclusions
18

Nortel Confidential Information

Top N most congested cells identification


> Counters Granularity: Hour
For each Cell : 4 metrics calculation
RbRejected@BH where BH=Hour[Max(RbSetReq)]
days

RbRejected@BkH where BkH=Hour[Max(RbRejected)]


days
RbRejected
days hour
BHBKRate(%) = RbRejected / RbSetReq (@BH)
Condition1= NbrDays[BHBKRate(%)> 0%]
Condition2= NbrDays[BHBKRate(%)> 2%]
N.B RbRejected corresponds to counter VS.RadioBearerEstablishmentUnsucces (without screening 0)
RbSetReq corresponds to counter VS.RadioBearerSetupRequest

Method aims to create a list of most representative cells in terms of


blocking by taking into account the amount of rejections at different
periods of the day but also the persistence in time of the issue.
19

Nortel Confidential Information

Top N most congested cells identification


CellsRanking
[ RbRejected]
day hour

CellsRanking
[RbRejected@BH]

Cells Filtering
[Condition1>X & Condition2>Y ]
X>Y

CellsRanking
[RbRejected@BkH]

Bottleneck Identification
[RbRejected Screening]

In our Study:
Obs.Period =14 days
X=7
Y=4
20

Final cells List to investigate


Nortel Confidential Information

Top N most congested cells


identification - Result
Cell
GRE_ALSACE_LORRAINE_U21
DJN_HAVRE_U11
BOURGOIN_EST_U11
BIONNE_TDFT_U31
ST_ETN_GARE_U11
GRE_EUROPOLE_U11
FERREOL_U31
ECHIROLLES_U11
FERREOL_U11
GRE_JEAN_JAURES_U11
HYERES_COUBERTIN_U11
BEAUREGARD_U21
NIMES_MONT_DUPLAN_U21
MARIGNANE_CENTRE_U21
ECHIROLLES_U31
AIX_CARCASSONNE_U21
FONTAINE_CHARM_U21
GRE_EUROPOLE_U21
HAUTS_VOREPPE_TDF_U21
DJN_CFRT_U11
DJN_HAVRE_U21
GRE_JEAN_JAURES_U31
TOULON_VALETTE_SUD_U21
RABATEAU_U31
GRE_AMPERE_DRAC_U11
SALON_SUD_U31
BEAUREGARD_U31
ALLAUCH_U31
ISTRES_LES_COGNETS_U11
VIENNE_PIPET_U11
GARIBALDI_U11
MARSEILLE_PARTY_U21
FONTAINE_OUCHE_U31
TERRON_U21
ST_EGREVE_BONNAIS_U11
GRE_EAUX_CLAIR_U31
MRS_MONTE_CRISTO_TDFT_U21
BIONNE_TDFT_U21
MARIGNANE_CENTRE_U11
GRE_AMPERE_DRAC_U21
SEYSSINET_NORD_U11
TOUR_PIN_U31
CROLLES_U11
TRAMOYES_U31
CORDERIE_U11
DJN_MARCS_OR_U21
LES_CHARTREUX_U21
ST_GELY_DU_FESC_TDFT_U31
DJN_VICTOR_HUGO_U11
GRE_ABBAYE_U31
MAUGUIO_U31
ST_JEROME_U31
AIGUELONGUE_U11
DJN_HAVRE_U31

21

BK Day

Setup Day

Bk rate day

Bk & BH

Setup & BH

Bk rate BH

Bk & BkH

Setup & BkH

Bk rate BhK

796
515
279
338
504
297
317
245
267
247
279
179
249
172
193
310
144
161
145
161
215
122
145
133
135
97
114
107
101
113
103
99
95
128
98
107
112
122
83
85
84
78
77
140
79
70
99
61
74
69
64
58
58
85

75504
23104
21612
27698
1962
20546
20132
17564
17850
20098
7600
12697
5761
13129
14707
4504
11916
13154
9049
14210
10344
8082
8726
10587
9464
9278
5931
8335
10379
9373
7631
8565
9542
995
6121
7513
8915
10167
7014
8765
6999
5121
5500
3911
3923
8143
3248
8044
8775
6977
4713
3889
5308
3788

1,1%
2,2%
1,3%
1,2%
25,7%
1,4%
1,6%
1,4%
1,5%
1,2%
3,7%
1,4%
4,3%
1,3%
1,3%
6,9%
1,2%
1,2%
1,6%
1,1%
2,1%
1,5%
1,7%
1,3%
1,4%
1,0%
1,9%
1,3%
1,0%
1,2%
1,3%
1,2%
1,0%
12,9%
1,6%
1,4%
1,3%
1,2%
1,2%
1,0%
1,2%
1,5%
1,4%
3,6%
2,0%
0,9%
3,0%
0,8%
0,8%
1,0%
1,4%
1,5%
1,1%
2,2%

230
142
188
124
143
101
81
105
80
86
86
103
76
100
76
70
74
84
59
58
52
73
55
58
48
62
52
52
49
54
60
51
51
43
60
43
41
40
38
44
35
37
47
20
33
35
25
29
32
37
41
29
36
21

6126
2382
2313
2797
370
2483
2054
2036
1949
2213
818
1369
713
1378
1690
739
1393
1524
1034
1457
1101
1029
1035
1107
1118
1056
730
1052
1179
1063
844
998
1506
263
724
1106
971
1191
1023
1040
758
594
644
515
609
959
453
821
993
788
578
547
621
484

3,8%
6,0%
8,1%
4,4%
38,6%
4,1%
3,9%
5,2%
4,1%
3,9%
10,5%
7,5%
10,7%
7,3%
4,5%
9,5%
5,3%
5,5%
5,7%
4,0%
4,7%
7,1%
5,3%
5,2%
4,3%
5,9%
7,1%
4,9%
4,2%
5,1%
7,1%
5,1%
3,4%
16,3%
8,3%
3,9%
4,2%
3,4%
3,7%
4,2%
4,6%
6,2%
7,3%
3,9%
5,4%
3,6%
5,5%
3,5%
3,2%
4,7%
7,1%
5,3%
5,8%
4,3%

516
227
227
224
158
193
192
185
164
151
117
124
129
125
124
112
143
111
107
100
85
92
92
93
94
96
82
81
88
76
76
81
82
72
71
77
67
65
88
58
62
62
53
64
57
62
55
77
54
50
50
61
51
41

4638
2045
1909
2352
314
1782
1507
1559
1519
1740
651
954
539
1124
1265
431
1040
1045
843
1062
897
690
809
838
627
672
532
858
727
804
619
720
747
157
507
633
710
786
572
709
542
394
357
275
430
693
271
514
655
443
433
321
457
334

11,1%
11,1%
11,9%
9,5%
50,3%
10,8%
12,7%
11,9%
10,8%
8,7%
18,0%
13,0%
23,9%
11,1%
9,8%
26,0%
13,8%
10,6%
12,7%
9,4%
9,5%
13,3%
11,4%
11,1%
15,0%
14,3%
15,4%
9,4%
12,1%
9,5%
12,3%
11,3%
11,0%
45,9%
14,0%
12,2%
9,4%
8,3%
15,4%
8,2%
11,4%
15,7%
14,8%
23,3%
13,3%
8,9%
20,3%
15,0%
8,2%
11,3%
11,5%
19,0%
11,2%
12,3%

Nortel Confidential Information

rank
1
3
6
6
7
9
12
12
16
17
21
22
22
23
25
26
31
31
43
44
52
52
53
55
61
70
74
78
79
81
83
86
87
89
91
92
105
106
115
140
154
159
168
173
177
177
185
189
195
208
211
224
234
272

days with
Bk>2%
4
7
7
5
8
4
4
6
5
5
8
7
8
6
5
11
6
6
5
4
7
6
6
6
7
5
7
7
7
5
7
5
6
11
7
6
6
6
5
6
5
7
7
7
7
6
8
6
4
7
9
6
7
6

days with
Bk>0%
7
8
9
10
8
7
7
7
7
8
8
9
9
9
7
11
7
7
8
7
9
7
7
8
9
7
8
9
7
7
7
8
7
11
7
7
9
7
7
9
7
7
7
7
7
7
8
7
7
7
9
7
7
7

Two filter criteria are


used:
- BK > 2 % for at least 4
days form observation
period
- BK > 0 % for at least 7
days form observation
period
The rank is calculated
following Step 1

54 regularly
blocked
cells with
high impact
on user
perception

Bottlenecks identification
The different blocking causes to be investigated are identified
through VS.RadioBearerEstablishmentUnsuccess screenings)

Scr. 0: invalid RAB parameters value not to be considered as blocking


cause, thus not taken into account in the blocking rate computation
Scr. 1: unavailable dl code resources Code shortage indication
Scr. 2: unavailable dl power resources DL power shortage indication
Scr. 3: Unspecified covers CEM resources and other possible issues during
call setup (Core Network UP Init. failures...)
Scr. 4: Lack of bandwidth Iux bandwidth allocation shortage (all interfaces:
Iub/Iu/Iur) this screening is replaced by screenings #10, #12 and #14 in
UA04.2
Scr. 5: Lack of CID Iux CIDs allocation shortage (all interfaces: Iub/Iu/Iur)
this screening is replaced by screenings #9, #11 and #13 in UA04.2
Scr. 6: CAC RNC Processing resources TMU resources shortage (overload
situations)
Scr. 9: Lack of CID on the Iu introduced in UA04.2
Scr. 10: Lack of bandwidth on the Iu introduced in UA04.2
Scr. 11: Lack of CID on the Iur introduced in UA04.2
Scr. 12: Lack of bandwidth on the Iur introduced in UA04.2
Scr. 13: Lack of CID on the Iub introduced in UA04.2
Scr. 14: Lack of bandwidth on the Iub introduced in UA04.

22

Nortel Confidential Information

Bottlenecks identification - Result


Cell
GRE_ALSACE_LORRAINE_U21
DJN_HAVRE_U11
BOURGOIN_EST_U11
BIONNE_TDFT_U31
ST_ETN_GARE_U11
GRE_EUROPOLE_U11
FERREOL_U31
ECHIROLLES_U11
FERREOL_U11
GRE_JEAN_JAURES_U11
HYERES_COUBERTIN_U11
BEAUREGARD_U21
NIMES_MONT_DUPLAN_U21
MARIGNANE_CENTRE_U21
ECHIROLLES_U31
AIX_CARCASSONNE_U21
FONTAINE_CHARM_U21
GRE_EUROPOLE_U21
HAUTS_VOREPPE_TDF_U21
DJN_CFRT_U11
DJN_HAVRE_U21
GRE_JEAN_JAURES_U31
TOULON_VALETTE_SUD_U21
RABATEAU_U31
GRE_AMPERE_DRAC_U11
SALON_SUD_U31
BEAUREGARD_U31
ALLAUCH_U31
ISTRES_LES_COGNETS_U11
VIENNE_PIPET_U11
GARIBALDI_U11
MARSEILLE_PARTY_U21
FONTAINE_OUCHE_U31
TERRON_U21
ST_EGREVE_BONNAIS_U11
GRE_EAUX_CLAIR_U31
MRS_MONTE_CRISTO_TDFT_U21
BIONNE_TDFT_U21
MARIGNANE_CENTRE_U11
GRE_AMPERE_DRAC_U21
SEYSSINET_NORD_U11
TOUR_PIN_U31
CROLLES_U11
TRAMOYES_U31
CORDERIE_U11
DJN_MARCS_OR_U21
LES_CHARTREUX_U21
ST_GELY_DU_FESC_TDFT_U31
DJN_VICTOR_HUGO_U11
GRE_ABBAYE_U31
MAUGUIO_U31
ST_JEROME_U31
AIGUELONGUE_U11
DJN_HAVRE_U31

23

BK Day

Setup Day

Bk rate day

Bk & BH

Setup & BH

Bk rate BH

796
515
279
338
504
297
317
245
267
247
279
179
249
172
193
310
144
161
145
161
215
122
145
133
135
97
114
107
101
113
103
99
95
128
98
107
112
122
83
85
84
78
77
140
79
70
99
61
74
69
64
58
58
85

75504
23104
21612
27698
1962
20546
20132
17564
17850
20098
7600
12697
5761
13129
14707
4504
11916
13154
9049
14210
10344
8082
8726
10587
9464
9278
5931
8335
10379
9373
7631
8565
9542
995
6121
7513
8915
10167
7014
8765
6999
5121
5500
3911
3923
8143
3248
8044
8775
6977
4713
3889
5308
3788

1,1%
2,2%
1,3%
1,2%
25,7%
1,4%
1,6%
1,4%
1,5%
1,2%
3,7%
1,4%
4,3%
1,3%
1,3%
6,9%
1,2%
1,2%
1,6%
1,1%
2,1%
1,5%
1,7%
1,3%
1,4%
1,0%
1,9%
1,3%
1,0%
1,2%
1,3%
1,2%
1,0%
12,9%
1,6%
1,4%
1,3%
1,2%
1,2%
1,0%
1,2%
1,5%
1,4%
3,6%
2,0%
0,9%
3,0%
0,8%
0,8%
1,0%
1,4%
1,5%
1,1%
2,2%

230
142
188
124
143
101
81
105
80
86
86
103
76
100
76
70
74
84
59
58
52
73
55
58
48
62
52
52
49
54
60
51
51
43
60
43
41
40
38
44
35
37
47
20
33
35
25
29
32
37
41
29
36
21

6126
2382
2313
2797
370
2483
2054
2036
1949
2213
818
1369
713
1378
1690
739
1393
1524
1034
1457
1101
1029
1035
1107
1118
1056
730
1052
1179
1063
844
998
1506
263
724
1106
971
1191
1023
1040
758
594
644
515
609
959
453
821
993
788
578
547
621
484

3,8%
6,0%
8,1%
4,4%
38,6%
4,1%
3,9%
5,2%
4,1%
3,9%
10,5%
7,5%
10,7%
7,3%
4,5%
9,5%
5,3%
5,5%
5,7%
4,0%
4,7%
7,1%
5,3%
5,2%
4,3%
5,9%
7,1%
4,9%
4,2%
5,1%
7,1%
5,1%
3,4%
16,3%
8,3%
3,9%
4,2%
3,4%
3,7%
4,2%
4,6%
6,2%
7,3%
3,9%
5,4%
3,6%
5,5%
3,5%
3,2%
4,7%
7,1%
5,3%
5,8%
4,3%

rank
1
3
6
6
7
9
12
12
16
17
21
22
22
23
25
26
31
31
43
44
52
52
53
55
61
70
74
78
79
81
83
86
87
89
91
92
105
106
115
140
154
159
168
173
177
177
185
189
195
208
211
224
234
272

Blocking Cause
Unspecified
LackOfBandwidth & Unspecified
Unspecified
Unspecified
LackOfBandwidth
LackOfBandwidth & Unspecified
LackOfBandwidth & Unspecified
LackOfBandwidth & Unspecified
LackOfBandwidth & Unspecified
Unspecified
LackOfBandwidth
LackOfBandwidth & Unspecified
LackOfBandwidth
Unspecified
Unspecified
LackOfBandwidth
Unspecified
LackOfBandwidth & Unspecified
Unspecified
Unspecified
LackOfBandwidth & Unspecified
Unspecified
LackOfBandwidth & Unspecified
Unspecified
Unspecified
Unspecified
LackOfBandwidth & Unspecified
Unspecified
Unspecified
LackOfBandwidth & Unspecified
Unspecified
Unspecified
Unspecified
LackOfBandwidth
Unspecified
LackOfBandwidth
Unspecified
Unspecified
Unspecified
LackOfBandwidth & Unspecified
Unspecified
Unspecified
Unspecified
Unspecified
Unspecified
Unspecified
Unspecified
Unspecified
LackOfBandwidth & Unspecified
Unspecified
Unspecified
Unspecified
Unspecified
LackOfBandwidth & Unspecified

Nortel Confidential Information

days with
Bk>2%
4
7
7
5
8
4
4
6
5
5
8
7
8
6
5
11
6
6
5
4
7
6
6
6
7
5
7
7
7
5
7
5
6
11
7
6
6
6
5
6
5
7
7
7
7
6
8
6
4
7
9
6
7
6

days with
Bk>0%
7
8
9
10
8
7
7
7
7
8
8
9
9
9
7
11
7
7
8
7
9
7
7
8
9
7
8
9
7
7
7
8
7
11
7
7
9
7
7
9
7
7
7
7
7
7
8
7
7
7
9
7
7
7

Main observations:
- Practically no DL
Power or Codes
blocking monitored
during Obs. Period.
(this is specific to the
entire network, not
only TopN list)
- Unspecified &
LackOfBandwidth are
the only blocking
causes monitored in
the network.

Agenda
> Introduction
Rational, Traffic Evolution, Call profile, Benchmark

> Capacity Monitoring method

Blocking & Resources Load Characterization


Congestion and Resources Load at network level
Top N most congested cells identification
Bottlenecks identification
Bottlenecks troubleshooting & analysis

> Capacity Monitoring Use case Method application


> Conclusions
24

Nortel Confidential Information

Bottleneck analysis
Unspecified
Blocking detection

> The Unspecified RB Establishment Unsuccess is not always directly


linked to a specific UTRAN bottleneck, thus the association with the load
is not obvious for this blocking cause.
> This blocking cause is identified through screening Unspecified of the
VS.RadioBearerEstablishmentUnsuccess counter - See bullet 6 in the RB
Allocation Procedure Data Flow Diagram.
> It covers the following Call Setup failures causes:
CEM resources shortage
User Plane Initialization failure between RNC and Core Network (Iu issues).
From UA4.2 the HSDPA CAC rejected calls are also pegging this counter.

> In general, it is a good method to detect the CEM resources shortage as


the CEMAllocFail counter has some limitations (see CEM analysis).
> HSDPA CAC is also pegging the UnsuccesfulHsdpaCac counter (#0958)

25

Nortel Confidential Information

Bottleneck analysis
Unspecified
Detailed investigation method
RB Establishment Unsuccess

no

Unspecified > 0

Stop

yes

RL Reconfiguration Prepare Unsuccess


RLReconfFailure RB Establishment Unsuccess
Unspecified

no
RL Reconfiguration Cancel > 0
And correlated in time with
RB Establishment Unsuccess
Unspecified

no
Unsucc. HSDPA CAC > 0
And correlated in time with
RB Establishment Unsuccess
Unspecified
26

yes

yes

Call allocation failure due to lack of CEM resources


(HBBU or DBBU) => relay on CEM Bottleneck analysis

Possible RNC - Core Network UPlane initialization issues. Confirmed if


the issue occurs on several BTSs of the same RNC (depending on their
traffic load).
It is possible to identify the Core domain generating the issue by
checking the RadioLinkReconfigurationCancel.DlAccessStratumConf5 for CS or
DlAccessStratumConf2/6/7 for PS
Check Core Network Counters & Alarms for more details

yes

The HSDPA limit of 20 UEs per cell is reached (HSDPA CAC)


Nortel Confidential Information

Bottleneck analysis
Unspecified

RadioBearerEstablishmentUnsuccess.Unspecified vs. RadioLinkReconfigurationCancel

Example
> During observation period (w10 &
w11) a large amount of

160
140

120

VS.RadioBearerEstablishmentUnsuccess.Un
specified was monitored on a large

number of ORF BTSs. It was


perfectly correlated with the

100
80
60

40

VS.RadioLinkReconfigurationCancel:

20

> The investigations showed that only

VS.RadioLinkReconfigurationCancel.DlAcce
ssStratumConf5 (CS voice) was

pegged, thus a possible issue


between RNC and the Core CS.

VS.RadioLinkReconfigurationCancel

VS.RadioBearerEstablishmentUnsuccess.Unspecified
7000

> When investigated at RNC level, the


issue was more visible on 6 RNCs
connected to the same MSC (and
probably to the same ALI board)
> The issue disappeared after some
operations of MSC side (RNC
reparenting & ALI redimensioning) but
it seems to be back again after 3
weeks (to be checked):
27

VS.RadioBearerEstablishm entUnsuccess.Unspecified

6000

5000
GRENOBLE1
4000

MARSESTM2
MONTPELL1
MARSESTM1

3000

DIJONC1
TOULON_SF1

2000

1000

Nortel Confidential Information

Bottleneck analysis
BTS Channel elements (CEM)
Blocking Detection

>

The CEM blocking is particular annoying as this resource is not managed


by the RNC. Thus, if no CEM resources available, the call is released (no
iRM downgrade based on this resources load is possible until UA5.0)

>

Call admission blocking due to CEM resources shortage can be detected


by two methods:

1. Using the VS.RadioBearerEstablishmentUnsuccess.Unspecified counter


see the Unspecified analysis above.
2. Using BTS counter CEMAllocFail. Method drawbacks:

>

This counter provides the % of CEM resources requests that are refused from
total CEM resources requests. All requests are considered: RL Setup SRB or
TRB, RL Reconfiguration Up or Down thus it will be slightly incremented even
in case of high blocking.
It is rounded to Integer (if 0.8% is calculated => 0% is displayed)

Recommendation: use the first method and only confirm (if it is the case)
with the CEMAllocFail method.
28

Nortel Confidential Information

Bottleneck analysis
BTS Channel elements (CEM)
Resource Load evaluation
> CEM load (DBBU) is measured by means of CEMUsed (BTS counter)
>

The CEMUsed screenings (Avg, Min, Max.) allows to:

>

Min during the night it will provide info on Common Channels utilization (usually 18% for
STSR1 on 1 iCEM64)
Avg Average load during one Hour used for dimensioning purposes
Max allows to determine the moments when the alarm thresholds were exceeded
triggering blocking.
RAB Type (UL/DL)
Block. thresholds

CEMUsed booking levels for


the most used RABs in ORF
in case of 1 iCEM64:

CS 12.2/12.2

98%

PS 64/64, 128/128, 128/384, 128/0


(HSDPA)

85%

PS 64/384

72%

PS 384/0 (HSDPA)

65%

>

Recommendation: set an Yellow alarm monitoring trigger for CEMUsed.Avg


exceeding 50% and a red one for CEMUsed.Max exceeding constantly 70%.

>

CEM load (HBBU) cannot be directly monitored in UA4.2. Alternative through


DlAsConfIdAvgNbrEstab.Max is proposed (next slide). Complete solution in UA5.0
29

Nortel Confidential Information

Bottleneck analysis
BTS Channel elements (CEM)
Detailed investigation method
no

Unspecified analysis
shows CEM blocking

Stop

yes (Optional)
yes
CEMAllocFail >

Call allocation
failure due to
lack of CEM

CEMUsed.Max

> 70%

yes
Plan to add additional
CEM (DBBU) resources

30

no

CEMAllocFail could display 0 even with


High PS blocking if there are many
resources requests that are accepted (small
RABs) => continue CEM investigation

yes
no

UA4.2
UA5.0
MAX
[DlAsConfIdAvgNbrEstab.Max
HsdpaUesperHbbu.Max
=X
(scr.27)] ~ X

yes

Plan to add additional


CEM (HBBU) resources

no
-Potential issues with CEM configuration
(check CCP configuration).
- STSR2 is activated and the traffic is
unbalanced between frequencies => Add
additional CEM resources
Nortel Confidential Information

X = 20 (UA4.2)
X = 48 (UA5.0)

Bottleneck analysis
Iub Interface
Blocking Detection

> The Iub resources management is done at RNC Level. From UA4.2 this
resource will be taken into account in the iRM @ CAC mechanism.
> Call admission blocking due to Iub resources shortage can be detected
through the counter:

VS.RadioBearerEstablishmentUnsuccess.LackOfBandwidth & LackOfCid


In UA4.1 this counter is not making the difference between different Iux interfaces,
thus additional check is needed => if problem is seen only on some (isolated)
BTSs, then it is an Iub problem.
From UA4.2, specific screenings per Iux interface will be available (better
precision during bottleneck identification phase)

> Additional indication about call admission blocking and iRM upgrade
procedures fail due to lack of Iub resources is provided by:
VS.RadioLinkReconfigurationPrepareUnsuccess.INodeRefusal (enhanced by
two new screenings in UA4.2 LackBwthIub & LackCid)
> When the mobility is affected by the lack of Iub resources the following
counter will be pegged: VS.RadioLinkSetupUnsuccess.INodeRefusal
(enhanced by two new screenings in UA4.2 LackBwthIub & LackCid)
31

Nortel Confidential Information

Bottleneck analysis
Iub Interface
Resource Load evaluation

> Iub Load indicators available in UA4.1:


Real load of the Iub calculated from BTS Transport counters (average/hour):
Monitored_Iub_Load = (VS.AAL2NbReceivedCell + VS.AAL5NbReceivedCell) / 4528

The Iub Load estimated by the RNC and used by iRM CAC:
RNC_Estimated_Iub_load = ( RadioLinkEstablishedPerCell.Avg) * ECR * SrHO% (at cell level)

The ECRs are calculated from configured EBRs.

> If the EBR configuration is very conservative (AF 100% considered for all
RBs, especially the big PS RBs), the gap between the Monitored_Iub_Load
and RNC_Estimated_Iub_load will be important and the blocking will be seen
with very low Iub traffic (currently the case in ORF).
> From UA4.2, the real Iub load is taken into consideration in the iRM and it
will be possible to monitor the percentage of time a certain load level was
reached through iRM specific counters:
#1155 VS.IRMTimeCellDlIubTransportColorYellow (recommended threshold @ 70%)
#1148 VS.IRMTimeCellDlIubTransportColorRed (recommended threshold @ 90%)
32

Nortel Confidential Information

Bottleneck analysis
Iub Interface

RB Establishment Unsuccess
LackOfBandwidth will be replaced in UA4.2 by
LackOfBandwidthOnIu
LackOfBandwidthOnIur
LackOfBandwidthOnIub
And Lack of CID
LackOfCidOnIu
LackOfCidOnIur
LackOfCidOnIub

Detailed investigation method

RB Establishment Unsuccess

no

Lack of bandwidth > 0


Or

Lack of CID > 0

yes
Call allocation failure
due to Iub Lack of
bandwidth or CIDs

yes (optional)

RL Reconfig Prepare Unsuccess


INodeRefusal RB Establishment Unsuccess
LackOfBandwidth

yes
Call allocation & iRM Sched. upgrade
failure due to lack of Iub Resources

Monitored_Iub_Load ~
IRMTimeCellDlIubTransportColorRed X%
RNC_Estimated_Iub_load?

yes
Plan to add additional
PCM resources
33

RadioLink Reconfiguration Prepare Unsuccess


INodeRefusal will be completed in UA4.2 with:
LackBwthIub
LackOfCid

Stop

no

UA4.2
UA4.1

no

Transport configuration
allows EBR Tuning?
Nortel Confidential Information

- X can vary with EBR config., traffic type


(R99/HSDPA) and Y2R/R2Y Iub
thresholds
- For R99, EBR at max and Y2R/R2Y Iub =
90%/80%, X is expected to be 1%
Additional analysis is needed for HSDPA

yes

Do EBR Tuning

Agenda
> Introduction
Rational, Traffic Evolution, Call profile, Benchmark

> Capacity Monitoring method

Blocking & Resources Load Characterization


Congestion and Resources Load at network level
Top N most congested cells identification
Bottlenecks identification
Bottlenecks troubleshooting & analysis

> Capacity Monitoring Use case Method application


> Conclusions
34

Nortel Confidential Information

Capacity Monitoring Use cases


Use case 1: Low RB Setup and high blocking
Use case 2: High RB Setup and high blocking
Use case 3: High RB Setup and low Blocking

35

Nortel Confidential Information

USE CASE 1: Low RBSetup but high blocking


Summary
> Site to be analyzed : ST_ETN_GARE
> Cell : ST_ETN_GARE_U11
> Capacity Analysis:
-

One day from the Obs. period (14/03) is chosen for detailed analysis
Blocking causes identification: LackOfBandwidth & Unspecified
Bottlenecks identification: Iub & CEM
Traffic analysis:

High amount of PS traffic : 9 Erl /BH (several hours) & 166 Kbps @BH
PS384 main traffic: Setup/Request = 77 / 179 = 43% admission / day /site
PS384 converted to PS64 due to Iub blocking: 80 (44%) / day /site
PS384 lost due to Iub blocking: 22 (13%) / day /site

- Recommendations for this site:


- Tuning EBRs or add PCM
- add CEM
36

Nortel Confidential Information

USE CASE 1: Low RBSetup but high blocking


Choosing blocking day for detailed analysis
RB Blocking Rate at BH (ST_ETN_GARE_U11)
70%

Blocking (%)

50%

60

Day chosen for


detailed analysis

50

40%

40

30%

30

20%

20

10%

10

0%

Nb of RBs

60%

70

RB Blocking Rate at BH
RB Setup Request at BH
RB Blocked at BH

> Cell very blocked during the entire week (exception being the week-end)

14/03 is chosen for deeper analysis the Blocking rate is not


the highest at this date, but the cell has the highest number of
RB Setup Requests and RB Blocking.

37

Nortel Confidential Information

USE CASE 1: Low RBSetup but high blocking


Blocking causes identification
Blocking causes identification (ST_ETN_GARE_U11)
40
35

35
32

30

29
27

25

RBEstablishmentUnsuccess - total

Peaks of
Lack of Bandwidth
& Unspecified

20

RBEstablishmentUnsuccess.Unspecified
RBEstablishmentUnsuccess.LackOfBandw idth
RBEstablishmentUnsuccess.UnavailableDlCodeResources
RBEstablishmentUnsuccess.UnavailableDlPow erResources

15
13
12

12

13

10
8
5
0

4
3
1

5
4

3
3
2
2
1
1
1
1
1
0
00
0
0
0
0
0
0
0
0
0
0
0
00
0
0
00
00:00 01:00 02:00 03:00 04:00 05:00 06:00 07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00

> High rate of Lack of Bandwidth is observed:


From UA4.2 Iub/Iu/Iur specific screenings will be available.

> A second cause of blocking is Unspecified:


Additional investigation is needed to clearly identified the cause of unspecified
blocking (see next slide)

38

No Codes & Power blocking issues

Lack of Bandwidth and Unspecified blocking identified


Nortel Confidential Information

USE CASE 1: Low RBSetup but high blocking


Bottlenecks analysis
Bottlenecks analysis - "Unspecified" detailed investigation
35
31
30

28
29

27

25
RLReconfPrepareUnsuccess.RadioLinkReconfigurationFailure
20

RBEstablishmentUnsuccess.Unspecified
RLReconfPrepareUnsuccess.INodeRefusal

15

13
13

12

12

10

RBEstablishmentUnsuccess.LackOfBandw idth

14

12

VS.RadioLinkReconfigurationCancel

8
5

7
6

3
4
2
2
1
1
11
1
0
0
0
0
0
0
0
00
0
0
0
00
0
0
0
00
0
00:00 01:00 02:00 03:00 04:00 05:00 06:00 07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00
3

> RL Reconfiguration Cancel is 0, so no Core Network issues are foreseen


> RL Reconfiguration is failing in the Preparation phase:
Cause INodeRefusal is confirming the Iux lack of bandwidth (linked to
LackofBandwidth cause)
=>As the issue is not seen on the entire RNC we can conclude that Iux = Iub
Cause RadioLinkReconfigurationFailure confirms a lack of resources at NodeB
level (generally caused by CEM resources).

Iub and CEM blocking are confirmed


39

Nortel Confidential Information

USE CASE 1: Low RBSetup but high blocking


Blocked traffic type identification
Rejected Traffic vs Blocking
30

25
21
RB Setup Unsucc PS64

Nb of RBs

20
17
16

RB Setup Unsucc PS384


RB Setup Unsucc CS

15

RB Setup Unsucc PS128


10

10

11

RBEstablishmentUnsuccess.Unspecified

10

RBEstablishmentUnsuccess.LackOfBandwidth

7
6
5
2
00

00

00

00

00

00

00

00

2
0

1
0

3
2

22
00

4
2

00

00

00

00

> RB Setup Unsuccess is calculated per RB type as difference between RB Setup


Request and RB Setup Success.
> The CS and PS128 traffic are practically not blocked

40

CS is a small RB which passes even if high blocking is present at Iub and CEM level.
PS128 no blocking as very few requests are present.

The only traffic blocked is the PS384 and PS64


Nortel Confidential Information

USE CASE 1: Low RBSetup but high blocking


Blocking traffic analysis
Max RL Establsihed vs Blocking
12

35

Peak of PS384
call duration (700s)

10

9
8

30
9
25

Nb of RL

20

6
5

15

Max of 3 PS384

4
3

3
2

22
1
0

0
01:00

10
2

00:00

02:00

Max Nb of RL PS64

1
0

03:00

1
0

04:00

00

00

00

05:00

06:00

07:00

Max Nb of RL PS384

2
1

1
00

08:00

Max Nb of RL CS

09:00

10:00

11:00

Max Nb of RL PS128

12:00

13:00

14:00

15:00

16:00

17:00

18:00

RBEstablishmentUnsuccess.Unspecified

19:00

20:00

21:00

22:00

RBEstablishmentUnsuccess.LackOfBandwidth

>

The max of 3 PS384 accepted by the Iub (with the current EBR configuration) is reached
during 4 hours only in the observed cell (not counting the traffic in other cells).

>

This is combined with a high PS384 Call duration (200 s in average, 700s peak at 14h00)

>

The Iub capacity with the current


EBR config - see tables:

41

if x PS384
1
2
3

PS64 ?
15
9
3

PS128?
7
4
1

if x PS64
7
8
9
10

PS128?
7
6
6
5

PS384?
2
2
2
1

Iub limit is reached constantly during the day => blocking


Nortel Confidential Information

23:00

RB Blocked

10

USE CASE 1: Low RBSetup but high blocking


Traffic blocking consequences
PS64 Setup correlation with Iub Blocking
30

25

20

15

10

RB Setup Request PS64

RBEstablishmentUnsuccess.LackOfBandwidth

> Due to rapid blocking at Iub level (max 3 PS384 per Iub with 1 PCM), a
RB Reallocation mechanism is triggered at Core PS level (see details in
next slide). Due to this mechanism, a PS384 failed setup is replaced by a
PS64 RB request.

This mechanism is allocating a lower PS RAB (PS64) maintained


during entire call duration (not eligible to any iRM Upgrade) => user
perception: Good for call admission but Bad for service quality
42

Nortel Confidential Information

USE CASE 1: Low RBSetup but high blocking


RequestedUEMaximum
Bit Rate not Available
(20)
Node B
RNC

CN - PS

PS call initial connection (RRC phase)


PS call RAB allocation phase starts with PS384
RAB Assignment Request (DL384/UL64)

No Ressources
RAB Assignment Response (Fail: 20)
RAB Assignment Request (DL64/UL64)

PS call RAB allocation phase continue with PS64

> If no resources are available at RNC level, the RB Assignment Request


(for PS384 or PS128) is answered with a RAB Assignment Response
(Fail: 20) which is immediately followed by a new request for a lower RAB
(instead of Iu Release Command)
to be confirmed if it works only with Iub or all other resources managed by
RNC (Codes & Power)
It not works with the CEM resources which are managed by the BTS.

Using this Core PS mechanism is helpful to reduce Iub blocking,


but it will work against RB adaptation feature (UA4.2) allocated
RB will not be eligible to any upgrade.
43

Nortel Confidential Information

USE CASE 1: Low RBSetup but high blocking


Load details all resources
Load per Resource type
100%
90%

14/03

80%

Load (%)

70%

iRM Green2Yellow limit

60%
50%
40%
30%
20%
10%
0%

Pow er load Max

Codes Load Max

CEM Used Max

IUB Load

> Iub load curve represents the average Iub usage during each hour.
> Power, Codes and CEM Load curves present the Max load.

Power & Codes load is below iRM Green2Yellow threshold (50%)

CEM load is significantly high explaining the CEM blocking

IuB load is low but the Iub blocking is present > to be investigated
44

Nortel Confidential Information

USE CASE 1: Low RBSetup but high blocking


CEM Blocking vs. load (week11)
CEM Load vs CEM Blocking (w eek)

load (%)

90%

10

14/03
PS Blocking limit

80%

70%

60%

50%

40%

30%

20%

10%

0%

RB Estab Blocked by CEM

CEM Load Max (%)

RB Blocked

100%

CEM Used Avg (%)

> CEM Blocking is observed only when the PS384 & PS64 blocking limit is reached
this happens constantly during the loaded period of the week (working days)
> CEMAllocFail was < 1% during all this period as the number of CEM resources
request was much higher that the refusal > not a good indicator for CEM blocking

CEM Blocking limit is constantly reached during working days

45

Recommendation => plan to add CEM


Nortel Confidential Information

USE CASE 1: Low RBSetup but high blocking


Iub Blocking vs. load (week11)
Iub traffic: Real vs Estimated

3500

35

14/03

Peak of
PS384 Setup
on Cell U21

3000

30
25

cells/s

2500
20
2000
15
1500

Mostly
CS traffic

1000

10

500

RB Estab on current cell Blocked by Iub

ECR Estimated

RB Blocked

4000

ECR realy passed

> Important gap is seen between the real ECR and the ECR estimated by the RNC
> Configuring EBRs at AF 100% for the PS traffic (especially PS384) is creating
important false blocking.
> If the EBR cannot be reduced (transmission architecture constraints), additional
PCM is to be added on Iub.

EBRs for PS are overestimated possible parameters tuning


46

Alternative: additional PCMs

Nortel Confidential Information

Agenda
> Introduction
Rational, Traffic Evolution, Call profile, Benchmark

> Capacity Monitoring method

Blocking & Resources Load Characterization


Congestion and Resources Load at network level
Top N most congested cells identification
Bottlenecks identification
Bottlenecks troubleshooting & analysis

> Capacity Monitoring Use case Method application


> Conclusions
47

Nortel Confidential Information

Conclusions
> Very high blocking was observed during the study (w10-w11). Important
part was coming from non-UTRAN causes.
> UTRAN Blocking is mainly due to Iub lack of bandwidth and CEM
resources.
> No blocking coming from OVSF Codes or RF Power was monitored.
> Resources load seems to be below the alarm limit for almost all BTS
(some exceptions for CEM). Planning exercise could be foreseen to
prepare for future traffic growth.
> Real load at Iub level is significantly lower than Estimated load due to
very conservative EBR Values:
EBR Values
DL RB
PS 64
PS 128
PS 384

ORF
69632
136832
408000

Operator 1 Operator 2
x 2/3
x 3/5
x 1/2

x1
x 3/4
x 1/2

> Consequence of Iub "virtual load" and Blocking is an iRM like mechanism
triggered at Core PS level forcing to lower PS RABs. In UA4.2 this system
behavior will negatively impact RB Adaptation feature.
48

Nortel Confidential Information

Thanks for your patience!

49

Nortel Confidential Information

50

Nortel Confidential Information

ORF Capacity Monitoring Study


- Detailed presentation
Wireless Network Engineering
CTF/W20

51

Nortel Confidential Information

Agenda
> Introduction & Executive summary
> Capacity Monitoring method

Blocking & Resources Load Characterization


Congestion and Resources Load at network level
Top N most congested cells identification
Bottlenecks identification
Bottlenecks troubleshooting & analysis

> Capacity Monitoring Use cases


> iRM & Benchmark

52

Nortel Confidential Information

UTRAN capacity analysis introduction


> Objective:
Networks have been deployed focusing on the service availability at the expense
of capacity
Traffic is growing constantly :
Traffic evolution during last 6 months
200000000
180000000
160000000
140000000
120000000
100000000
80000000
60000000

CS DL traffic KBytes

PS DL Traffic KBytes

15 A pr 06

8 A pr 06

1 A pr 06

2 5 -m a rs -0 6

1 8 -m a rs -0 6

1 1 -m a rs -0 6

0 4 -m a rs -0 6

25 Feb 06

18 F eb 06

11 F eb 06

4 Feb 06

2 8 -ja n v -0 6

2 1 -ja n v -0 6

1 4 -ja n v -0 6

31 Dec 05

24 D ec 05

17 Dec 05

10 Dec 05

3 Dec 05

2 6 -n o v -0 5

1 9 -n o v -0 5

1 2 -n o v -0 5

0 5 -n o v -0 5

2 9 -o c t -0 5

2 2 -o c t -0 5

1 5 -o c t -0 5

0 8 -o c t -0 5

0 1 -o c t -0 5

20000000

0 7 -ja n v -0 6

40000000

total PS + CS

Operators start to experience some congestion in their networks and foresee


further capacity issues.
They expect capacity analysis guidelines in order to be ready for capacity
planning.

> Scope:
This presentation has been prepared in the scope of UA4.1 pointing out also
possible UA4.2 evolutions (when the case)
53

Nortel Confidential Information

UTRAN capacity analysis introduction


> Study coverage
The capacity assessment is done for a period of two weeks: w10 &
w11
All the cells of the networks were analyzed (5495 cells)
All the RNCs (19 RNCs)

> Software release


Nortel OAM Release OAM5.0
Nortel UTRAN Release UA4.1 (17 RNCs) and UA4.2 (Lyonecul1 &
Lyontrio1)

> Elements of NodeB configuration:


Multi-carrier BTS : only some of the VO sites

54

X=1

X=2

X=3

X=4

BTS with X CEM

2058

28

BTS with X PCM

2060

29

Nortel Confidential Information

Executive summary
> Very high blocking was observed during the study (w10-w11).
Important part was coming from non-UTRAN causes.
> UTRAN Blocking is mainly due to Iub lack of bandwidth and CEM
resources.
> Real load at Iub level is significantly lower than Estimated load due
to very conservative EBR Values.
> Consequence of Iub "virtual load" and Blocking is an iRM like
mechanism triggered at Core PS level forcing to lower PS RABs. In
UA4.2 this system behavior will negatively impact RB Adaptation
feature.
> No blocking coming from OVSF Codes or RF Power was monitored.
> Due to low load and blocking of Iub resources, the iRM @ CAC is
triggered only due to RL color.
> Reduction of AON T1 could be foreseen in order to gain extra
capacity
55

Nortel Confidential Information

Agenda
> Introduction & Executive summary
> Capacity Monitoring method

Blocking & Resources Load Characterization


Congestion and Resources Load at network level
Top N most congested cells identification
Bottlenecks identification
Bottlenecks troubleshooting & analysis

> Capacity Monitoring Use cases


> iRM & Benchmark

56

Nortel Confidential Information

UTRAN congestion detection principles


> Principle of congestion
detection
Combination of blocking + load
indicators for a specific resource

Blocking?

> 3 states

=0

>0

No blocking No action
Blocking, low load Capacity
analysis and tuning required
Blocking + High load Capacity
analysis and tuning required
and/or resource addition

Load

Low

Status Yellow
Capacity analysis/tuning

High
Status Red
Capacity analysis/tuning
and/or
Resources addition

57

Status Green
No action required

Nortel Confidential Information

Scope of
package

Blocking Characterization
> Blocking situation definition : each time the UE requests an UTRAN
resource and is not getting it because the resource is not available.
Call Phase

Blocking Cause

Effect

Monitoring detection
method

Call Admission

Lack of resources at
call setup

Call admission failure

- RB Establishment Unsuccess

Lack of resources to
perform iRM
transitions (AON
Upsize, iRM Sched
Upgrade)

Call is not reconfigured - RL Reconfiguration Unsuccess


(iRM Reconf)
(impact on user
throughput)
- RL Setup Unsuccess (AON)

Call
Reconfiguration

Mobility

- RL Reconfiguration Unsuccess
and Cancel

- AON/iRM Scheduling failure


counters

- RL Setup Unsuccess
No resources
Additional RL not
available for additional added in the Active Set - RL Addition Unsuccess
RL
(risk of call drop)

- RL & RB Reconfiguration
Unssuccess for HSDPA

This study is mainly focused on Blocking during Call Admission phase as it


is the most impacting for call success.
In general, if the call admission blocking is solved, the blocking on other call
phases is also solved (same bottlenecks involved).
58

Nortel Confidential Information

Resources Load
Main possible bottleneck resources monitored at UTRAN Level:
> BTS Channel Elements - BTS Resource

The load can be monitored at BTS level


Blocking of this resource RB rejection or RL Setup/Addition/Reconfiguration
failures

> Iub ATM

The load can be monitored at BTS level (AAL2&AAL5 traffic)


alternative solution monitor the Traffic load per BTS at RNC level
Blocking of this resource RB rejection

> RF power - BTS Resource

The load can be monitored per cell via BTS or RNC counters
Blocking of this resource RB rejection

> OVSF Codes - RNC Resource

The load can be monitored per cell via RNC counters


Blocking of this resource RB rejection

> RNC CPU

59

The load can be monitored at card level through RNC counters


Blocking of this resource Overload mechanism => RB rejection
Nortel Confidential Information

Agenda
> Introduction & Executive summary
> Capacity Monitoring method

Blocking & Resources Load Characterization


Congestion and Resources Load at network level
Top N most congested cells identification
Bottlenecks identification
Bottlenecks troubleshooting & analysis

> Capacity Monitoring Use cases


> iRM & Benchmark

60

Nortel Confidential Information

Congestion and Load on the Network


> Congestion at network level is evaluated through Blocking
Rate distribution per cell.
Blocking Rate per cell is computed as a ratio between the RB Blocking at
BH and RB Setup Request at BH for the entire observation period.
The BH is based on the most loaded Hour of the day in terms of Radio
Bearer Setup Request and it is calculated for each cell, each day. It can
be different for each cell, each day.

> Load is analyzed by bottleneck type at the BH of each Cell:


CEM Load through Max of CEMUsed.Avg&Max per cell at BH.
Iub Load through Max of Iub Load DL metric calculated per day
DL Power Load trough Tx Carrier Power Avg & Max at BH divided
by PA Power available@ref_point (calculated)
Codes load through Average of Free codes SF128 Min at BH metric
TMU load through Average and Maximum TMU Load for the entire
period (for each RNC)

> All metrics used will be available in a metric library to be


delivered with the study.
61

Nortel Confidential Information

Congested cells @ Network level


Global Blocking
Blocking Rate at cell BH Distribution
4989

100%

4500

90%

4000

80%

3500

70%

2807

3000
2500

60%
50%

2102

2000

40%

1500

% of cells

Nb of Cells concerned

5000

30%

1083

1000

20%

273

500

37

>7%

>10%

>20%

>30%

0
>=0%

>0%

>0,5%

>2%

>5%

10%

104

0%

Blocking rate > x %


Blocked cells Count

Blocked cells %

> Main contributor to this high blocking level was a Core CS dimensioning issue
(MSC Trion) which affected almost all Nortel RNCs.
> The rest is coming from Iub congestion.

62

Significant amount of blocking (> 2%) observed during w10 &


w11 at Cell BH for 22% of the cells
Nortel Confidential Information

CEM load @ Network level


CEM Load Distribution (Avg & Max @ BH)

4979
4757

4757

4500

100%
90%

4144

Nb of Cells concerned

4000

80%

3482

3500

70%

3000

60%

2717

2500

50%

2000

40%

1500

30%

1060
765

1000
500

20%

116

15

279
4

0
>10%

>20%

>30%

% of cells

5000

>40%

>50%

>60%

64

>70%

17

10%

>80%

>90%

0%

>95%

CEMUsed > x %
Max Load Count

Avg Load Count

Max Load (%)

Avg Load (%)

> During the Observation period, most part of the CEMs were loaded in average
less than 50% (the attention limit)
> When investigating the Max CEMUsed for the same period, we can see that 64
cells (and the associated CEMs) exceeded the 70% load which represents the
upper blocking limit of PS384.

63

37 BTS (64 cells) are reaching the dangerous zone of the CEM
Nortel Confidential Information

Iub load @ Network level


Iub real Load Distribution (Max@BH)
100%

4484

4500

90%

4000

80%

3500

70%

3016

3000

60%

2500

50%

2000

40%

1500

% of cells

Nb of Cells concerned

5000

30%

1074

1000

20%

311

500

112

48

18

>20%

>25%

>30%

>35%

40%

10%
0%

>2%

>5%

>10%

>15%

Iub Load > x %


Max Load Count

Max Load (%)

> As it is based on a BTS counter, the monitored Iub load (@ BH) has to be seen
as the average load per hour. In the graph, the Max of these averages (for the
observation period) is captured.

Iub utilization is very low, not exceeding 35% of the Iub bandwidth.
This is not preventing from having Iub blocking as the admission
criteria is based on the estimated Iub load (at the RNC)
64

Nortel Confidential Information

DL Power load @ Network level


DL RF Power Distribution (Avg & Max)
49574954

4780
4390

4500

100%

4326

90%

Nb of Cells concerned

4000

80%

3500

70%

3080

3000

60%

2503

2500

50%

2000

40%

1486

1500

30%

1000

283

500

% of cells

5000

20%

514
83

26

151

12

45

10%
0%

>10%

>20%

>25%

>30%

>35%

>40%

>45%

>50%

>60%

>70%

>80%

DL Pow er Load > x %


Max Load Count

Avg Load Count

Max Load (%)

Avg Load (%)

> During the Observation period, most part of the cells were loaded in average less
than 50%
> When investigating the Max load for the same period, we can see that only 1 cell
punctually exceeded 80% (Yellow2Red iRM power threshold).

DL power load is relatively low in the network. The loaded sites


being mostly sites from VO zone (JONAGE_COTTAGE, MEYZIEU_CARREAU,
MEYZIEU_GD_LARGE, MEYZIEU_OUEST) which use PA/2 (due to STSR2)
65

Nortel Confidential Information

Codes load @ Network level


OVSF Codes Load Distribution (Avg & Max @ BH)
4933

4797

4873

100%

4500

90%

3746

Nb of Cells concerned

4000

80%

3377

3500

70%

3000

60%

2500

50%

2000

40%

1352

1500

30%

745

1000

20%

459

500

% of cells

5000

93

54

10

12

10%

0%
>5%

>10%

>20%

>30%

>40%

>50%

>60%

>70%

>80%

Codes Load > x %


Max Load Count

Avg Load Count

Max Load (%)

Avg Load (%)

> During the Observation period, all the cells were loaded in average less than 50%
> When investigating the Max load for the same period, we can see that only 2 cells
punctually exceeded 70% (Yellow2Red iRM codes threshold).

66

Low codes load => only two cells being able to reach a max load
leading to a red iRM cell color.
Nortel Confidential Information

RNC CPU load @ Network level


TMU load Avg & Max
100
90
80

TM U Load (%)

70
GRENOBLE1 Avg

60

GRENOBLE1 Max
LYONECUL1 Avg

50

LYONECUL1 Max
MONTPELL1 Avg

40

MONTPELL1 Max

30
20
10
0
13/3/ 13/3/ 13/3/ 13/3/ 14/3/ 14/3/ 14/3/ 14/3/ 15/3/ 15/3/ 15/3/ 15/3/ 16/3/ 16/3/ 16/3/ 16/3/ 17/3/ 17/3/ 17/3/ 17/3/ 18/3/
06
06
06
06
06
06
06
06
06
06
06
06
06
06
06
06
06
06
06
06
06
0:00 6:00 12:00 18:00 0:00 6:00 12:00 18:00 0:00 6:00 12:00 18:00 0:00 6:00 12:00 18:00 0:00 6:00 12:00 18:00 0:00

> Only 3 RNCs are captured in this graph (3 most loaded from TMU point of view)
> For these 3 RNCs, the TMU load is remaining in acceptable limits:

Average < 20%


Max < 35% (the only exception is Montpell1 with some local issues on 15-03-06)

TMU load is far from being a concern for ORF network


67

Nortel Confidential Information

Agenda
> Introduction & Executive summary
> Capacity Monitoring method

Blocking & Resources Load Characterization


Congestion and Resources Load at network level
Top N most congested cells identification
Bottlenecks identification
Bottlenecks troubleshooting & analysis

> Capacity Monitoring Use cases


> iRM & Benchmark

68

Nortel Confidential Information

Top N most congested cells identification


> Counters Granularity: Hour
For each Cell : 4 metrics calculation
RbRejected@BH where BH=Hour[Max(RbSetReq)]
days

RbRejected@BkH where BkH=Hour[Max(RbRejected)]


days
RbRejected
days hour
BHBKRate(%) = RbRejected / RbSetReq (@BH)
Condition1= NbrDays[BHBKRate(%)> 0%]
Condition2= NbrDays[BHBKRate(%)> 2%]
N.B RbRejected corresponds to counter VS.RadioBearerEstablishmentUnsucces (without screening 0)
RbSetReq corresponds to counter VS.RadioBearerSetupRequest

Method aims to create a list of most representative cells in terms of


blocking by taking into account the amount of rejections at different
periods of the day but also the persistence in time of the issue.
69

Nortel Confidential Information

Top N most congested cells identification


CellsRanking
[ RbRejected]
day hour

CellsRanking
[RbRejected@BH]

Cells Filtering
[Condition1>X & Condition2>Y ]
X>Y

CellsRanking
[RbRejected@BkH]

Bottleneck Identification
[RbRejected Screening]

In our Study:
Obs.Period =14 days
X=7
Y=4
70

Final cells List to investigate


Nortel Confidential Information

Top N most congested cells identification


> Advantages:
Ranking is helping to focus on cells with important amount of traffic
and blocking (putting in lower priority low loaded cells or low blocked
cells) this step is not mandatory in case of a day bay day monitoring
the temporal factor is considered during the filtering only cells with
constant blocking are picked up
The double filter, even it is not mandatory (simple filter could be
enough for most of the cases), add to detect cells exceeding a given
threshold (condition 1) and also cells with constant load (condition 2)

> Drawback:
only blocking at BH global (CS+PS) is considered risk to focus loose
one blocking period if the BH PS BH CS
=> The method can be improved by analyzing two different BH (CS&PS)

71

Nortel Confidential Information

Top N most congested cells


identification - Result
Cell
GRE_ALSACE_LORRAINE_U21
DJN_HAVRE_U11
BOURGOIN_EST_U11
BIONNE_TDFT_U31
ST_ETN_GARE_U11
GRE_EUROPOLE_U11
FERREOL_U31
ECHIROLLES_U11
FERREOL_U11
GRE_JEAN_JAURES_U11
HYERES_COUBERTIN_U11
BEAUREGARD_U21
NIMES_MONT_DUPLAN_U21
MARIGNANE_CENTRE_U21
ECHIROLLES_U31
AIX_CARCASSONNE_U21
FONTAINE_CHARM_U21
GRE_EUROPOLE_U21
HAUTS_VOREPPE_TDF_U21
DJN_CFRT_U11
DJN_HAVRE_U21
GRE_JEAN_JAURES_U31
TOULON_VALETTE_SUD_U21
RABATEAU_U31
GRE_AMPERE_DRAC_U11
SALON_SUD_U31
BEAUREGARD_U31
ALLAUCH_U31
ISTRES_LES_COGNETS_U11
VIENNE_PIPET_U11
GARIBALDI_U11
MARSEILLE_PARTY_U21
FONTAINE_OUCHE_U31
TERRON_U21
ST_EGREVE_BONNAIS_U11
GRE_EAUX_CLAIR_U31
MRS_MONTE_CRISTO_TDFT_U21
BIONNE_TDFT_U21
MARIGNANE_CENTRE_U11
GRE_AMPERE_DRAC_U21
SEYSSINET_NORD_U11
TOUR_PIN_U31
CROLLES_U11
TRAMOYES_U31
CORDERIE_U11
DJN_MARCS_OR_U21
LES_CHARTREUX_U21
ST_GELY_DU_FESC_TDFT_U31
DJN_VICTOR_HUGO_U11
GRE_ABBAYE_U31
MAUGUIO_U31
ST_JEROME_U31
AIGUELONGUE_U11
DJN_HAVRE_U31

72

BK Day

Setup Day

Bk rate day

Bk & BH

Setup & BH

Bk rate BH

Bk & BkH

Setup & BkH

Bk rate BhK

796
515
279
338
504
297
317
245
267
247
279
179
249
172
193
310
144
161
145
161
215
122
145
133
135
97
114
107
101
113
103
99
95
128
98
107
112
122
83
85
84
78
77
140
79
70
99
61
74
69
64
58
58
85

75504
23104
21612
27698
1962
20546
20132
17564
17850
20098
7600
12697
5761
13129
14707
4504
11916
13154
9049
14210
10344
8082
8726
10587
9464
9278
5931
8335
10379
9373
7631
8565
9542
995
6121
7513
8915
10167
7014
8765
6999
5121
5500
3911
3923
8143
3248
8044
8775
6977
4713
3889
5308
3788

1,1%
2,2%
1,3%
1,2%
25,7%
1,4%
1,6%
1,4%
1,5%
1,2%
3,7%
1,4%
4,3%
1,3%
1,3%
6,9%
1,2%
1,2%
1,6%
1,1%
2,1%
1,5%
1,7%
1,3%
1,4%
1,0%
1,9%
1,3%
1,0%
1,2%
1,3%
1,2%
1,0%
12,9%
1,6%
1,4%
1,3%
1,2%
1,2%
1,0%
1,2%
1,5%
1,4%
3,6%
2,0%
0,9%
3,0%
0,8%
0,8%
1,0%
1,4%
1,5%
1,1%
2,2%

230
142
188
124
143
101
81
105
80
86
86
103
76
100
76
70
74
84
59
58
52
73
55
58
48
62
52
52
49
54
60
51
51
43
60
43
41
40
38
44
35
37
47
20
33
35
25
29
32
37
41
29
36
21

6126
2382
2313
2797
370
2483
2054
2036
1949
2213
818
1369
713
1378
1690
739
1393
1524
1034
1457
1101
1029
1035
1107
1118
1056
730
1052
1179
1063
844
998
1506
263
724
1106
971
1191
1023
1040
758
594
644
515
609
959
453
821
993
788
578
547
621
484

3,8%
6,0%
8,1%
4,4%
38,6%
4,1%
3,9%
5,2%
4,1%
3,9%
10,5%
7,5%
10,7%
7,3%
4,5%
9,5%
5,3%
5,5%
5,7%
4,0%
4,7%
7,1%
5,3%
5,2%
4,3%
5,9%
7,1%
4,9%
4,2%
5,1%
7,1%
5,1%
3,4%
16,3%
8,3%
3,9%
4,2%
3,4%
3,7%
4,2%
4,6%
6,2%
7,3%
3,9%
5,4%
3,6%
5,5%
3,5%
3,2%
4,7%
7,1%
5,3%
5,8%
4,3%

516
227
227
224
158
193
192
185
164
151
117
124
129
125
124
112
143
111
107
100
85
92
92
93
94
96
82
81
88
76
76
81
82
72
71
77
67
65
88
58
62
62
53
64
57
62
55
77
54
50
50
61
51
41

4638
2045
1909
2352
314
1782
1507
1559
1519
1740
651
954
539
1124
1265
431
1040
1045
843
1062
897
690
809
838
627
672
532
858
727
804
619
720
747
157
507
633
710
786
572
709
542
394
357
275
430
693
271
514
655
443
433
321
457
334

11,1%
11,1%
11,9%
9,5%
50,3%
10,8%
12,7%
11,9%
10,8%
8,7%
18,0%
13,0%
23,9%
11,1%
9,8%
26,0%
13,8%
10,6%
12,7%
9,4%
9,5%
13,3%
11,4%
11,1%
15,0%
14,3%
15,4%
9,4%
12,1%
9,5%
12,3%
11,3%
11,0%
45,9%
14,0%
12,2%
9,4%
8,3%
15,4%
8,2%
11,4%
15,7%
14,8%
23,3%
13,3%
8,9%
20,3%
15,0%
8,2%
11,3%
11,5%
19,0%
11,2%
12,3%

Nortel Confidential Information

rank
1
3
6
6
7
9
12
12
16
17
21
22
22
23
25
26
31
31
43
44
52
52
53
55
61
70
74
78
79
81
83
86
87
89
91
92
105
106
115
140
154
159
168
173
177
177
185
189
195
208
211
224
234
272

days with
Bk>2%
4
7
7
5
8
4
4
6
5
5
8
7
8
6
5
11
6
6
5
4
7
6
6
6
7
5
7
7
7
5
7
5
6
11
7
6
6
6
5
6
5
7
7
7
7
6
8
6
4
7
9
6
7
6

days with
Bk>0%
7
8
9
10
8
7
7
7
7
8
8
9
9
9
7
11
7
7
8
7
9
7
7
8
9
7
8
9
7
7
7
8
7
11
7
7
9
7
7
9
7
7
7
7
7
7
8
7
7
7
9
7
7
7

Two filter criteria are


used:
- BK > 2 % for at least 4
days form observation
period
- BK > 0 % for at least 7
days form observation
period
The rank is calculated
following Step 1

54 regularly
blocked
cells with
high impact
on user
perception

Bottlenecks identification
The different blocking causes to be investigated are identified
through VS.RadioBearerEstablishmentUnsuccess screenings)

Scr. 0: invalid RAB parameters value not to be considered as blocking


cause, thus not taken into account in the blocking rate computation
Scr. 1: unavailable dl code resources Code shortage indication
Scr. 2: unavailable dl power resources DL power shortage indication
Scr. 3: Unspecified covers CEM resources and other possible issues during
call setup (Core Network UP Init. failures...)
Scr. 4: Lack of bandwidth Iux bandwidth allocation shortage (all interfaces:
Iub/Iu/Iux) this screening is replaced by screenings #10, #12 and #14 in
UA04.2
Scr. 5: Lack of CID Iux CIDs allocation shortage (all interfaces: Iub/Iu/Iux)
this screening is replaced by screenings #9, #11 and #13 in UA04.2
Scr. 6: CAC RNC Processing resources TMU resources shortage (overload
situations)
Scr. 9: Lack of CID on the Iu introduced in UA04.2
Scr. 10: Lack of bandwidth on the Iu introduced in UA04.2
Scr. 11: Lack of CID on the Iur introduced in UA04.2
Scr. 12: Lack of bandwidth on the Iur introduced in UA04.2
Scr. 13: Lack of CID on the Iub introduced in UA04.2
Scr. 14: Lack of bandwidth on the Iub introduced in UA04.

73

Nortel Confidential Information

Bottlenecks identification - Result


Cell
GRE_ALSACE_LORRAINE_U21
DJN_HAVRE_U11
BOURGOIN_EST_U11
BIONNE_TDFT_U31
ST_ETN_GARE_U11
GRE_EUROPOLE_U11
FERREOL_U31
ECHIROLLES_U11
FERREOL_U11
GRE_JEAN_JAURES_U11
HYERES_COUBERTIN_U11
BEAUREGARD_U21
NIMES_MONT_DUPLAN_U21
MARIGNANE_CENTRE_U21
ECHIROLLES_U31
AIX_CARCASSONNE_U21
FONTAINE_CHARM_U21
GRE_EUROPOLE_U21
HAUTS_VOREPPE_TDF_U21
DJN_CFRT_U11
DJN_HAVRE_U21
GRE_JEAN_JAURES_U31
TOULON_VALETTE_SUD_U21
RABATEAU_U31
GRE_AMPERE_DRAC_U11
SALON_SUD_U31
BEAUREGARD_U31
ALLAUCH_U31
ISTRES_LES_COGNETS_U11
VIENNE_PIPET_U11
GARIBALDI_U11
MARSEILLE_PARTY_U21
FONTAINE_OUCHE_U31
TERRON_U21
ST_EGREVE_BONNAIS_U11
GRE_EAUX_CLAIR_U31
MRS_MONTE_CRISTO_TDFT_U21
BIONNE_TDFT_U21
MARIGNANE_CENTRE_U11
GRE_AMPERE_DRAC_U21
SEYSSINET_NORD_U11
TOUR_PIN_U31
CROLLES_U11
TRAMOYES_U31
CORDERIE_U11
DJN_MARCS_OR_U21
LES_CHARTREUX_U21
ST_GELY_DU_FESC_TDFT_U31
DJN_VICTOR_HUGO_U11
GRE_ABBAYE_U31
MAUGUIO_U31
ST_JEROME_U31
AIGUELONGUE_U11
DJN_HAVRE_U31

74

BK Day

Setup Day

Bk rate day

Bk & BH

Setup & BH

Bk rate BH

796
515
279
338
504
297
317
245
267
247
279
179
249
172
193
310
144
161
145
161
215
122
145
133
135
97
114
107
101
113
103
99
95
128
98
107
112
122
83
85
84
78
77
140
79
70
99
61
74
69
64
58
58
85

75504
23104
21612
27698
1962
20546
20132
17564
17850
20098
7600
12697
5761
13129
14707
4504
11916
13154
9049
14210
10344
8082
8726
10587
9464
9278
5931
8335
10379
9373
7631
8565
9542
995
6121
7513
8915
10167
7014
8765
6999
5121
5500
3911
3923
8143
3248
8044
8775
6977
4713
3889
5308
3788

1,1%
2,2%
1,3%
1,2%
25,7%
1,4%
1,6%
1,4%
1,5%
1,2%
3,7%
1,4%
4,3%
1,3%
1,3%
6,9%
1,2%
1,2%
1,6%
1,1%
2,1%
1,5%
1,7%
1,3%
1,4%
1,0%
1,9%
1,3%
1,0%
1,2%
1,3%
1,2%
1,0%
12,9%
1,6%
1,4%
1,3%
1,2%
1,2%
1,0%
1,2%
1,5%
1,4%
3,6%
2,0%
0,9%
3,0%
0,8%
0,8%
1,0%
1,4%
1,5%
1,1%
2,2%

230
142
188
124
143
101
81
105
80
86
86
103
76
100
76
70
74
84
59
58
52
73
55
58
48
62
52
52
49
54
60
51
51
43
60
43
41
40
38
44
35
37
47
20
33
35
25
29
32
37
41
29
36
21

6126
2382
2313
2797
370
2483
2054
2036
1949
2213
818
1369
713
1378
1690
739
1393
1524
1034
1457
1101
1029
1035
1107
1118
1056
730
1052
1179
1063
844
998
1506
263
724
1106
971
1191
1023
1040
758
594
644
515
609
959
453
821
993
788
578
547
621
484

3,8%
6,0%
8,1%
4,4%
38,6%
4,1%
3,9%
5,2%
4,1%
3,9%
10,5%
7,5%
10,7%
7,3%
4,5%
9,5%
5,3%
5,5%
5,7%
4,0%
4,7%
7,1%
5,3%
5,2%
4,3%
5,9%
7,1%
4,9%
4,2%
5,1%
7,1%
5,1%
3,4%
16,3%
8,3%
3,9%
4,2%
3,4%
3,7%
4,2%
4,6%
6,2%
7,3%
3,9%
5,4%
3,6%
5,5%
3,5%
3,2%
4,7%
7,1%
5,3%
5,8%
4,3%

rank
1
3
6
6
7
9
12
12
16
17
21
22
22
23
25
26
31
31
43
44
52
52
53
55
61
70
74
78
79
81
83
86
87
89
91
92
105
106
115
140
154
159
168
173
177
177
185
189
195
208
211
224
234
272

Blocking Cause
Unspecified
LackOfBandwidth & Unspecified
Unspecified
Unspecified
LackOfBandwidth
LackOfBandwidth & Unspecified
LackOfBandwidth & Unspecified
LackOfBandwidth & Unspecified
LackOfBandwidth & Unspecified
Unspecified
LackOfBandwidth
LackOfBandwidth & Unspecified
LackOfBandwidth
Unspecified
Unspecified
LackOfBandwidth
Unspecified
LackOfBandwidth & Unspecified
Unspecified
Unspecified
LackOfBandwidth & Unspecified
Unspecified
LackOfBandwidth & Unspecified
Unspecified
Unspecified
Unspecified
LackOfBandwidth & Unspecified
Unspecified
Unspecified
LackOfBandwidth & Unspecified
Unspecified
Unspecified
Unspecified
LackOfBandwidth
Unspecified
LackOfBandwidth
Unspecified
Unspecified
Unspecified
LackOfBandwidth & Unspecified
Unspecified
Unspecified
Unspecified
Unspecified
Unspecified
Unspecified
Unspecified
Unspecified
LackOfBandwidth & Unspecified
Unspecified
Unspecified
Unspecified
Unspecified
LackOfBandwidth & Unspecified

Nortel Confidential Information

days with
Bk>2%
4
7
7
5
8
4
4
6
5
5
8
7
8
6
5
11
6
6
5
4
7
6
6
6
7
5
7
7
7
5
7
5
6
11
7
6
6
6
5
6
5
7
7
7
7
6
8
6
4
7
9
6
7
6

days with
Bk>0%
7
8
9
10
8
7
7
7
7
8
8
9
9
9
7
11
7
7
8
7
9
7
7
8
9
7
8
9
7
7
7
8
7
11
7
7
9
7
7
9
7
7
7
7
7
7
8
7
7
7
9
7
7
7

Main observations:
- Practically no DL
Power or Codes
blocking monitored
during Obs. Period.
(this is specific to the
entire network, not
only TopN list)
- Unspecified &
LackOfBandwidth are
the only blocking
causes monitored in
the network.

Agenda
> Introduction & Executive summary
> Capacity Monitoring method

Blocking & Resources Load Characterization


Congestion and Resources Load at network level
Top N most congested cells identification
Bottlenecks identification
Bottlenecks troubleshooting & analysis

> Capacity Monitoring Use cases


> iRM & Benchmark

75

Nortel Confidential Information

Bottleneck analysis
RB Allocation Procedure (1/2)
UE

Node B

RNC

CN - CS
RAB Assignment Request

1
2

CAC
Radio Link Reconfiguration Prepare
Radio Link Reconfiguration Ready

4
5

UP / DL Synchronization
UP / UL Synchronization

AAL2 / ERQ
AAL2 / ECF
UP / Initialization
UP / Initialization Ack.

Radio Link Reconfiguration Commit

Radio Bearer Setup


Radio Bearer Setup Complete
8

76

RAB Assignment Response

Nortel Confidential Information

Bottleneck analysis
RB Allocation Procedure (2/2)
1

#0625 - RB Setup Request (Downlink Access Stratum)

#0631.0 - RB Establishment Unsuccess cause Invalid RAB parameters value

#0631 - RB Establishment Unsuccess


1 - Unavailable code resource
2 - Unavailable Power resource
4 - Lack of bandwidth (replaced by screenings #10, #12 and #14 in UA04.2)
5 - Lack of CID (replaced by screenings #9, #11 and #13 in UA04.2)
6 - RNC Processing resources
9 - Lack of CID on IU (introduced in UA04.2)
10 - Lack of bandwidth on IU (introduced in UA04.2)
11 - Lack of CID on IUR (introduced in UA04.2)
12 - Lack of bandwidth on IUR (introduced in UA04.2)
13 - Lack of CID on IUB (introduced in UA04.2)
14 - Lack of bandwidth on IUB (introduced in UA04.2)

#0023 Radio Link Reconfiguration Request (Downlink Access Stratum) (introduced in UA04.2)
5 #0007 - Radio Link Reconfiguration Success (Downlink Access Stratum)
#0008 - Radio Link Reconfiguration Prepare Unsuccess
0 - Reception of RL Reconfiguration Failure
1 - Time-out
2 - Rrm refusal (ie: no more power available, no more code available, ...)
3 - I-Node refusal (replaced by screenings #6, #7 and #8 in UA04.2)
6 - Lack of CID on the Iub (introduced in UA04.2)
7 - Lack of bandwidth on the Iub (introduced in UA04.2)
8 - I-Node refusal (ie: other I-Node resource not available ...) (introduced in UA04.2)

77

#0631.3 - RB Establishment Unsuccess cause Unspecified

#0010 - Radio Link Reconfiguration Commit (Downlink Access Stratum)


#0009 - Radio Link Reconfiguration Cancel (Downlink Access Stratum)

#0601 - Radio Bearer Setup Success (Downlink Access Stratum)


#0602 - Radio Bearer Setup Unuccess (Failure cause)
Nortel Confidential Information

Bottleneck analysis
Unspecified
Blocking detection

> The Unspecified RB Establishment Unsuccess is not always directly


linked to a specific UTRAN bottleneck, thus the association with the load
is not obvious for this blocking cause.
> This blocking cause is identified through screening Unspecified of the
VS.RadioBearerEstablishmentUnsuccess counter - See bullet 6 in the RB
Allocation Procedure Data Flow Diagram.
> It covers the following Call Setup failures causes:
CEM resources shortage
User Plane Initialization failure between RNC and Core Network (Iu issues).
From UA4.2 the HSDPA CAC rejected calls are also pegging this counter.

> In general, it is a good method to detect the CEM resources shortage as


the CEMAllocFail counter has some limitations (see CEM analysis).
> HSDPA CAC is also pegging the UnsuccesfulHsdpaCac counter (#0958)

78

Nortel Confidential Information

Bottleneck analysis
Unspecified
Detailed investigation method
RB Establishment Unsuccess

no

Unspecified > 0

Stop

yes

RL Reconfiguration Prepare Unsuccess


RLReconfFailure RB Establishment Unsuccess
Unspecified

no
RL Reconfiguration Cancel > 0
And correlated in time with
RB Establishment Unsuccess
Unspecified

no
Unsucc. HSDPA CAC > 0
And correlated in time with
RB Establishment Unsuccess
Unspecified
79

yes

yes

Call allocation failure due to lack of CEM resources


(HBBU or DBBU) => relay on CEM Bottleneck analysis

Possible RNC - Core Network UPlane initialization issues. Confirmed if


the issue occurs on several BTSs of the same RNC (depending on their
traffic load).
It is possible to identify the Core domain generating the issue by
checking the RadioLinkReconfigurationCancel.DlAccessStratumConf5 for CS or
DlAccessStratumConf2/6/7 for PS
Check Core Network Counters & Alarms for more details

yes

The HSDPA limit of 20 UEs per cell is reached (HSDPA CAC)


Nortel Confidential Information

Bottleneck analysis
Unspecified

RadioBearerEstablishmentUnsuccess.Unspecified vs. RadioLinkReconfigurationCancel

Example
> During observation period (w10 &
w11) a large amount of

160
140

120

VS.RadioBearerEstablishmentUnsuccess.Un
specified was monitored on a large

number of ORF BTSs. It was


perfectly correlated with the

100
80
60

40

VS.RadioLinkReconfigurationCancel:

20

> The investigations showed that only

VS.RadioLinkReconfigurationCancel.DlAcce
ssStratumConf5 (CS voice) was

pegged, thus a possible issue


between RNC and the Core CS.

VS.RadioLinkReconfigurationCancel

VS.RadioBearerEstablishmentUnsuccess.Unspecified
7000

> When investigated at RNC level, the


issue was more visible on 6 RNCs
connected to the same MSC (and
probably to the same ALI board)
> The issue disappeared after some
operations of MSC side (RNC
reparenting & ALI redimensioning) but
it seems to be back again after 3
weeks (to be checked):
80

VS.RadioBearerEstablishm entUnsuccess.Unspecified

6000

5000
GRENOBLE1
4000

MARSESTM2
MONTPELL1
MARSESTM1

3000

DIJONC1
TOULON_SF1

2000

1000

Nortel Confidential Information

Bottleneck analysis
BTS Channel elements (CEM)
Blocking Detection

>

The CEM blocking is particular annoying as this resource is not managed


by the RNC. Thus, if no CEM resources available, the call is released (no
iRM downgrade based on this resources load is possible until UA5.0)

>

Call admission blocking due to CEM resources shortage can be detected


by two methods:

1. Using the VS.RadioBearerEstablishmentUnsuccess.Unspecified counter


see the Unspecified analysis above.
2. Using BTS counter CEMAllocFail. Method drawbacks:

>

This counter provides the % of CEM resources requests that are refused from
total CEM resources requests. All requests are considered: RL Setup SRB or
TRB, RL Reconfiguration Up or Down thus it will be slightly incremented even
in case of high blocking.
It is rounded to Integer (if 0.8% is calculated => 0% is displayed)

Recommendation: use the first method and only confirm (if it is the case)
with the CEMAllocFail method.
81

Nortel Confidential Information

Bottleneck analysis
BTS Channel elements (CEM)
Resource Load evaluation
> CEM load (DBBU) is measured by means of CEMUsed (BTS counter)
>

The CEMUsed screenings (Avg, Min, Max.) allows to:

>

Min during the night it will provide info on Common Channels utilization (usually 18% for
STSR1 on 1 iCEM64)
Avg Average load during one Hour used for dimensioning purposes
Max allows to determine the moments when the alarm thresholds were exceeded
triggering blocking.
RAB Type (UL/DL)
Block. thresholds

CEMUsed booking levels for


the most used RABs in ORF
in case of 1 iCEM64:

CS 12.2/12.2

98%

PS 64/64, 128/128, 128/384, 128/0


(HSDPA)

85%

PS 64/384

72%

PS 384/0 (HSDPA)

65%

>

Recommendation: set an Yellow alarm monitoring trigger for CEMUsed.Avg


exceeding 50% and a red one for CEMUsed.Max exceeding constantly 70%.

>

CEM load (HBBU) cannot be directly monitored in UA4.2. Alternative through


DlAsConfIdAvgNbrEstab.Max is proposed (next slide). Complete solution in UA5.0
82

Nortel Confidential Information

Bottleneck analysis
BTS Channel elements (CEM)
Detailed investigation method
no

Unspecified analysis
shows CEM blocking

Stop

yes (Optional)
yes
CEMAllocFail >

Call allocation
failure due to
lack of CEM

CEMUsed.Max

> 70%

yes
Plan to add additional
CEM (DBBU) resources

83

no

CEMAllocFail could display 0 even with


High PS blocking if there are many
resources requests that are accepted (small
RABs) => continue CEM investigation

yes
no

UA4.2
UA5.0
MAX
[DlAsConfIdAvgNbrEstab.Max
HsdpaUesperHbbu.Max
=X
(scr.27)] ~ X

yes

Plan to add additional


CEM (HBBU) resources

no
-Potential issues with CEM configuration
(check CCP configuration).
- STSR2 is activated and the traffic is
unbalanced between frequencies => Add
additional CEM resources
Nortel Confidential Information

X = 20 (UA4.2)
X = 48 (UA5.0)

Bottleneck analysis
Iub Interface
Blocking Detection

> The Iub resources management is done at RNC Level. From UA4.2 this
resource will be taken into account in the iRM @ CAC mechanism.
> Call admission blocking due to Iub resources shortage can be detected
through the counter:

VS.RadioBearerEstablishmentUnsuccess.LackOfBandwidth & LackOfCid


In UA4.1 this counter is not making the difference between different Iux interfaces,
thus additional check is needed => if problem is seen only on some (isolated)
BTSs, then it is an Iub problem.
From UA4.2, specific screenings per Iux interface will be available (better
precision during bottleneck identification phase)

> Additional indication about call admission blocking and iRM upgrade
procedures fail due to lack of Iub resources is provided by:
VS.RadioLinkReconfigurationPrepareUnsuccess.INodeRefusal (enhanced by
two new screenings in UA4.2 LackBwthIub & LackCid)
> When the mobility is affected by the lack of Iub resources the following
counter will be pegged: VS.RadioLinkSetupUnsuccess.INodeRefusal
(enhanced by two new screenings in UA4.2 LackBwthIub & LackCid)
84

Nortel Confidential Information

Bottleneck analysis
Iub Interface
Resource Load evaluation

> Iub Load indicators available in UA4.1:


Real load of the Iub calculated from BTS Transport counters (average/hour):
Monitored_Iub_Load = (VS.AAL2NbReceivedCell + VS.AAL5NbReceivedCell) / 4528

The Iub Load estimated by the RNC and used by iRM CAC:
RNC_Estimated_Iub_load = ( RadioLinkEstablishedPerCell.Avg) * ECR * SrHO% (at cell level)

The ECRs are calculated from configured EBRs.

> If the EBR configuration is very conservative (AF 100% considered for all
RBs, especially the big PS RBs), the gap between the Monitored_Iub_Load
and RNC_Estimated_Iub_load will be important and the blocking will be seen
with very low Iub traffic (currently the case in ORF).
> From UA4.2, the real Iub load is taken into consideration in the iRM and it
will be possible to monitor the percentage of time a certain load level was
reached through iRM specific counters:
#1155 VS.IRMTimeCellDlIubTransportColorYellow (recommended threshold @ 70%)
#1148 VS.IRMTimeCellDlIubTransportColorRed (recommended threshold @ 90%)
85

Nortel Confidential Information

Bottleneck analysis
Iub Interface

RB Establishment Unsuccess
LackOfBandwidth will be replaced in UA4.2 by
LackOfBandwidthOnIu
LackOfBandwidthOnIur
LackOfBandwidthOnIub
And Lack of CID
LackOfCidOnIu
LackOfCidOnIur
LackOfCidOnIub

Detailed investigation method

RB Establishment Unsuccess

no

Lack of bandwidth > 0


Or

Lack of CID > 0

yes
Call allocation failure
due to Iub Lack of
bandwidth or CIDs

yes (optional)

RL Reconfig Prepare Unsuccess


INodeRefusal RB Establishment Unsuccess
LackOfBandwidth

yes
Call allocation & iRM Sched. upgrade
failure due to lack of Iub Resources

Monitored_Iub_Load ~
IRMTimeCellDlIubTransportColorRed X%
RNC_Estimated_Iub_load?

yes
Plan to add additional
PCM resources
86

RadioLink Reconfiguration Prepare Unsuccess


INodeRefusal will be completed in UA4.2 with:
LackBwthIub
LackOfCid

Stop

no

UA4.2
UA4.1

no

Transport configuration
allows EBR Tuning?
Nortel Confidential Information

- X can vary with EBR config., traffic type


(R99/HSDPA) and Y2R/R2Y Iub
thresholds
- For R99, EBR at max and Y2R/R2Y Iub =
90%/80%, X is expected to be 1%
Additional analysis is needed for HSDPA

yes

Do EBR Tuning

Bottleneck analysis
BTS DL RF Power
Blocking Detection

>

The DL Power management is done by the RNC. This resource is taken


into account in the iRM @ CAC mechanism (the blocking due to this
resource can be controlled through cell color thresholds).

>

Call admission blocking due to DL RF Power shortage can be detected


through the counter: VS.RadioBearerEstablishmentUnsuccess.
UnavailableDlPowerResources.

>

Additional indication about call admission blocking and iRM upgrade


procedures fail due to lack of power is provided by:
VS.RadioLinkReconfigurationPrepareUnsuccess.RrmRefusal

>

When the mobility is affected by the lack of power the following counter
will be pegged: VS.RadioLinkSetupUnsuccess.RrmRefusal

>

The RrmRefusal cause is also covering the lack of OVSF codes


resources.

87

Nortel Confidential Information

Bottleneck analysis
BTS DL RF Power
Resource Load evaluation

>

Power utilization is measured by means of VS.AvgTxPower counter

>

Power usage is computed based on PA_Power @ Reference point


Power_Load (%) = VS.AvgTxPower / PA_Power@Refpoint
PA_Power@Refpoint = maxPowerAmplification TotalLoss

>

Call Admission is applied on 85% of Power_Traffic:


Power_Traffic= PA_Power@Refpoint Power_Common_Channel

>

For details on maxPowerAmplification, TotalLoss & Power Common


Channel computation see UPUG

>

Recommendation: set an alarm monitoring trigger for Power_Load Max


(based on VS.AvgTxPower.Max) exceeding constantly 80%

88

Nortel Confidential Information

Bottleneck analysis
BTS DL RF Power
Detailed investigation method
RB Establishment Unsuccess
UnavailableDlPowerResources > 0

no

Stop

yes (optional)

yes

RL Reconfig Prepare Unsuccess


RrmRefusal > RB Establishment Unsuccess
UnavailableDlPowerResources

yes
Call allocation failure due to
lack of DL Power resources

Call allocation & iRM Sched. upgrade


failure due to lack of DL Power resources

Power_Load.Max (%) > 80 %

Power_Load % = VS.AvgTxPower / PA_Power@Refpoint


PA_Power@Refpoint = Min(PAMax Loss, MaxTxPower)

yes
- Reduction of iRM Power Thresholds
- Tuning of Power allocation of Radio Bearer
- Tuning of Power allocation of common Channels
- Add a new capacity Layer

89

Nortel Confidential Information

Bottleneck analysis
OVSF Codes
Blocking Detection

>

The OVSF codes management is done by the RNC. This resource is taken
into account in the iRM @ CAC mechanism (the blocking due to this
resource can be controlled through cell color thresholds).

>

Call admission blocking due to OVSF Codes shortage can be detected


through the counter: VS.RadioBearerEstablishmentUnsuccess.
UnavailableDlCodeResources.

>

Additional indication about call admission blocking and iRM upgrade


procedures fail due to lack of codes is provided by:
VS.RadioLinkReconfigurationPrepareUnsuccess.RrmRefusal

>

When the mobility is affected by the lack of codes the following counter will
be pegged: VS.RadioLinkSetupUnsuccess.RrmRefusal

>

The RrmRefusal cause is also covering the lack of DL Power resources.

90

Nortel Confidential Information

Bottleneck analysis
OVSF Codes
Resource Load evaluation

>

Codes load is measured by means of

VS.IRMTimeFreeDlCodesBySpreadingFactor counter. It is recommended to

use SF128 as it has a better granularity.


>

Following metric is to be used to monitor the Codes load:


128_Codes_load (%) = (128 - VS.IRMTimeFreeDlCodesBySpreadingFactor.128) / 128

>

The VS.IRMTimeFreeDlCodesBySpreadingFactor screenings (Avg, Min,


Max.) allows to:

>

Max during the night it will provide info on Common Channels utilization
(usually 3-4%)
Min allows to determine the moments when the alarm thresholds were
exceeded triggering blocking.

OVSF Codes booking levels for


the most used DL RBs in ORF:

91

RB Type (DL)

SF128_Blocking threshold

CS12.2

99%

PS 64

97%

PS 128

90%

PS 384

78%

Nortel Confidential Information

Bottleneck analysis
OVSF Codes
Detailed investigation method
RB Establishment Unsuccess
UnavailableDlCodeResources > 0

no

Stop

yes (optional)
RL Reconfig Prepare Unsuccess
RrmRefusal > RB Establishment Unsuccess
UnavailableDlCodeResources

yes

yes
Call allocation failure due
to lack of OVSF codes

128_Codes_load_Max (%) >

75%

yes
Plan to add additional
frequency (additional
codes tree)

92

Call allocation & iRM Sched. upgrade


failure due to lack of OVSF codes

yes (optional)
iRM threshold for codes (green2Yellow &
yellow2RedCLCThreshold) reduction is another
alternative but it has to be correlated with operator
strategy against user service quality

Nortel Confidential Information

Bottleneck analysis
Blocked traffic investigation
> Even it is an optional step in the capacity bottlenecks study, it
is an important pointy of investigation which can provide
important information on the cause and the source of
blocking.
> Once the bottleneck identified, the blocking investigation can
be pushed further by analyzing the difference:
VS.RadioBearerSetupRequest.DlAccessStratumConfX VS.RadioBearerSetupSuccess.DlAccessStratumConfX (X from 0 to 29)

> This will reveal the RB blocked an can indicate some actions
to take:
If CS is blocked, there is not too much tuning to perform resources
addition is needed.
If PS is blocked there is still iRM tuning to foresee before deciding to
add resources.
93

Nortel Confidential Information

Agenda
> Introduction & Executive summary
> Capacity Monitoring method

Blocking & Resources Load Characterization


Congestion and Resources Load at network level
Top N most congested cells identification
Bottlenecks identification
Bottlenecks troubleshooting & analysis

> Capacity Monitoring Use cases


> iRM & Benchmark

94

Nortel Confidential Information

Capacity Monitoring Use cases


Use case 1: Low RB Setup and high blocking
Use case 2: High RB Setup and high blocking
Use case 3: High RB Setup and low Blocking

95

Nortel Confidential Information

USE CASE 1: Low RBSetup but high blocking


Summary
> Site to be analyzed : ST_ETN_GARE
> Cell : ST_ETN_GARE_U11
> Capacity Analysis:
-

One day from the Obs. period (14/03) is chosen for detailed analysis
Blocking causes identification: LackOfBandwidth & Unspecified
Bottlenecks identification: Iub & CEM
Traffic analysis:

High amount of PS traffic : 9 Erl /BH (several hours) & 166 Kbps @BH
PS384 main traffic: Setup/Request = 77/179 = 43% admission / day /site
PS384 converted to PS64 due to Iub blocking: 80 (44%) / day /site
PS384 lost due to Iub blocking: 22 (13%) / day /site

- Recommendations for this site:


- Tuning EBRs or add PCM
- add CEM
96

Nortel Confidential Information

USE CASE 1: Low RBSetup but high blocking


Choosing blocking day for detailed analysis
RB Blocking Rate at BH (ST_ETN_GARE_U11)
70%

Blocking (%)

50%

60

Day chosen for


detailed analysis

50

40%

40

30%

30

20%

20

10%

10

0%

Nb of RBs

60%

70

RB Blocking Rate at BH
RB Setup Request at BH
RB Blocked at BH

> Cell very blocked during the entire week (exception being the week-end)

97

14/03 is chosen for deeper analysis the Blocking rate is


not the highest at this date, but the cell has the highest
number of RB Setup Requests and RB Blocking.
Nortel Confidential Information

USE CASE 1: Low RBSetup but high blocking


Blocking causes identification
Blocking causes identification (ST_ETN_GARE_U11)
40
35

35
32

30

29
27

25

RBEstablishmentUnsuccess - total

Peaks of
Lack of Bandwidth
& Unspecified

20

RBEstablishmentUnsuccess.Unspecified
RBEstablishmentUnsuccess.LackOfBandw idth
RBEstablishmentUnsuccess.UnavailableDlCodeResources
RBEstablishmentUnsuccess.UnavailableDlPow erResources

15
13
12

12

13

10
8
5
0

4
3
1

5
4

3
3
2
2
1
1
1
1
1
0
00
0
0
0
0
0
0
0
0
0
0
0
00
0
0
00
00:00 01:00 02:00 03:00 04:00 05:00 06:00 07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00

> High rate of Lack of Bandwidth is observed:


From UA4.2 Iub/Iu/Iur specific screenings will be available.

> A second cause of blocking is Unspecified:


Additional investigation is needed to clearly identified the cause of unspecified
blocking (see next slide)

98

No Codes & Power blocking issues

Lack of Bandwidth and Unspecified blocking identified


Nortel Confidential Information

USE CASE 1: Low RBSetup but high blocking


Bottlenecks analysis
Bottlenecks analysis - "Unspecified" detailed investigation
35
31
30

28
29

27

25
RLReconfPrepareUnsuccess.RadioLinkReconfigurationFailure
20

RBEstablishmentUnsuccess.Unspecified
RLReconfPrepareUnsuccess.INodeRefusal

15

13
13

12

12

10

RBEstablishmentUnsuccess.LackOfBandw idth

14

12

VS.RadioLinkReconfigurationCancel

8
5

7
6

3
4
2
2
1
1
11
1
0
0
0
0
0
0
0
00
0
0
0
00
0
0
0
00
0
00:00 01:00 02:00 03:00 04:00 05:00 06:00 07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00
3

> RL Reconfiguration Cancel is 0, so no Core Network issues are foreseen


> RL Reconfiguration is failing in the Preparation phase:
Cause INodeRefusal is confirming the Iux lack of bandwidth (linked to
LackofBandwidth cause)
=>As the issue is not seen on the entire RNC we can conclude that Iux = Iub
Cause RadioLinkReconfigurationFailure confirms a lack of resources at NodeB
level (generally caused by CEM resources).

Iub and CEM blocking are confirmed


99

Nortel Confidential Information

USE CASE 1: Low RBSetup but high blocking


Blocked traffic type identification
Rejected Traffic vs Blocking
30

25
21
RB Setup Unsucc PS64

Nb of RBs

20
17
16

RB Setup Unsucc PS384


RB Setup Unsucc CS

15

RB Setup Unsucc PS128


10

10

11

RBEstablishmentUnsuccess.Unspecified

10

RBEstablishmentUnsuccess.LackOfBandwidth

7
6
5
2
00

00

00

00

00

00

00

00

2
0

1
0

3
2

22
00

4
2

00

00

00

00

> RB Setup Unsuccess is calculated per RB type as difference between RB Setup


Request and RB Setup Success.
> The CS and PS128 traffic are practically not blocked

CS is a small RB which passes even if high blocking is present at Iub and CEM level.
PS128 no blocking as very few requests are present.

The only traffic blocked is the PS384 and PS64

100

Nortel Confidential Information

USE CASE 1: Low RBSetup but high blocking


Blocking traffic analysis
Max RL Establsihed vs Blocking
12

35

Peak of PS384
call duration (700s)

10

9
8

30
9
25

Nb of RL

20

6
5

15

Max of 3 PS384

4
3

3
2

22
1
0

0
01:00

10
2

00:00

02:00

Max Nb of RL PS64

1
0

03:00

1
0

04:00

00

00

00

05:00

06:00

07:00

Max Nb of RL PS384

2
1

1
00

08:00

Max Nb of RL CS

09:00

10:00

11:00

Max Nb of RL PS128

12:00

13:00

14:00

15:00

16:00

17:00

18:00

RBEstablishmentUnsuccess.Unspecified

19:00

20:00

21:00

22:00

RBEstablishmentUnsuccess.LackOfBandwidth

>

The max of 3 PS384 accepted by the Iub (with the current EBR configuration) is reached
during 4 hours only in the observed cell (not counting the traffic in other cells).

>

This is combined with a high PS384 Call duration (200 s in average, 700s peak at 14h00)

>

The Iub capacity with the current


EBR config - see tables:

101

if x PS384
1
2
3

PS64 ?
15
9
3

PS128?
7
4
1

if x PS64
7
8
9
10

PS128?
7
6
6
5

PS384?
2
2
2
1

Iub limit is reached constantly during the day => blocking


Nortel Confidential Information

23:00

RB Blocked

10

USE CASE 1: Low RBSetup but high blocking


Traffic blocking consequences
PS64 Setup correlation with Iub Blocking
30

25

20

15

10

RB Setup Request PS64

RBEstablishmentUnsuccess.LackOfBandwidth

> Due to rapid blocking at Iub level (max 3 PS384 per Iub with 1 PCM), a
RB Reallocation mechanism is triggered at Core PS level (see details in
next slide). Due to this mechanism, a PS384 failed setup is replaced by a
PS64 RB request.

This mechanism is allocating a lower PS RAB (PS64) maintained


during entire call duration (not eligible to any iRM Upgrade) => user
perception: Good for call admission but Bad for service quality
102

Nortel Confidential Information

USE CASE 1: Low RBSetup but high blocking


RequestedUEMaximum
Bit Rate not Available
(20)
Node B
RNC

CN - PS

PS call initial connection (RRC phase)


PS call RAB allocation phase starts with PS384
RAB Assignment Request (DL384/UL64)

No Ressources
RAB Assignment Response (Fail: 20)
RAB Assignment Request (DL64/UL64)

PS call RAB allocation phase continue with PS64

> If no resources are available at RNC level, the RB Assignment Request


(for PS384 or PS128) is answered with a RAB Assignment Response
(Fail: 20) which is immediately followed by a new request for a lower RAB
(instead of Iu Release Command)
to be confirmed if it works only with Iub or all other resources managed by
RNC (Codes & Power)
It not works with the CEM resources which are managed by the BTS.

Using this Core PS mechanism is helpful to reduce Iub blocking,


but it will work against RB adaptation feature (UA4.2) allocated
RB will not be eligible to any upgrade.
103

Nortel Confidential Information

USE CASE 1: Low RBSetup but high blocking


Load details all resources
Load per Resource type
100%
90%

14/03

80%

Load (%)

70%

iRM Green2Yellow limit

60%
50%
40%
30%
20%
10%
0%

Pow er load Max

Codes Load Max

CEM Used Max

IUB Load

> Iub load curve represents the average Iub usage during each hour.
> Power, Codes and CEM Load curves present the Max load.

Power & Codes load is below iRM Green2Yellow threshold (50%)

CEM load is significantly high explaining the CEM blocking

IuB load is low but the Iub blocking is present > to be investigated
104

Nortel Confidential Information

USE CASE 1: Low RBSetup but high blocking


CEM Blocking vs. load (week11)
CEM Load vs CEM Blocking (w eek)

load (%)

90%

10

14/03
PS Blocking limit

80%

70%

60%

50%

40%

30%

20%

10%

0%

RB Estab Blocked by CEM

CEM Load Max (%)

RB Blocked

100%

CEM Used Avg (%)

> CEM Blocking is observed only when the PS384 & PS64 blocking limit is reached
this happens constantly during the loaded period of the week (working days)
> CEMAllocFail was < 1% during all this period as the number of CEM resources
request was much higher that the refusal > not a good indicator for CEM blocking

CEM Blocking limit is constantly reached during working days

105

Recommendation => plan to add CEM


Nortel Confidential Information

USE CASE 1: Low RBSetup but high blocking


Iub Blocking vs. load (week11)
Iub traffic: Real vs Estimated

3500

35

14/03

Peak of
PS384 Setup
on Cell U21

3000

30
25

cells/s

2500
20
2000
15
1500

Mostly
CS traffic

1000

10

500

RB Estab on current cell Blocked by Iub

ECR Estimated

RB Blocked

4000

ECR realy passed

> Important gap is seen between the real ECR and the ECR estimated by the RNC
> Configuring EBRs at AF 100% for the PS traffic (especially PS384) is creating
important false blocking.
> If the EBR cannot be reduced (transmission architecture constraints), additional
PCM is to be added on Iub.

EBRs for PS are overestimated possible parameters tuning


106

Alternative: additional PCMs

Nortel Confidential Information

Capacity Monitoring Use cases


Use case 1: Low RB Setup and high blocking
Use case 2: High RB Setup and high blocking
Use case 3: High RB Setup and low Blocking

107

Nortel Confidential Information

USE CASE 2: High BRSetup & high blocking


Summary
> Site to be analyzed : DJN_HAVRE
> Cell : DJN_HAVRE_U11
> Capacity Analysis :
- One day from the Obs. period (15/03) is chosen for detailed
analysis
- Blocking causes identification: LackOfBandwidth & Unspecified
- Unspecified cause investigation => Core CE issue
- Bottlenecks identification: Iub
- Traffic analysis:
- CS -> main requested traffic
- PS -> mix of PS128 and PS384

- Recommendations for this cell/site:


- Tuning EBRs for Iub or add PCM

108

Nortel Confidential Information

USE CASE 2: High RBSetup & high blocking


Choosing blocking day for detailed analysis

20%

Day chosen for


detailed analysis

Radio Bearer Blocking at BH


250

18%
16%

200

14%
12%

150

10%

RB Blocking Rate BH
RB Setup Request at BH

8%

100

RB Blocking at BH

6%
4%

50

2%
0%

0
0607080910111213141516171819mars- mars- mars- mars- mars- mars- mars- mars- mars- mars- mars- mars- mars- mars06
06
06
06
06
06
06
06
06
06
06
06
06
06

109

15/03 is identified as the most blocked day: deeper analysis


is to be performed
Nortel Confidential Information

USE CASE 2: High RBSetup & high blocking


Blocking causes identification
Blocking cause identification

35

Peaks of
Lack of Bandwidth
& Unspecified

30
25
20
15
10
5
0
RBEstablishmentUnsuccess - Total

RB EstablishmentUnsuccess.Unspecified

RB EstablishmentUnsuccess.LackOfBandwidth

RB EstablishmentUnsuccess.UnavailableDlCodeResources

RB EstablishmentUnsuccess.UnavailableDlPowerResources

> High rate of Unspecified is observed:


Additional investigation is needed to clearly identified the cause of unspecified
blocking
> A second cause of blocking is lack of Bandwidth:
Iu cause is discarded because the issue is not spread over entire RNC.
Iur cause is to be foreseen only if the BTS is on RNC border
is most probably Iub as it is seen only on this site

110

No Codes & Power blocking issues

Lack of Bandwidth and Unspecified blocking identified


Nortel Confidential Information

USE CASE 2: High RBSetup & high blocking


Bottlenecks analysis
Botte lneck Analysis

35
30
25
20
15
10
5

23:00

22:00

21:00

20:00

19:00

18:00

17:00

16:00

15:00

14:00

13:00

12:00

11:00

10:00

09:00

08:00

07:00

06:00

05:00

04:00

03:00

02:00

01:00

00:00

RB EstablishmentUnsuccess.Unspecified

RL ReconfigurationCancel

RB EstablishmentUnsuccess.LackOfBandwidth

RL ReconfigurationPrepareUnsuccess.INodeRefusal

RLReconfigPrepareUnsuccess.ReconfigurationFailure

> No RLReconfigPrepUnsuccess.ReconfigFailure Leads to CEM Blocking cause


discard
> Clear correlation between Unspecified reason and RLRecofingCancel
Root cause may be on CoreNetwork to be deeply investigated (see next slide)
> Clear correlation between LackOfBandwidth Blocking and InodeRefusal cause of
RLReconfig Failure

111

Lack of Bandwidth confirmed and CEM blocking discarded


Nortel Confidential Information

USE CASE 2: High RBSetup but High blocking


Bottlenecks analysis Unspecified

Bottlneck analysis "Unspecified" investiguation


35
30
25
20

> This Cell belongs to Rnc Genoble1 : Cell


Curve for Unspecified blocking follows
the one for the Rnc which confirms a
global Issue on Rnc
=> ( see slide 30 on Unspecified analysis)

10
5
0

RLReconfigCancel.CsSpeech
RL ReconfigurationCancel - Total
RB EstablishmentUnsuccess.Unspecified
"Unspecified" correlation between Cell and corresponding Rnc
100
90
80
70
60
50
40
30
20
10
0

1000
900
800
700
600
500
400
300
200
100
0
06- 07- 08- 09- 10- 11- 12- 13- 14- 15- 16- 17- 18- 19mars- mars- mars- mars- mars- mars- mars- mars- mars- mars- mars- mars- mars- mars06
06
06
06
06
06
06
06
06
06
06
06
06
06
DJN_HAVRE_U11

112

Unspecified cause linked to CS CoreNetwork issue


Nortel Confidential Information

Dijon

"Unspecified" Blocking for Rnc

a possible issue between RNC and the


CoreNetwork CS

15

"Unspecified" Blocking for Cell

> Only CS calls are affected indicates

USE CASE 2: High RBSetup & high blocking


Blocking traffic analysis
Max RL Established vs Blocking
14

30

Bandwith Blocking
& PS traffic peak

12

25

Nb of RL

10

20

8
15
6
10

2
0

0
00:00 01:00 02:00 03:00 04:00 05:00 06:00 07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00

Max RL PS 384

Max RL PS 128

Max RL PS 64

Max RL Cs

RB EstablishmentUnsuccess.Unspecified

RB EstablishmentUnsuccess.LackOfBandwidth

> Important amount of Voice traffic


> Bandwidth-Blocking correlated to PS traffic amount and the apparition of
PS64 Rb Setup

113

Lack of Bandwidth cause linked to PS Traffic peaks


Nortel Confidential Information

USE CASE 2: High RBSetup but high


blocking
Load details all resources

Max Load per Resource (5 days evolution)

CEM Peaks

100%

15/03

90%
80%
70%
60%
50%
40%
30%
20%

Max Codes Load

IuB Load

Max CEM Load

> Power, Codes and CEM Load curves present the Max load.

CEM will be the next blocking issue

There is a Margin on Power and Code usage


Nortel Confidential Information

21:00

18:00

15:00

12:00

09:00

06:00

03:00

00:00

21:00

18:00

15:00

Max Power load

> Iub load curve presents the Average load.

114

12:00

09:00

06:00

03:00

00:00

21:00

18:00

15:00

12:00

09:00

06:00

03:00

00:00

21:00

18:00

15:00

12:00

09:00

06:00

03:00

00:00

21:00

18:00

15:00

12:00

09:00

06:00

03:00

0%

00:00

10%

USE CASE 2: High RBSetup but high


blocking
Iub Blocking vs. load

IuB Traffic: Real vs Estimated

3000

25

15/03
20

2000
15
1500
10

RB Blocked

Equivalent Cell Rate

2500

1000

500

RB EstablishmentUnsuccess.LackOfBandwidth

Iub Estimated ECR

Iub Real Cell Rate

> Important gap is seen between the real ECR and the ECR estimated by the RNC
> Configuring EBRs at AF 100% for the PS traffic (especially PS384) is creating
important false blocking.
> If the EBR cannot be reduced (transmission architecture constraints), additional
PCM is to be added on Iub.

> EBRs for PS are overestimated possible parameters tuning


> Alternative: additional PCMs
115

Nortel Confidential Information

Capacity Monitoring Use cases


Use case 1: Low RB Setup and high blocking
Use case 2: High RB Setup and high blocking
Use case 3: High RB Setup and low Blocking

116

Nortel Confidential Information

USE CASE 3: High RBSetup but Low blocking


Summary
> Site to be analyzed : GRE_ALSACE_LORRAINE
> Cell : GRE_ALSACE_LORRAINE_U21
> Capacity Analysis :

One day from the Obs. period (13/03) is chosen for detailed
analysis
Blocking causes identification: Unspecified
Unspecified cause investigation => Core CE issue
Bottlenecks identification: No UTRAN botleneck
Traffic analysis:
CS -> main requested traffic
PS -> very low PS traffic level

Recommendations for this cell/site:


None

117

Nortel Confidential Information

USE CASE 3: High RBSetup but Low blocking


Choosing blocking day for detailed analysis
Day chosen for
detailed analysis

> Cell with High Traffic and High-but-ponctual blocking on few days

118

13/03 is identified as the most blocked day with the higher


traffic: deeper analysis is to be performed
Nortel Confidential Information

USE CASE 3: High RBSetup but Low blocking


Blocking causes identification
Blocking cause Identification (13/03)
120
100
80
60
40
20
0

RB Establishement Unsuccess - Total

RB Establishement Unsuccess.Unspecified

RB Establishement Unsuccess.LackOfBandwidth

RB Establishement Unsuccess.UnavailableDlCodeResources

RB Establishement Unsuccess.UnavailableDlPowerResources

> No Codes , Power, lack of Bandwidth blocking


> Total Blocking curve is equal to Unspecified Blocking cause

119

Unspecified is the root cause of blocking


Nortel Confidential Information

USE CASE 3: High RBSetup but Low blocking


Bottlenecks analysis - "Unspecified" detailed investigation

Bottlenecks analysis Unspecified


> VS.RadioBearerEstablishmentUnsuccess.
Unspecified is perfectly correlated with the
VS.RadioLinkReconfigurationCancel
> Only CS calls are affected; thus a possible
issue between RNC and the Core Network
CS
> This Cell belongs to Rnc Genoble1 : Cell
Curve for Unspecified blocking follows the
one for the Rnc which confirms a global Issue
on Rnc
=> ( see slide 30 on Unspecified analysis)

140
120
100
80
60
40
20
0

RBEstablishmentUnsuccess.Unspecified

RLReconfigCancel

RLReconfPrepareUnsuccess

RLReconfCancel.CsSpeech

Unspecified Correlation between Cell and Corresponding Rnc

4000

1000

3500

900
800

3000

700

2500

600

2000

500

1500

400
300

1000

200

500

100

0
0607080910111213141516171819mars-06 mars-06 mars-06 mars-06 mars-06 mars-06 mars-06 mars-06 mars-06 mars-06 mars-06 mars-06 mars-06 mars-06
GRE_ALSACE_LORRAINE_U21

120

RNC: GRENOBLE1

Unspecified cause linked to CS Core Network issue


Nortel Confidential Information

USE CASE 3: High RBSetup but Low blocking


traffic type distribution

Max RadioLink Established vs Blocking

25

120

Nb of RL

80
15
60
10
40
5

Nb of Blocking

100

20

20

Max Nb of RL Cs
Max Nb of RL Ps 128
RB EstablishmentUnsuccess.Unspecified

Max Nb of RL Ps64
Max Nb of RL Ps 384

> Very few Ps Calls


> The Cell is Speech Oriented : This Confirms that Cell is not a serious candidate to
Capacity Issue

121

The Cell is Speech Oriented ; Low Blocking probability


Nortel Confidential Information

USE CASE 3: High RBSetup but Low blocking


Load details all resources

Load per Resource Type

100%
90%

13/03

80%
70%
60%
50%
40%
30%
20%
10%
0%

00:00 07:00 14:00 21:00 04:00 11:00 18:00 01:00 08:00 15:00 22:00 05:00 12:00 19:00 02:00 09:00 16:00 23:00
Max Codes Load

IUB Load

Max CEM Load

Max Power load

> More than 30% Margin on CEM Load before blocking , considering the
same CS profile
> Power and Codes Load curves far from the iRM GreenToYellow threshold
(50%)
> considering the CS oriented traffic, the Iub load should reflect the estimated
one
enough margin for additional traffic

122

Load analysis confirms the absence of capacity issues


Nortel Confidential Information

Agenda
> Introduction & Executive summary
> Capacity Monitoring method

Blocking & Resources Load Characterization


Congestion and Resources Load at network level
Top N most congested cells identification
Bottlenecks identification
Bottlenecks troubleshooting & analysis

> Capacity Monitoring Use cases


> iRM & Benchmark

123

Nortel Confidential Information

iRM Benchmark
Summary

> Benchmark based on Configuration information and Monitoring results


coming from 3 different Operators.

one representative RNC per Operator is monitored (usually the most loaded)
One RNC is in UA4.1 (Operator 2) and the other two (Lyonecul1 & Operator 1)
are in UA4.2 but with basically no UA4.2 features activated (only UA4.1 features)
Observation period: one week (in April)

> RRM State Machine for UA4.1 and UA4.2 evolution


> iRM Configuration for the 3 operators:

iRM Table,
Cell/RL color & iRM Preemption
iRM Scheduling Downgrade & Upgrade
Always_ON,

> Monitored results for the three Operators:

Traffic profile in terms volume (Kbytes) & type of PS traffic.


High level Blocking analysis (RNC level)
Cell/RL color & iRM Preemption
IRM Downgrade (iRM&CAC)
iRM Scheduling Downgrade & Upgrade
Always_ON

> Conclusions
124

Nortel Confidential Information

iRM Benchmark
RRM State Machine: UA4.1
Reference

- From UA4.1 the RRN state


machine becomes complex
with many interactions.
- Even more complex in
UA4.2 see next slide

iRM Cac

Quality Trigger

IC Target = IRM Table(Reference)

iRM Sched Upgrade


DTxCP

Default : Selected = IC Target

Selected = MIN(IS Target, IC Target)

iRM
Scheduling
Downgraded

Activity Trigger

iRM Sched Downgrade


BLER

Always
On
Downsize
d

if Preempted is
384, 256 or 128

Capacity Trigger

Selected
Selected

Always-On
Upsize

Always On
Release

Always-On
Downsize

iRM Preemption

Preempted
125

Nortel Confidential Information

Idle

iRM Benchmark
RRM State Machine: UA4.2
Reference
Quality Trigger

iRM Cac
IC Target = IRM Table(Reference)

iRM Sched Upgrade


DTxCP
iRM
Scheduling
Downgraded

RB Rate Adaptation
(Upsize)

RB Rate Adaptation
(Downsize)

Default : Selected = IC Target

Selected = MIN(IS Target, IC Target)

Selected = MAX(Current, MIN(IC


Target, RA Target))

Selected = MIN(RA Target, IC Target)

Selected
Selected

iRM Sched Downgrade


BLER / DTxCP

Always-On
Upsize
Always
On
Downsize
d

if Preempted is
384, 256 or 128

Capacity Trigger

Activity Trigger

Always On
Release

Always-On
Downsize

iRM Preemption

Preempted

Idle
126

Nortel Confidential Information

iRM Benchmark
RRM Configuration UA4.1
Parameter
irmDlCoverageThreshold
irmDlPowerThreshold

ORF

Operator 1

Operator2

-100 -100

-100

-103

15

14

15

12

yellow2GreenCLCThreshold

40

40

52

green2YellowCLCThreshold

50

50

56

red2YellowCLCThreshold

60

60

65

yellow2RedCLCThreshold

70

70

70

yellow2GreenPLCThreshold

40

40

65

green2YellowPLCThreshold

50

50

70

red2YellowPLCThreshold

75

60

80

yellow2RedPLCThreshold

80

70

85

PS384 RL Green Cell Green

384

384

384

PS384 RL Green Cell Yellow

384

128

128

PS384 RL Green Cell Red

64

64

64

PS384 RL Red Cell Green

128

128

128

PS384 RL Red Cell Yellow

128

64

128

PS384 RL Red Cell Red

64

64

64

PS128 RL Green Cell Green

128

128

128

PS128 RL Green Cell Yellow

128

64

128

PS128 RL Green Cell Red

64

64

64

PS128 RL Red Cell Green

128

64

128

PS128 RL Red Cell Yellow

128

64

128

PS128 RL Red Cell Red

64

64

64

127

RL Color thresholds are close in


terms of Ec/N0. Operator2 more
permissive in terms of RSCP.
Cell Color thresholds for codes are
very close for the 3 Operators. Again
Operator 2 is more permissive for the
power thresholds
Different strategies for iRM table:
ORF no difference between green
and yellow in terms of decision. Only
cell red can lead to PS64
Operator1 aggressive downgrade
policy especially for RL Red
Operator 2 is playing moderate.
Cell Red is triggering PS64 (as ORF)

Nortel Confidential Information

iRM Benchmark

Parameter

RRM Configuration UA4.1

Always_ON timers are completely


different configured: very aggressive in
case of Operator 1 & 2 (due to DCH
used as downsize state) and very
relax in case of ORF (due to FACH
used as downsize state)

ORF

Operator 1

Operator2

timerT1-gold

20000

5000

3000

timerT2-gold

120000

10000

20000

timerT1-silver

20000

5000

3000

timerT2-silver

120000

10000

20000

targetDlRbSetForIrmScheduling6

3 (false)

3 (false)

targetDlRbSetForIrmScheduling7

2000

5000

2000

default

default

default

yes

yes

no

isIrmSchedulingUpgradeAllowedFro
mThisUS

128

64

isIrmSchedulingUpgradeAllowedToT
hisUS

384

128

384

23

27

26.5

384

128

384

5000

5000

5000

384

128

128

55000

500

500

timeToTrigger (PS384 -> PSxx)


radioQualityMeasurementQuantity
isIrmSchedulingUpgradeAllowed

iRM Scheduling Downgrade is mainly


used for PS384 (only Operator1 using
it for PS128 too). Very long time to
trigger for Operator1.
iRM Scheduling Upgrade is only used
by ORF and Operator 1.
iRM Preemption is used by ORF and
Operator 1. Thresholds used for
congested state are same as for the
Yellow2Red/Red2Yellow (iRM table)

FlipflopUpDowngradeTimer

thresholdTransCodePower (dBm)
highBitRate
timeToTrigger
deltaThresholdTransCodePower
intermediateRate
deltaTimeToTrigger

isIrmPreemptionAllowed

yes

yes

no

normal2CongestedPLCThreshold

80

70

70

congested2NormalPLCThreshold

75

60

60

normal2CongestedCLCThreshold

70

70

70

60

60

congested2NormalCLCThreshold
128

Nortel Confidential Information

Same as iRM
Y2R & R2Y

60

Lower than iRM


Y2R & R2Y

iRM Benchmark
Orange FR - RRM diagram UA4.1
384

RL Good
Green
Cell

Yellow Cell

384

384

Reference RB

EcNo: -12
RSCP: -100
Red Cell
Green
Cell

64

Codes

RL Bad
Yellow Cell

Red
Cell

128

64

128

70

60

50

40

Power

384
128

128

64

iRM
Sched.
TimerT1 = 20s

Always On (FACH)
129

Idle

TimerT2 = 120s

Nortel Confidential Information

80

75

50

40

iRM Benchmark
Operator 1 - RRM diagram UA4.1
384

RL Good
Green
Cell

Yellow Cell

384

128

Reference RB

EcNo: -15
RSCP: -100
Red Cell
Green
Cell

64

Codes

RL Bad
Yellow Cell

Red
Cell

64

64

128

70

60

50

40

Power

384
iRM
Sched.

128

64

64
TimerT1 = 5s

Always On (DCH)
130

Idle

TimerT2 = 10s

Nortel Confidential Information

70

60

50

40

iRM Benchmark
Operator 2 - RRM diagram UA4.1
384

RL Good
Green
Cell

Yellow Cell

384

128

Reference RB

EcNo: -14
RSCP: -103
Red Cell
Green
Cell

64

Codes

RL Bad
Yellow Cell

Red
Cell

128

64

128

70

65

56

52

Power

384
128

128

64

iRM
Sched.
TimerT1 = 3s

Always On (DCH)
131

Idle

TimerT2 = 20s

Nortel Confidential Information

85

80

70

65

iRM Benchmark
Monitoring Results Traffic Type & Volume
DedicatedDownlinkKbytesRlc (All traffic & PS only)

RBSetupRequest PS Repartition (%)


60000000

100%
90%

50000000

80%

40000000

70%
KByts

60%
50%
40%

30000000
20000000

30%

10000000

20%
10%

0
Day 1

0%
Day 1

Day 2

ORF PS384

ORF PS128

Day 3
OP1 PS384

Day 4
Op1 PS128

Op2 PS384

Day 5
Op2 PS128

Day 2

Day 3

Day 4

RNC: LYON1 (PS)

RNC: Op1 (PS)

RNC: Op2 (PS)

RNC: LYON1 (All)

RNC: Op1 (All)

RNC: Op2 (All)

Day 5

> Three different operators with three different traffic allocation strategies:

Orange mix of PS128 & PS384 RBs


Operator 1 mainly PS 384 (more than 95%)
Operator 2 mainly PS 128 (more than 90%)

> In terms of Volume: The most loaded is Operator 2 followed by Operator 1


which is also the most loaded in terms of PS traffic.

For this benchmark, ORF RNC (Lyonecul1) is the less loaded RNC
in terms of total traffic and PS traffic, but it has an homogeneous
PS traffic mix.
132

Nortel Confidential Information

iRM Benchmark
Monitoring Results High Level Blocking analysis
Blocking rate comparison (RNC Level)
0,80%

Lyn1 Blocking cause = Iub


Op1 Blocking cause = Codes
Op2 Blocking cause = RF Power

0,70%
0,60%
0,50%
0,40%
0,30%
0,20%
0,10%
0,00%

Blocking rate Light % Lyn1

Blocking rate Light % Op1

Blocking rate Light % Op2

> A Blocking rate Light taking into account all causes except unspecified is
calculated:
The Unspecified cause is not considered as it concerns only the CEM which is
not managed by the iRM in UA4.1 & UA4.2.
All other causes (Iub, Codes, RF Power) are covered by this metric.

> Operators 1 & 2 are almost not blocked at all (< 0,1%), blocking coming
from Codes (Op1) & RF Power (Op2)

Lyonecul1 is the most blocked RNC, the entire blocking coming


from Iub.
133

Nortel Confidential Information

iRM Benchmark
Monitoring Results Cell Color analysis (RNC level)
cells of only one
site (BRON) were
yellow 100% of time
To be investigated

Cell Color evolution (RNC Level)


1,4%
1,2%
1,0%
0,8%
0,6%
0,4%
0,2%
0,0%

Avg PS RL established
Cell Color Red % Lyn1

Cell Color Yellow % Lyn1

Cell Color Red % Op1

Cell Color Yellow % Op1

Cell Color Red % Op2

Cell Color Yellow % Op2

90
80
70
60
50
40
30
20
10
0

Avg Nb of PS RBs established Lyn1

>
>

Avg Nb of PS RBs established Op1

Avg Nb of PS RBs established Op2

The ratio of time when the Cell color was Red & Yellow for each of the 2 RNCs is presented on
the first graph (upper).
Average Established PS RBs per RNC/hour are presented in the lower graph allowing to
correlate with the cell color.

No R and very few Y cell color for Lyn1. iRM@CAC decision is based
practically only on RL color.

Op 1: Cell Color is following the traffic load = > full iRM table usage

134

Op 2 is slightly passing to Y and R (high thresholds for power)


Nortel Confidential Information

iRM Benchmark
Monitoring Results Cell Color analysis (Cell level)
BRON site cells
not included

>

Monitoring interval is 5 days = 120 H = 7200 min

>

BRON site cell color is not included

>

Graphs show the number of cells spending more than X s in Y and R state (during the 5 days)

Lyn1 has very few cells in Yellow and for very short time. The traffic is
low and blocked by Iub before to affect the codes&power.

Op1 has lower number of cells congested (in Y and R state) than Op2
but the time spend in congestion is much higher => high downgrade
135

Nortel Confidential Information

iRM Benchmark
Monitoring Results iRM Preemption analysis (RNC level)
iRM Preemption (RNC Level)
0,16%

Codes congestion

0,14%

Power congestion

0,12%
0,10%
0,08%
0,06%
0,04%
0,02%
0,00%

IrmPreemptionBecauseOfOvsfCodes Lyn1

IrmPreemptionBecauseOfOvsfCodes Op1

IrmPreemptionBecauseOfOvsfCodes Op2

IrmPreemptionBecauseOfPower Lyn1

IrmPreemptionBecauseOfPower Op1

IrmPreemptionBecauseOfPower Op2

> The iRM Preemption analysis can confirm the cause of cell color change.
> Even the iRM preemption is not activated in Operator 2, the iRM preempt.
counters are pegged each time the congested level is reached.

No iRM congestion monitored for Lyonecul1 => confirms 0 Cell Red

Operator 1 the OVSF codes are the main cause of Cell color
change

Operator 2 the RF Power is the main cause for Cell color change

136

Nortel Confidential Information

iRM Benchmark
Monitoring Results RL Color analysis
RL Red Evolution (week)
30%

07:00 AM

% of RL Red

25%
20%
15%
10%
5%
0%

RNC: LYON1

RNC: Op1

RNC: Op2

> Operator 2 is clearly the worst in terms of RL red (17% in average). Possible
coverage issues as it uses low PCpich values (29.5 dbm). ORF & OP1 use similar
Common Channels settings.
> Interesting phenomenon: Operator 2 has peaks of RL Red et 7:00 AM each work day
of the week (day1 was public holyday). Coverage problems during home-work route?

No RL Color dependence on traffic is observed (flat during the day)

The RL Color seems more related to coverage issues.

Lyonecul1 is better than Op2 and comparable with Op1


137

Nortel Confidential Information

iRM Benchmark
Monitoring Results iRM CAC Downgrade analysis
iRM CAC Downgrade ratio (PS384)
25%

Traffic peaks

20%
15%
10%
5%
0%

IrmcacDow ngrade Ratio PS384 Lyn1

IrmcacDow ngrade Ratio PS384 Op1

IrmcacDow ngrade Ratio PS384 Op2

> Monitoring obs. shows that PS384 is the only RB downgraded at admission, thus
the iRM CAC Downgrade ratio was computed based on PS384 RB requests (more
accurate).
> In case Lyonecul1 and Operator 2 the iRM CAC Downgrade ratio is flat during the
day which confirms the involvement of RL Color only.
> Operator 2 having a predominant PS128 traffic is less exposed to iRM downgrades
> Operator 1 IRM Downgrade curve is clearly impacted by cell color (peaks of traffic
during the day)

Lyn1 downgrade driven only by RL color based on coverage issues.


No change in UA4.2 (R99 cells), as the Iub load in iRM is performing
cell color calculation based on real load not RNC estimated Iub load.
On sites with HSDPA, the Iub load will probably change the cell color.

138

Nortel Confidential Information

iRM Benchmark
Monitoring Results iRM Scheduling Downgrade analysis
iRM Sched Downgrade (PS384) vs Avg Nb of PS384

iRM Scheduling Downgrade Ratio (PS384)


10%
35

8%

30
25

6%

20

4%

15
10

2%

0%

Day 1

Day 1

Day 2

Day 3

Day 4

Day 5

IRMSchedulingDowngrade PS384 (%) Lyn1


IRMSchedulingDowngrade PS384 (%) Op1

Day 2

Day 3

Day 4

Day 5

iRM Sched vs Avg Nb of PS384 Lyn1


iRM Sched vs Avg Nb of PS384 Op1
iRM Sched vs Avg Nb of PS384 Op2

IRMSchedulingDowngrade PS384 (%) Op2

> The iRM Scheduling is triggered only for PS384 for all three operators.
> First graph (left) presents standard iRM Scheduling Downgrade Ratio metric. This
standard metric is based on RAB Establishment Success, thus exposed to large
variation in case of high number of RAB Requests (case of AON T1/T2 very low).
> A second graph (right) is presenting the number of iRM Scheduling Downgrade per
Average PS384 established during the day (PS384 call duration is considered)

iRM Scheduling downgrade seems to behave normally on Lyonecul1

Op1 has much higher PS load but also higher TimeToTrigger which
compensate => Op1 same level of iRM Sched downgrade as Lyn1

139

Nortel Confidential Information

iRM Benchmark
Monitoring Results iRM Scheduling Upgrade analysis
IRM Scheduling Upgrade/Downgrade Ratio
30%
25%
20%
15%
10%
5%
0%
Day 1

Day 2

Day 3

Day 4

Day 5

IRM Sched Upgrade/Downgrade Ratio Lyn1


IRM Sched Upgrade/Downgrade Ratio Op1

> The iRM Scheduling upgrade is not activated in Operator 2 (so no data
available)
> The two different criteria for upgrade and downgrade (in UA4.2) are making
difficult an evaluation of the feature efficiency

iRM Scheduling upgrade seems to have the same behavior for the
two operators despite difference in thresholdTransCodePower ->
possible cause the different fallback RB used (64 in Op1 & PS128
Lyn1).
140

Nortel Confidential Information

iRM Benchmark
Monitoring Results Always ON analysis (T1)
AON Down Step1 / Average Number of PS RB Established (Erl PS)
8000

T1 = 3s

7000
6000
5000

T1 = 5s

4000
3000
2000

T1 = 20s

1000
0
Day 1

Day 2

Day 3

AON Dow n Step1 / Average PS RB Established Lyn1

Day 4

Day 5

AON Dow n Step1 / Average PS RB Established Op1

AON Dow n Step1 / Average PS RB Established Op2

> The graph is presenting the number of Always ON Downsize per Average
number of PS established during the day (PS call duration is considered
in this case). Clear dependency: T1 value => AON downgrade / PS call

High T1 means long perceived PS RAB durations (low number of


downsizes per PS RAB) => waste of UPlane resources & low PS
sessions fragmentation => increased blocking on bottlenecks.

Low T1 is part of Op1 & Op2 strategy to deal with high PS traffic
141

Nortel Confidential Information

iRM Benchmark
Monitoring Results Always ON analysis (T2)
AON Release / PS Call
0,9
0,8

T2 = 10s

0,7
0,6

T2 = 20s

0,5
0,4
0,3
0,2

T2 = 120s

0,1
0,0
Day 1

Day 2

Day 3

AON Dow n Step2 /PS Setup Lyn1

AON Dow n Step2 /PS Setup Op1

Day 4

Day 5

AON Dow n Step2 /PS Setup Op2

> The graph is presenting the number of Always ON Downsize Step2 per PS
call Established. Clear dependency: T2 value => AON release / PS call
Low T2 generates several PS RAB Establishments for one User
session -> 80% of PS calls in Op1 are ending through AON release
Very few of the PS calls are ended by Always On release in Lyn1 (<
20%): good for user perception but RNC FACH context consuming
(not an issue for the moment)
142

Nortel Confidential Information

iRM Benchmark
Conclusions

> ORF RNC (Lyonecul1) is less loaded that the other two RNCs in the
benchmark (in terms of total traffic and PS traffic), but it has an homogeneous
PS traffic mix.
> Lyonecul1 is the most blocked RNC from the three and the entire blocking is
coming from Iub (and slightly Iur). The other Operators are not facing this
issue as use different EBR Configuration:
EBR Values
DL RB
PS 64
PS 128
PS 384

ORF
69632
136832
408000

Operator 1 Operator 2
x 2/3
x 3/5
x 1/2

x1
x 3/4
x 1/2

> The Cell color is practically unused for Lyonecul1 (traffic is low and blocked by
Iub before to affect the codes & power). iRM@CAC decision base only on RL
color.
> Normal behavior/settings of iRM Scheduling downgrade.
> Long T1 timer is generating less AON downsize procedures (thus less CPlane
signaling) but is also wasting valuable UPlane resources (generating
additional Blocking).
> T1+T2 duration is seriously impacting the number of PS RAB established and
the RAB duration monitored through UTRAN counters (see call profile ORF
vs. Operator 1 next slide). Low T1 is a good method to reduce admission
blocking
Nortel Confidential Information
143
Back to highlights

Call Profile Resume


ORF
Metric name
BHCA
Mean Call Duration
Active Set Update per second
Mean Sector per user
Percentage of softer HO
Inter-RAT cell reselection
SMS
Nb of AO Downsize per PS call
Nb of AO Upsize per call
Nb of AO Release per call
Nb of IRM Sched Downgrade per
call
% of 3G-2G HO per CS call
% of 3G-2G HO per PS call
Number of 2G-3G CS HHO

PS
CS V CS D 384
1,05
95

0.01
108

0,01
636

Operator 1
PS
128

PS
Ps 64 CS V CS D 384

0,06 0,001 0,93


110 452,0 55

0,01
48

0,19
34

PS
128
0,02
34,9

Operator 3
PS
Ps 64 CS V CS D 384
0,01
23

2,04
82

0,06
103

0,26
50

0,12

0,18

0,145

1,71

1,95

1,6

27%
1,36
0,33
2,36
1,88
0,25

35%
1,81
0,27
1,57
0,79
0,68

24%
1,23
0,42
1,75
1,2
0,44

0,07

0,01

0,08

11%
13%
0

23%
14%
0

5%
10%
0

PS
128
0,03
119

Ps 64
N/A
N/A

Significant differences between ORF profile and other operators:


Low number of PS calls/Sub but very long calls (impact on UPlane) >
AON timers settings consequence.
Mean sector per user reasonable value (OK from capacity point of view).
Very low number of AON release per call unnecessary resources usage.
144

Nortel Confidential Information

Thanks for your patience!

145

Nortel Confidential Information

Backup

146

Nortel Confidential Information

147

Nortel Confidential Information

You might also like