Professional Documents
Culture Documents
0 Power Saving
(Including DRX) Solution Integration and
Optimization Guide
Approved By Date
Change History
Contents
Change History.................................................................................................................................2
1 Feature Application Policies.......................................................................................................5
1.1 DRX................................................................................................................................................................................5
1.1.1 Weak Coverage Scenarios...........................................................................................................................................5
1.1.2 Heavy Traffic Scenarios..............................................................................................................................................5
1.1.3 High-Speed Movement Scenarios (Including Highways, Metros, and Railways)......................................................6
1.1.4 Low-Band Wide-Coverage Scenarios.........................................................................................................................6
1.2 Dynamic DRX................................................................................................................................................................7
1.3 UE-Specific Differentiated DRX...................................................................................................................................8
1.3.1 Introduction.................................................................................................................................................................8
1.3.2 Parameter Configurations............................................................................................................................................9
1.3.3 Effect Evaluation Method............................................................................................................................................9
1.1 DRX
To support DRX, UEs must comply with 3GPP Release 8 or later.
It is recommended that DRX be disabled by default. If feature activation is required,
performance-preferred parameter settings are recommended first, and then power saving-
preferred parameter settings are recommended. DRX applies to the following services
where UEs on the network support DRX and focuses on power saving:
Services insensitive to delay or with sporadic data transmission, such as web
browsing, email, and FTP services
Services that use a few small packets, such as Presence services
Periodic transmission of continuous small packets, such as Voice over Internet
Protocol (VoIP) services
The signaling overhead of 10 RRC Connection Reconfiguration messages is equal to that of one E-
RAB setup procedure (including the service request procedure).
the UE power consumption can be set for UEs with a short battery life. In comparison, a
DRX parameter group used to improve the UE performance can be set for UEs with a long
battery life.
Step 2 Run the ADD UECOMPAT command to create an index and set UE-specific differentiated
DRX parameters.
For UEs with different TACs, set the UECompat.UeInfoType parameter to
IMEISV_TAC, and set the UECompat.ImeisvTac parameter.
For UEs with different TAC and SVN combinations, set the UECompat.UeInfoType
parameter to IMEISV_TAC_SVN, and set the UECompat.ImeisvTac and the
UECompat.ImeisvSvn parameters.
Step 3 Set the UECompat.UeDiffDrxSwitch parameter to ON.
For UEs with different TACs, run the following command:
ADD UECOMPAT: Index=1, UeInfoType=IMEISV_TAC, ImeisvTac=87654321,
UeDiffDrxSwitch=ON;
For UEs with different TAC and SVN combinations, run the following command:
ADD UECOMPAT: Index=0, UeInfoType=IMEISV_TAC_SVN, ImeisvTac=12345678,
ImeisvSvn=1,UeDiffDrxSwitch=ON;
Step 4 Run the ADD UEDIFFCONFIG commands with the parameters LocalCellId,
UeInfoIndex, Qci, and DrxParaGroupId specified. Among these parameters, the
UeInfoIndex parameter belongs to the UeCompact MO and the value of the
DrxParaGroupId parameter belongs to the DrxParaGroup MO.
ADD UeDiffConfig: LocalCellId=5, UeInfoIndex=0,QCI=9, DrxParaGroupId=3;
ADD UeDiffConfig: LocalCellId=5, UeInfoIndex=1,QCI=9, DrxParaGroupId=4;
After the configuration is completed, the eNodeB will deliver UE-specific DRX
configurations to UEs.
----End
This chapter describes recommended parameter settings. This does not mean that other
parameter settings cannot be used. There are various parameter setting combinations, but it
is impossible that all parameter setting combinations are verified. For details about impact
of parameter setting changes, see sections 7.1.7"Specific Methods for Feature Problem
Location" and 7.1.5"Parameter Check."
If the EPC configures the same ARP for QCIs, the DRX parameter group for the
preferentially established QCI bearer is delivered to the UE, which may not be the
expected parameter setting. In versions later than eRAN11.0, the eNodeB selects
the DRX parameter group for the QCI of services with the priority defined in 3GPP
based on switch control. It is recommended that DrxPolicyMode be set to
3GPPDEFINEDQCIPRIORITY when DRX is enabled again. Operators can also
configure the QCI priority as required and determine the DRX parameter group
based on the QCI priority if the long DRX cycle is the same for multiple QCIs (for
example, DrxPolicyMode is set to QCIPRIORITY).
When a UE simultaneously has bearers of QCI 1 and QCI 5 as well as a default bearer
(for example, of QCI 6 to QCI 9), the configuration rules for the
DrxParaGroup.LongDrxCycle parameter are as follows:
− If the default bearer and QCI 1 bearer have the same
DrxParaGroup.LongDrxCycle value (for example, 40 ms), the QCI 5 bearer must
have a larger DrxParaGroup.LongDrxCycle value (for example, 80 ms) than the
default bearer and QCI 1 bearer.
− If the default bearer has a larger parameter value than QCI 1 bearer, the default and
QCI 5 bearers must have the same DRX parameter group (that is, their
DrxParaGroup MOs must have the same parameter values).
The OnDurationTimer value should be less than the ShortDrxCycle value.
Otherwise, the short cycle is not configured. The OnDurationTimer value should be
less than the LongDrxCycle value. Otherwise, the DRX parameters are not
configured.
The LongDrxCycle value should be an integer multiple of the ShortDrxCycle value.
Otherwise, the short DRX cycle is automatically adjusted to a proper value. The
LongDrxCycle and ShortDrxCycle values must be a multiple of ten. Otherwise, the
values are automatically adjusted to a multiple of ten.
Due to impacts of the sounding reference signal (SRS) reporting period and CQI
reporting period, OnDurationTimer specifies a period in which the SRS and CQI are
reported once at the same time. The valid value of OnDurationTimer, which is
carried in the DRX reconfiguration message over the air interface, is greater than a
configured value. Due to impacts of TA configurations after SRS is enabled, the valid
value of LongDrxCycle, which is carried in the DRX reconfiguration message over
the air interface, may be less than a configured value to ensure synchronization when
TimingAdvCmdOptSwitch is set to OFF.
Table 1.1 Recommended parameter settings for VoIP services carried on the bearer with QCI 1
Parameter ID MML Parameter Description Unit Recommended
Command Value for Power
Saving-Preferred
Parameter Settings
It is recommended that DRX be disabled when performance-preferred parameter settings are used.
If DRX is enabled, and the value of the LongDrxCycle parameter for a QCI of 1 is set to a value
greater than 20 ms, the eNodeB does not activate downlink semi-persistent scheduling.
Default values in command output of MOD DRXPARAGROUP for a QCI of 1 or 5 include:
OnDurationTimer(PSF10)
DrxInactivityTimer(PSF5)
DrxReTxTimer(PSF8)
LongDrxCycle(SF20)
SupportShortDrx(UU_DISABLE)
Table 1.2 Recommended parameter settings for data services carried on the bearer with QCI 5,
QCI 6, QCI 8, or QCI 9
Parameter MML Unit Recommended Value Recommended Value
Command for Performance- for Power Saving-
Preferred Parameter Preferred Parameter
Settings Settings
DrxParaGroupId MOD - 3 3
DRXPARAG
EnterDrxSwitch ROUP - ON ON
LongDrxCycle Subframe SF40 SF320
OnDurationTimer Subframe PSF2 PSF10
DrxInactivityTimer Subframe PSF80 PSF100
DrxReTxTimer Subframe PSF8 PSF8
SupportShortDrx - UU_ENABLE UU_ENABLE
Recommended parameter settings for services carried on the bearer with QCI 5 are consistent
with parameter settings for services carried on the default bearer.
If false SR detection occurs on the live network, it is recommended that PuschDtxSwitch and
UlEnhancedSrSchSwitch be set to ON, decreasing the impact of false SR detection on network
KPIs.
Default values in output of MML commands for a QCI of 4, 6, 7, 8, or 9 are performance-
preferred parameter settings. Default values in output of MML commands for a QCI of 2 or 3 are
NULL.
UlSynTimerDynDrx s 20(10)
This parameter is set according
to information collection
results, and the signaling
reduction mode is used when
this parameter is set to 20.
Parameters added or modified in eRAN11.1 are marked in green, as shown in the following table.
DrxStatus Default value: Indicates whether to use None This parameter is used only
FALSE(FALSE) normal or special DRX. If when special DRX
this parameter is set to parameters are applied to
TRUE, ordinary DRX UEs with the SPID.
parameters are applied to
UEs with the SPID. If this
parameter is set to FALSE,
special DRX parameters
are applied to UEs with the
SPID.
QciPriorityForDr Default value: 9 Indicates the QCI-specific None This parameter is configured
x priority for selecting a based on operators'
DRX parameter group. A requirements, which affects
larger value of this the power saving
parameter indicates a lower performance of UEs.
priority. If the bearers for a
UE have multiple QCIs and
the DrxPolicyMode
parameter is set to
QCIPRIORITY(QCI
priority), the eNodeB
selects the DRX parameter
group for the UE based on
the QCI-specific priorities.
The value ranges from 1 to
254.
DrxAlgSwitch OFF(Off) Indicates the DRX switch. None This parameter is set to ON
as required. The cell-level
parameter DrxAlgSwitch in
the CellDrxPara MO is
added in eRAN11.1 to
replace this eNodeB-level
parameter. The configuration
interface in this version
supports synchronization and
delivery of the settings of
both the eNodeB-level and
cell-level parameters. If the
eNodeB-level parameter is
set to ON(On), the setting of
the eNodeB-level parameter
takes effect in all cells served
by the eNodeB. If the
eNodeB-level parameter is
set to OFF(Off), the setting
of the cell-level parameter
takes effect individually in
the cells.
ShortDrxSwitch ON(On) Indicates whether to enable None Short DRX cycles reduce the
short DRX cycles. traffic delay.
VolteGapDrxExc OFF(Off) Indicates whether the None The value ON(On) is
lusiveSwitch eNodeB delivers the intra- recommended in an inter-
RAT gap-assisted RAT or inter-frequency
measurement configuration networking scenario. This is
and DRX configuration to because the speech quality of
a UE performing a voice UEs that are performing
service at the same time. If voice services and start gap-
this parameter is set to assisted measurements
ON(On), the eNodeB should be ensured further in
instructs the UE to perform such scenarios. If this
intra-RAT gap-assisted parameter is set to ON, the
measurements and removes eNodeB removes the DRX
DRX configuration at the configuration that is
same time. If this delivered to a UE performing
parameter is set to a voice service when the
OFF(Off), the eNodeB can eNodeB delivers the gap-
deliver both configurations assisted measurement
to the UE. The configuration to the UE. This
DrxForMeasSwitch and reduces the possibility of
VolteGapDrxExclusiveSw packet loss problems that
itch parameters cannot be may occur due to untimely
set to ON(On) at the same scheduling in the cooperation
time. between gap-assisted
measurement and DRX.
However, this decreases the
power saving performance.
PucchIRCEnhan ON Indicates whether the None Setting this parameter has the
ce eNodeB determines following impacts:
interference strength. The The number of detected
eNodeB determines false alarms decreases.
interference strength only
when this switch is 1,
The uplink BLER
which ensures accurate decreases.
adaptive switchover The detected performance
between maximum ratio deteriorates slightly.
combining (MRC) and The ping delay is a little
interference rejection prolonged.
combining (IRC). This
parameter takes effect only
when PucchIrcSwitch of
the IrcSwitch parameter is
1.
CellUlschAlgo.U Off Indicates whether to enable None The VoLTE packet delay can
lEnhancedVoipS the estimation of traffic be shortened, the packet loss
chSw- volume dynamically rate can be reduced, and the
UlVoLTEDataSiz scheduled for VoLTE voice service quality can be
eEstSwitch services in the uplink. If improved.
this option is selected, the
eNodeB estimates the
uplink traffic volume that
is dynamically scheduled
for VoLTE services, which
shortens the VoLTE packet
delay, reduces the packet
loss rate, and improves the
voice service quality.
ENBCELLRSV Off Bit 2 of RsvdSwPara3 is None It is recommended that this
DPARA: used to specify whether to option be selected for
RsvdSwPara3=R optimize the periodic CQI- optimization of the decreased
svdSwPara3_bit2 related counters when UEs CQI caused by false alarms.
have entered the DRX
sleep period. When this bit
is set to 1, the eNodeB
optimizes the periodic
CQI-related counters if SR-
based scheduling exists
during the DRX sleep
period and the eNodeB
extends the DRX activation
period. In addition, this bit
is used to determine
whether to enable DRX
modes in CA primary and
secondary carrier cells to
be the same in eRAN11.1.
If this bit is set to 1, DRX
modes in primary and
SriPeriodAdaptiv NOQCIADAPTIV Indicates whether to enable None The E-RAB setup success
e E SRI period adaptation. If rate increases.
this parameter is set to
NOQCIADAPTIVE, the
probability of updating the
SR period decreases, the
occurrence probability of
the problem that abnormal
UEs do not send SRIs
occasionally on the live
network decreases, and the
E-RAB setup success rate
increases.
DataAmountStat 30 Indicates the length of the 20ms This parameter is used for
Timer UE traffic measurement length measurement.
period. The traffic volume
of a UE during this period
is measured. Based on the
measurement result, the
DRX algorithm decides
whether the UE should
enter or exit DRX.
CqiMask OFF(Off) Indicates whether the cqi- None If this parameter is set to
Mask IE can be set. ON, the number of CQI,
PMI, RI, and PTI reports sent
on the PUCCH decreases,
and the downlink data
transmission performance
deteriorates in a cell with
DRX enabled. In addition, if
UEs in compliance with
3GPP Release 9 or later do
not support the cqi-Mask IE,
the incompatibility problem
between UEs and the
eNodeB may occur. As a
result, the downlink
throughput decreases in a
cell with DRX enabled.
DrxPolicyMode DEFAULT(Default Indicates the policy for None This parameter is set to
) selecting a DRX parameter QCIPRIORITY only when
group for a UE for which the QCI priority is redefined.
the bearers have multiple In most cases, the QCI
QCIs based on the long priority cannot be redefined.
DRX cycle length, ARP,
and 3GPP-defined priority.
DrxStartOffsetO OFF(Off) Indicates whether to enable None It is recommended that this
ptSwitch optimized random parameter be set to ON when
distribution of SRS resources are not
to UEs performing
coverage-based GSM
measurement.
LongDrxCycleS Default value: Indicates the length of a Subfra The RFSP function is not
pecial SF10 long DRX cycle that is me applicable because UEs
applied only to non-power- support power saving, and
saving UEs of which therefore this parameter is set
subscriber profile ID for to a default value.
RAT/frequency priority
(RFSP) indexes are
contained in the RFSP
index set or to UEs of
which capability
information indicates that
they do not support power
saving.
OnDurationTime Default value: SF5 See above. Subfra See above.
rSpecial me
DrxInactivityTim Default value: See above. Subfra See above.
erSpecial SF10 me
SupportShortDrx Default value: See above. Subfra See above.
Special UU_DISABLE me
ShortDrxCycleS Default value: See above. Subfra See above.
pecial SF10 me
DrxShortCycleTi Default value: 1 See above. Subfra See above.
merSpecial me
LongDrxCycleF SF320 Indicates the long DRX Subfra Set this parameter to the
orAnr cycle for intra-RAT me recommended value.
automatic neighbor relation
(ANR).
LongDRXCyclef LTE- Indicates the long DRX Subfra Set this parameter to the
orIRatAnr >GERAN:SF2560 cycle for inter-RAT ANR. me recommended value.
Other network:
SF1280
DrxInactivityTim PSF200 Indicates the length of the Subfra Set this parameter to the
erForAnr DRX inactivity timer for me recommended value.
ANR.
ReptSyncAvoidI NOT_CFG Indicates whether all None This parameter is set based
nd synchronization procedures on whether UE problems
repeatedly initiated by a need to be prevented on the
UE having incompatibility live network.
problems can trigger the
RRC connection
reconfiguration procedure.
DRX for special UEs in RRC_CONNECTED mode is customized for Deutsche Telekom, and
currently is not configured for other operators. Parameters used for DRX for special UEs in
RRC_CONNECTED mode are set for a certain terminal specified based on the subscriber profile ID
(SPID). This terminal may not support DRX, or whether this terminal supports DRX cannot be
identified, and therefore DRX parameters are set for this terminal according to operators'
requirements. If DRX for special UEs in RRC_CONNECTED mode is configured for other operators,
it is recommended that parameters be set to default values.
Step 6 Run the following command to bind a DRX parameter group to a QCI for a cell:
MOD CELLSTANDARDQCI: LocalCellId=0, Qci=QCI9, DrxParaGroupId=3;
Step 7 (Optional) If QCI 10 has been configured for a cell, run the following command to bind a
DRX parameter group to QCI 10 for the cell:
MOD CELLEXTENDEDQCI: ExtendedQci=10, LocalCellId=0, DrxParaGroupId=3;
Step 8 If the long cycle is greater than or equal to 80 ms, run the following command to modify
the time alignment timer:
MOD TATIMER: LocalCellId=0, TimeAlignmentTimer=SF10240;
Step 9 Run the following command to reduce RRC connection reconfiguration signaling generated
by UEs frequently entering and exiting DRX mode and prevent UEs from entering and
exiting DRX mode based on traffic statistics:
MOD CELLDRXPARA: LocalCellID=0,FddEnterDrxThd=1000,FddExitDrxThd=1000;
Step 10 Run the MOD DRXPARAGROUP command to turn on the switch of a specified DRX
parameter group and configure DRX parameters.
Performance-preferred parameters
MOD DRXPARAGROUP: LocalCellId=0, DrxParaGroupId=3, EnterDrxSwitch=ON,
OnDurationTimer=PSF2,DrxInactivityTimer=PSF80, DrxReTxTimer=PSF8,
LongDrxCycle=SF40, SupportShortDrx=UU_ENABLE,ShortDrxCycle=SF20,
DrxShortCycleTimer=1;
Step 11 (Optional) To allow the eNodeB to select DRX parameters based on QCI priorities, run the
following commands to specify a DRX parameter selection policy and set DRX priorities
used in DRX parameter selection:
MOD CELLDRXPARA: LocalCellId=0, DrxPolicyMode=QCIPRIORITY;
MOD CELLSTANDARDQCI: LocalCellId=0, Qci=QCI9, QciPriorityForDrx=9;
----End
EnterDrxSwitch - ON
EnterDrxSwitch - ON
EnterDrxSwitch - ON
EnterDrxSwitch - ON
After DRX is disabled, packets are sent more discretely at a longer interval, causing a
decrease of the downlink throughput in a cell based on traffic statistics. The average packet
size is smaller, and the modulation and coding scheme (MCS) indexes for small packets are
smaller. As a result, the average MCS index decreases.
1 Throughput and packet loss rate UEs' DRX activation period and This issue has been resolved
at the Packet Data Convergence sleep period are not considered in eRAN7.0.
Protocol (PDCP) layer when the priority of the product
deteriorate when DRX is queue is updated using a hash
enabled in a cell in high-load algorithm, greatly increasing the
scenarios (the number of scheduling delay of UEs processing
RRC_CONNECTED UEs is large-packet services. As a result,
greater than 140) in office Y in throughput is affected.
country E.
2 Two consecutive DTXs are UEs are supposed to be in the DRX This issue has been resolved
detected in downlink sleep period if two consecutive in V100R007C00SPC150
scheduling in office H in DTXs are detected in downlink (for details, see section
country C, and the eNodeB scheduling with DRX enabled, and 4.2"Feature Activation Risks
suspends downlink scheduling, therefore scheduling is stopped and and Workarounds"). In a
causing an insufficient number will start again within the next On version earlier than
of downlink scheduling times Duration period, causing a decrease V100R007C00SPC150, you
and a decrease in throughput. in the number of scheduling times can change thresholds for
and throughput. entering and exiting DRX to
300 and 800 respectively in
drive test (DT) scenarios to
temporarily prevent the
impact on throughput.
3 LBBPc boards are configured When DRX is enabled and This is a configuration issue,
on the eNodeB in office N in thresholds for entering and exiting which can be resolved by
country N. If the number of DRX are set to 300 and 800, changing thresholds for
UEs served by a single eNodeB respectively, a large number of entering and exiting DRX to
reaches 400 and the CPU usage RRC Connection Reconfiguration 1000 and 1000, respectively.
of a main control board exceeds messages are generated over the air
60%, the main control board is interface, causing CPU overload on
overloaded when DRX is the main control board. The
DTs due to E5375 or E3276 terminals, including uplink error terminals be replaced.
terminal issues in office J in packets, downlink error packets,
country C. and issues when real measurement
policies configured for serving cells
are reported.
8 OnDurationTimer is set to In scenarios where the length of the Set the length of the On
2ms (a short time) in office W On Duration timer is short, RRC Duration timer to the
in country C, decreasing the Connection Release messages are recommended value.
RRC connection sent without scheduling due to
reestablishment success rate. product restrictions.
9 Due to design restrictions of Design restrictions of QualComm There are no workarounds
QualComm chips in office N in chips are described as follows: for design restrictions on
country C, the ping delay If Long DRX Cycle is set to UEs.
increases by tens of 320ms, UEs enter a deep sleep state
milliseconds during tests after when no data needs to be sent
DRX is enabled. during a long time, and more than
50 ms is required when the UEs
recover from the deep sleep state,
causing an increase in the ping
delay.
10 After dynamic DRX is enabled After dynamic DRX is enabled, the Detailed descriptions have
in office F in country L, the timer used to govern the period in been supplemented in
number of state switching which the eNodeB maintains uplink eRAN8.1 Feature
between synchronization and synchronization is set to a value 2s Documentation.
out of synchronization longer than a configuration value
increases due to a DRX based on algorithm design (an
parameter delivery mechanism uplink synchronization protection
in out-of-synchronization mechanism is extended), causing a
mode. In addition, the UEs' decrease in the out of
sleep period is extended, synchronization probability. In
causing an increase in the addition, the probability of false SR
probability of false SR detection increases, causing an
detection and a decrease in CQI increase in CQI detection errors and
reporting without affecting user the proportion of CQIs in 0 to 4
experience. orders and a decrease in the average
CQI value.
11 The RRC connection Without UE PDP context during Optimize weak coverage on
reestablishment success rate connection reestablishment, the the live network.
decreases when power saving RRC connection reestablishment
mode for DRX is enabled in success rate is low. At this time, if
office X in country C. UEs are in DRX mode, the UE
inactivity timer expires in weak-
coverage areas. If RRC Connection
Release messages are initially sent
and resent in DTX mode in the
DRX On Duration period,
subsequent RRC Connection
Release messages cannot be resent
until UE instances on the system
side are released 450 ms after
length of the UE inactivity terminals, including iPhone 5 and resolve the terminal issue,
timer is increased to 200s iPhone 5S, but have nothing to do see chapter 1"Feature
before dynamic DRX is with dynamic DRX. When Application Policies."
enabled, and therefore the UeInactiveTimer is set to 200s,
service drop rate calculated UEs enter RRC_IDLE mode
using a new formula obviously without notifying the network side,
deteriorates, increasing from which will be calculated as the
0% to 1.5%. service drop on the network side.
Because UEs enter RRC_IDLE
mode automatically, UE service
experience is not affected.
21 In office X in country C, This is a known issue. When the For details about issue
because SRS is disabled by number of RRC_CONNECTED resolving, see section
default in eRAN6.0 and UEs on the LBBPc board exceeds 4.2"Feature Activation Risks
eRAN8.0, KPI deterioration on 100 and that on the LBBPd board and Workarounds."
the control plane is caused by exceeds 300 in a single cell, and
occupied resources in uplink DMRS TA and DRX are enabled
synchronization (DMRS TA) simultaneously, the DRX sleep
when the number of period is not considered in DMRS
RRC_CONNECTED UEs in a TA scheduling, causing KPI
single cell exceeds 100. When deterioration on the control plane.
DMRS TA and DRX are
enabled simultaneously, the
DRX sleep period is not
considered in DMRS TA
scheduling, causing KPI
deterioration on the control
plane. (DMRS is short for
demodulation reference signal,
and TA is short for tracking
area.)
22 Operators in office V complain Voice packet scheduling is not Reply to the operators that
that voice packet loss occurs performed in time when DRX- DRX-related parameter
when DRX-related parameter related parameter settings settings recommended by
settings recommended by Apple recommended by Apple are used. Apple have been applied in
are used. V100R007C00SPC230.
23 At 17:50 on February 9 in If ShortDrxCycle and This issue is caused by
office L in province Z, the DRXShortCycleTimer are set to terminals. Change the values
number of inter-frequency SF80(80 subframes) and 1, more of ShortDrxCycle and
handovers (from FDD to TDD) unnecessary signals are measured DRXShortCycleTimer to
increases, and the handover by UEs, such as iPhone 6, iPhone 6 SF40(40 subframes) and 2
success rate decreases, Plus, and Samsung S5, on the live to resolve this issue.
remaining at a level of about network. As a result, the number of
85% after DRX is enabled. handovers from FDD to TDD
increases, and the handover success
rate decreases.
24 After DRX is disabled on high- Scheduling is stopped after two Explain principles to
quality networks in office D in consecutive DTXs are detected operators.
province J, the RRC connection when DRX is enabled, then
reestablishment rate decreases RRC Connection Release
Version Workaround
Such issue has been resolved in V100R007C00SPC160. You can set RsvdSwPara1_bit17
in the MOD ENBRSVDPARA command to determine whether the workarounds take
effect. You can also set DrxBasedSriGapOptSwitch to resolve such issue in eRAN8.0.
If the number of UEs in a cell and uplink traffic volume exceed product specifications,
BBPs' CPU load will increase when DRX is enabled, causing a decrease in the number of
PRBs occupied by scheduling data in the uplink. Such issue has been resolved in
V100R007C00SPC160. Check whether a UE is in the gap-assisted measurement period or
DRX sleep period when selecting the UE for scheduling. If the UE is in the DRX sleep
period and cannot be scheduled, deselect the UE. If the UE has been selected, remove it
from candidate UEs.
When IP RAN inter-site CA is performed in light-traffic small-packet scenarios using the
SpeedTest, the SCC load activation threshold is small (configured in comparison testing
scenarios), traffic is light at the start phase, OnDurationTimer is set to 10ms, and the
downlink CA preallocation period is set to 10 ms fixedly. Therefore, when downlink
preallocation is performed within the sleep period each time, UEs enter the sleep period for
a long time and cannot be scheduled, affecting throughput. This issue has been resolved by
determining On Duration period in advance by SCC during real UE scheduling and
performing scheduling adjustment in the On Duration period by default in eRAN8.1. In
versions earlier than eRAN8.1, you can increase the length of the On Duration timer to be
longer than the downlink preallocation period, such as 20 ms, to resolve this issue.
Without UE PDP context during connection reestablishment, the RRC connection
reestablishment success rate is low when DRX is enabled. At this time, if UEs are in DRX
mode, the UE inactivity timer expires in weak-coverage areas. If RRC Connection Release
messages are initially sent and resent in DTX mode in the DRX On Duration period,
subsequent RRC Connection Release messages cannot be resent until UE instances on the
system side are released 450 ms after message scheduling in UEs' next DRX On Duration
period. Therefore, if Long DRX Cycle is set to a larger value in weak coverage scenarios,
the number of times RRC Connection Release messages are resent in DTX mode decreases,
the probability that UEs cannot receive RRC Connection Release message increases, and
the probability of no UE PDP context during connection reestablishment increases. Such
issues can be resolved by optimizing weak coverage on the live network.
Disable SRS if the number of RRC_CONNECTED UEs in a cell exceeds 100 with DRX
enabled. This may cause deterioration of the RRC setup success rate, E-RAB setup success
rate, and service drop rate.
With DRX enabled and SRS disabled, the DRX sleep period is not taken into consideration
when candidate UEs for uplink scheduling are to be selected. As a result, UEs which cannot
be scheduled in the DRX sleep period are selected by mistake. When the number of
RRC_CONNECTED UEs in a single cell exceeds 100, the RRC setup success rate, E-RAB
setup success rate, and service drop rate deteriorate.
The workaround is to enable SRS on the eNodeB side. However, MCS index for UEs'
uplink scheduling decreases and UEs' uplink peak throughput decreases after workaround
implementation. The final solution is described as follows:
Perform the following steps in sequence on macro and LampSite eNodeBs:
Step 12 Upgrade eNodeBs to BTS3900 V100R008C00SPC300 or later BTS3900 V100R008C00
versions, or upgrade eNodeBs to BTS3900 V100R009C00SPC150 or later versions.
Step 13 Enable SRS by running the following command:
MOD SRSCFG: LocalCellId=0, SrsCfgInd=BOOLEAN_TRUE;
Step 14 Turn on the SRS adaption switch by running the following command:
In BTS3900 V100R008C00SPC300 or later BTS3900 V100R008C00 versions
MOD ENBCELLRSVDPARA: LOCALCELLID=0, RSVDSWPARA0=RsvdSwPara0_bit19-1;
----End
----End
----End
VoLTE services are performed on some networks using QCI 1 parameter settings 4/4/40
recommended by Apple, causing voice packet loss. (4 is the length of the On Duration
timer, 4 is the length of the DRX inactivity timer, and 40 is the length of the DRX long
cycle timer.)However, parameter settings 10/80/40 recommended by Huawei are optimal,
ensuring voice quality for VoLTE services first, and then power saving. Sending two voice
packets together by iPhones at an interval of 40 ms is improper, which may easily cause
packet loss. One packet is sent at an interval of 20 ms for other models of mobile phones. If
operators require parameter settings recommended by Apple, perform an upgrade towards
V100R007C00SPC230, in which parameter adaptation is done.
The inactivity timer for UEs with the QCI 1 bearer is independently configured in dynamic
DRX scenarios in eRAN11.0. Assume that dynamic DRX is enabled and the length of the
UE inactivity timer indicated by UeInactTimerDynDrxQci1 in the RrcConnStateTimer
MO is less than the length of the connection release timer configured on the EPC (the
ongoing call is released after the connection release timer expires). In this case, voice
services cannot be set up if the following conditions are met:
The inactivity timer of a calling UE has expired before the called UE answers a call,
which causes the calling UE to enter the idle mode.
The calling UE reselects another VoLTE-incapable cell with a higher priority.
UeInactTimerDynDrxQci1 can be set to a larger value so that the calling party can be
paged after the called party answers the call, which prevents a call setup failure.
A delay variation of 0 to 40 ms is introduced in uplink packet sending and in the downlink
on all UEs including iPhone 6s under Apple-provided DRX-related parameter settings with
the QCI of 1. If UEs are not subject to delay variation processing, an end to end (E2E)
delay variation of 80 ms will be introduced, causing small MOS values. Enable HARQ
pending and data volume estimation optimization functions to increase MOS values. In
HARQ pending, the uplink voice BSR scheduling algorithm is optimized, and HARQ-
responded transmission time interval (TTI) for SR-based scheduling within the On Duration
period is used to perform BSR scheduling on UEs after the On Duration period is over. This
function will be incorporated into eRAN11.1. In data volume estimation optimization, the
voice data volume increasing in the buffer area within a period of time is estimated and the
data volume is increased in the next scheduling based on the uplink voice packet receiving
interval and packet size under Apple-provided DRX-related parameter settings. This
function has been incorporated into eRAN8.1.
Voice data volume estimation-based uplink scheduling optimization is controlled by
UlVoLTEDataSizeEstSwitch under the
CELLULSCHALGO.UlEnhancedVoipSchSw parameter.
services are performed and gap-assisted measurement starts. This parameter has not been
used on the live network. After VolteGapDrxExclusiveSwitch is set to ON(On), it is
recommended that changes in main KPIs be observed to identify other unknown impacts.
The RRC connection reestablishment rate may increase when DRX is enabled. Therefore,
In eRAN11.1, GapDrxExclusiveSwitch can be set to ON for the eNodeB to solve the
problem that gap-assisted measurement configuration and DRX configuration cannot be
delivered at the same time for voice and data services, without distinguishing the QCI. This
can optimize inaccurate inter-frequency measurement, reduce the number of RRC
connection reestablishments, and reduce the reestablishment-access ratio. However, power
generated during the period cannot be saved.
The reestablishment-access ratio is calculated using the following formula:
Reestablishment-access ratio = Number of RRC connection reestablishment requests/
(Number of RRC connection reestablishment requests + Number of RRC access requests)
This parameter has not been used on the live network. After GapDrxExclusiveSwitch is set
to ON(On), it is recommended that changes in main KPIs be observed to identify other
unknown impacts.
7 Troubleshooting
EnterDrxSwitch MOD - ON ON
DRXPARAGROU
LongDrxCycle P Subframe SF40 SF320
OnDurationTimer Subframe PSF2 PSF10
DrxInactivityTimer Subframe PSF80 PSF100
DrxReTxTimer Subframe SF8 SF8
SupportShortDrx - UU_ENABLE UU_ENABLE
A larger value of LongDrxCycle indicates a longer UE's sleep period and a longer delay. If
the value of LongDrxCycle is too large, the CQI reporting period is lengthened and
therefore the throughput decreases for the scheduling and multiple-input multiple-output
(MIMO) features, the handover success rate decreases, and the service drop rate increases.
If the value of LongDrxCycle is too small, UE power saving is insignificant. If
TimingAdvCmdOptSwitch is set to ON, it is recommended that LongDrxCycle be set to
a value less than or equal to 320ms. Otherwise, TAs may not be sent in time and the uplink
synchronization performance of UEs is affected. For smart terminals with a strict power
consumption requirement, the value SF320(320 subframes) is recommended, and
long DRX cycle. When TimingAdvCmdOptSwitch is set to ON, you are advised to set
TimeAlignmentTimer to sf10240. If TimeAlignmentTimer is set to a smaller value, UEs
in the DRX mode are more likely to enter the out-of-synchronization state in the uplink.
FddEnterDrxThd
FddExitDrxThd
Qci MOD CELLSTANDARDQCI
DrxParaGroupId
ExtendedQci MOD CELLEXTENDEDQCI
DrxParaGroupId
DrxParaGroupId MOD DRXPARAGROUP
EnterDrxSwitch
LongDrxCycle
OnDurationTimer
DrxInactivityTimer
DrxReTxTimer
SupportShortDrx
ShortDrxCycle
DrxShortCycleTimer
LongDrxCycleSpecial MOD DRX
OnDurationTimerSpecial
DrxInactivityTimerSpecial
SupportShortDrxSpecial
ShortDrxCycleSpecial
DrxShortCycleTimerSpecial
TimeAlignmentTimer MOD TATIMER
The following figure shows that reconfiguration message sent from the eNodeB to the UE
contains the DRX parameter settings. This indicates that the eNodeB instructs the UE to
enter the DRX mode. As shown in the following figure, the setup field under drx-Config
indicates that a UE is instructed to enter the DRX mode, whereas the release field indicates
that the UE is instructed to exit the DRX mode.
Identify whether UEs support DRX. A UE reports its capability during initial access. Bit 5
in the feature group indicators field of the RRC_UE_CAP_INFO message specifies
whether the UE supports DRX. (Bit 5 indicates that the UE supports a long DRX cycle,
whereas bit 4 indicates that the UE supports a short DRX cycle. The value of bit 4 is
ignored when that of bit 5 is invalid.) Seen from the message parsed in the following figure,
the value of bit 5 is 0, indicating that the UE does not support DRX.
7.1.7.3 Method of Checking Why UEs Cannot Enter the DRX Mode
Step 1 Run LST DRX to query whether the DRX switch is on. If the DRX switch is off, set
DrxAlgSwitch to ON.
Step 2 Check whether UEs support DRX. For details, see section 7.1.7.2"Method for DRX
Problem Identification."
Step 3 Query thresholds for entering and exiting DRX and value of TTI for data transmission of
UEs. If the value of TTI for data transmission is greater than or equal to the threshold for
exiting DRX, UEs do not enter the DRX mode.
The following figure shows DRX measurement-based IFTS 97 tracing data. If the value of
TTI for data transmission is greater than or equal to the threshold for exiting DRX, UEs do
not enter the DRX mode.
Step 4 Check whether a number of bearers with different QCIs are set up for the UE. If yes, check
whether each bearer supports DRX. Run LST CELLSTANDARDQCI to query the DRX
parameter group ID for the QCI in use.
Step 5 Run LST DRXPARAGROUP to query whether DRX is enabled based on the queried
DRX parameter group ID. If DRX is disabled, you need to enable DRX or set up other
bearers of QCIs supporting DRX.
----End
One DRX measurement message is recorded each time the DRX measurement timer
expires. When the value of the usSendingTTIRatio field is less than or equal to that of the
usThEnterDrx field, and the value of the ucRptType field is not equal to 1, the ucRptType
= 1 message is reported to instruct the UE to enter the DRX mode. When the value of the
usSendingTTIRatio field is greater than or equal to that of the usThExitDrx field, and the
value of the ucRptType field is not equal to 0, the ucRptType = 0 message is reported to
instruct the UE to exit the DRX mode. The current conditions are maintained in other cases.
Reporting ucRptType = 1 or ucRptType = 0 message does not indicate that the UE enters or
exits the DRX mode immediately. After the DRX parameters are delivered from the
eNodeB to the UE and the UE makes a response that the DRX reconfiguration is
completed, the eNodeB instructs the UE to enter or exit the DRX mode. The following table
lists descriptions of fields in 97 tracing.
usSendingTTIRatio Ratio of uplink scheduling The value of this field is calculated using the following
times to downlink formula:
scheduling times within the Value of usSendingTTIRatio = Total number of uplink
last measurement period and downlink scheduling times/Measurement period x
1000
Counter Description
Counter Description
Counter Description
Distribution of packet transmission intervals for UEs in a cell can be taken as a reference to
analyze traffic model changes when DRX is enabled.
Counter Description
L.Traffic.PktInterval.Num.Index0 Measures the number of times the packet transmission interval for
a UE in a cell ranges within index 0.
L.Traffic.PktInterval.Num.Index1 Measures the number of times the packet transmission interval for
a UE in a cell ranges within index 1.
L.Traffic.PktInterval.Num.Index2 Measures the number of times the packet transmission interval for
a UE in a cell ranges within index 2.
L.Traffic.PktInterval.Num.Index3 Measures the number of times the packet transmission interval for
a UE in a cell ranges within index 3.
L.Traffic.PktInterval.Num.Index4 Measures the number of times the packet transmission interval for
a UE in a cell ranges within index 4.
L.Traffic.PktInterval.Num.Index5 Measures the number of times the packet transmission interval for
a UE in a cell ranges within index 5.
L.Traffic.PktInterval.Num.Index6 Measures the number of times the packet transmission interval for
a UE in a cell ranges within index 6.
L.Traffic.PktInterval.Num.Index7 Measures the number of times the packet transmission interval for
a UE in a cell ranges within index 7.
L.Traffic.PktInterval.Num.Index8 Measures the number of times the packet transmission interval for
a UE in a cell ranges within index 8.
L.Traffic.PktInterval.Num.Index9 Measures the number of times the packet transmission interval for
a UE in a cell ranges within index 9.
Counter Description
Counter Description
Modification The total number of E-RABs of UEs when the UEs switch from the
Point uplink-synchronized state to the uplink out-of-synchronization state is
added in the denominator, and the total number of released E-RABs of
UEs in the uplink out-of-synchronization state, which is repeatedly
calculated, is subtracted.
Modification After the length of the UE inactivity timer is increased, the number of
Cause some normal E-RAB releases is converted into that of E-RABs of UEs in
the uplink out-of-synchronization state. Therefore, the total number of E-
RABs of UEs when the UEs switch from the uplink-synchronized state
to the uplink out-of-synchronization state is added in the denominator.
Service RRC Calculation (L.RRC.ConnReq.Succ.Mt + L.RRC.ConnReq.Succ.MoData +
Connection Formula L.RRC.ConnReq.Succ.EMC + L.RRC.ConnReq.Succ.HighPri +
Succ Rate (%) L.RRC.StateTrans.Unsyn2Syn)/(L.RRC.ConnReq.Att.Mt +
L.RRC.ConnReq.Att.MoData + L.RRC.ConnReq.Att.EMC +
L.RRC.ConnReq.Att.HighPri + L.RRC.StateTrans.Unsyn2Syn)
Modification The modification cause is the same as that of the RRC setup success rate.
Cause The number of E-RAB setup times may be greater than that of RRC
connection setup times. Therefore, in eRAN7.0, values of the L.E-
RAB.StateTrans.Unsyn2Syn.Att and L.E-
RAB.StateTrans.Unsyn2Syn.Succ counters are added to replace
denominator and numerator in the calculation formula.
Analysis of 89 Tracing
Cell-level 89 tracing indicates DRX mode tracing, as shown in the following figure.
The ucTracEvent field is recorded once an event is triggered. DRX mode tracing is often
used for DRX problem location. The current DRX mode is determined based on each DRX
event, and problem identification is performed in combination with tracing items, including
cell-level 34 tracing for downlink scheduling, cell-level 50 tracing for uplink scheduling,
cell-level 111 tracing for the TA state, and cell-level 14 tracing for random access. The
following table lists specific descriptions of each field.
Analysis of 98 Tracing
IFTS tracing indicates UE-level tracing and is used for tracing data of the first UE that
accesses a cell. A record is made at every TTI. The data volume is large, and therefore flow
bitOnDurationTim Running identifier of the 0: Indicates that the OnDuration timer is stopped.
erOn OnDuration timer 1: Indicates that the OnDuration timer is running.
bitInactivityTimer Running identifier of the 0: Indicates that the inactivity timer is stopped.
On inactivity timer 1: Indicates that the inactivity timer is running.
bitSrPending Indication for SR receiving 0: Indicates that an SR is not received.
1: Indicates that an SR is received.
ulCellDrxActiveCnt Activity period of all UEs in the DRX mode in a cell (unit: s)
ulCellDrxSleepCnt Sleep period of all UEs in the DRX mode in a cell (unit: s)
Because preallocation is not performed in DRX mode, the ping delay will increase by
a maximum of 8 ms. Therefore, first run the LST CELLALGOSWITCH command
to check whether the preallocation switch is turned on.
The following figure illustrates the uplink ping delay when the DRX switch is on. If
the long DRX cycle is 10 ms, the short DRX cycle is not used, and the length of the
On Duration timer is 2 ms, then the uplink ping delay is between (8 + x) ms and (25 +
x) ms.
can also obtain UE logs for analysis. For QualComm chips, you can use the QXDM
tool to obtain UE logs and use other tolls such as Wireshark and QCAT to analyze the
ping procedure segment by segment. Pay more attention to the following signaling on
the UE side:
Signaling generated when the UE enters the sleep state
MSG [09509/02/07]LTE ML1/High/DLM 12:11:08.333 lte_ml1_dlm_cfg.c 03401 DLM: cdrx req
ON->OFF
Problem Analysis
4. Identify the problem.
The operation log shows that there are no operations except a DRX operation at the
time point when the throughput changes.
The air interface log shows that DRX parameters are delivered.
The probe log shows that bit 5 is 1, indicating that the UE supports DRX.
OnDurationTimer PSF10
DRXInactivityTimer PSF100
DRXReTxTimer PSF8
LongDRXCycle SF320
ShortDRXCycle SF80
DRXShortCycleTimer 1
(2) When DRX is enabled, the number of RCU terminal scheduling times is far below
1000 per second (the maximum value) though the RCU is scheduled every second.
(3) The time point 9:32:18 (where the number of scheduling times is far below the
maximum value) is selected for further analysis. In frame 33, two consecutive
DTX states are detected and the sleep period starts. Then, there is not downlink
scheduling until the On Duration in frame 63 arrives. The downlink scheduling
interruption time is about (63 – 33) x 10 = 300 ms.
(4) In frame 73, three consecutive DTX states are detected and the sleep period starts.
Then, there is not downlink scheduling until the On Duration in frame 95 arrives.
The downlink scheduling interruption time is about (95 – 73) x 10 = 220 ms.
(5) In frame 105, two consecutive DTX states are detected and the sleep period starts.
Then, there is not downlink scheduling until the On Duration in frame 127 arrives.
The downlink scheduling interruption time is about (127 – 105) x 10 = 220 ms.
(6) The analysis results at other time points are the same as that at 9:32:18. That is,
two or more consecutive DTX states are detected and therefore the sleep period
starts, which affects the number of downlink scheduling times.
10. Conclusion
The problem in this office is the same as that in Chengdu. The eNodeB detects
two consecutive DTX states, considers the UE to be in the sleep state as there is
no response to downlink scheduling, and stops downlink scheduling. The eNodeB
has to wait for the next On Duration for scheduling and therefore the number of
downlink scheduling times decreases. All E392, E3276, and RCU terminals have
the same problem. For the RCU terminal, there is a higher probability that two
consecutive DTX states are detected and a more obvious decrease in the number
of downlink scheduling times.
In DRX mode, the eNodeB stops downlink scheduling after detecting two
consecutive DTX states. This mechanism is not rational and will be optimized in
later versions to reduce the impact of two consecutive DTX states on downlink
scheduling.
Problem Analysis
1. Check the preallocation switch.
Because preallocation is not performed in DRX mode, the ping delay will increase by
a maximum of 8 ms. However, preallocation is not the only cause of long delays.
There should be other causes.
1 OnDurationTimer 10 ms
2 DrxInactivityTimer 100 ms
3 DrxRetransmissionTimer 8 ms
4 LongDrxCycle 320 ms
5 ShortDrxCycle 20 ms
6 ShortDrxCycleTimer 1
7 TAtimer 10240 ms
(2) The terminal enters the sleep state at the time marked in red.
MSG [09509/15] LTE ML1/SleepMGR 05:32:03.538 lte_ml1_sleepmgr_stm.c 09515
[LTE_ML1_SLEEPMGR_STM] ONLINE_SLEEP_WAIT -> SLEEP (Entry Function)
(6) The application layer of the terminal receives the ping reply.
The previous log shows that the terminal enters the sleep state after DRX is enabled.
To send an SR, the terminal needs to transit from the sleep state to the active state. The
transition consumes 43 ms, as shown in the following figure.
When DRX is disabled, the terminal immediately sends an SR after receiving a ping
request from the application software. In this case, the internal processing delay is very
small, as shown in the following figure.
4. Conclusion
The previous analysis shows that long delays and variations are related to DRX. In
DRX mode, the terminal enters the sleep state if there is no data to transmit. When the
terminal needs to originate an uplink ping service, it must be awakened first and then it
can send an SR. As the wakening consumes a long time, the ping delay is long. This
problem is due to the design of QualComm chips.
Problem Analysis
5. Identify the problem.
Only parameters related to DRX are adjusted and the adjustment time is consistent
with the change time shown in the previous figures.
EnterDRXSwitch ON ON
LongDRXCycle SF320 SF320
OnDurationTimer PSF2 PSF10
DRXInactivityTimer PSF80 PSF100
DRXReTxTimer SF8 SF8
SupportShortDRX UU_ENABLE UU_ENABLE
The parameters on the live network are adjusted as listed in the previous table, where
the two settings are different from those recommended by Huawei, and the
TimeAlignmentTime value changed from 5120 to 10240. Setting OnDurationTimer
and DrxInactivityTimer to smaller values and LongDrxCycle to a large value will
increase the proportion of the UE staying in the sleep state and reduce UE power
consumption, but increase system latency and reduce scheduling opportunities. In
addition, changing the TimeAlignmentTimer value from 5120 to 10240 leads to
deterioration in uplink timing performance and may result in scheduling issues.
The parameters are adjusted again as recommended by Huawei for power saving.
Then, the RRC connection reestablishment success rate restores to a normal value. The
following figure shows the test result.
You can take X Solutions for Access and Paging Performance Optimization in the LTE
Network as a reference for analysis. The following provides the main analysis
procedure.
The traffic statistics show that reestablishment failures are mainly due to no contexts.
The result of Uu tracing on a site from 9:00 to 11:00 shows that reestablishment
failures on this site accounts for a large proportion.
Item Reestablishment on Reestablishment on Total
This Site Other Sites
The tracing result shows UE 267910 is released due to inactivity after access cell 3. A
UE's reestablishment request is rejected. This UE is UE 267910, according to the PCI
and CRNTI of the cell where the UE camps before reestablishment. After the request
is rejected, the UE sends another reestablishment request, with the same TMSI.
The previous analysis shows that the UE Inactivity Timer expires and the UE does not
receive the RRC connection release command sent by the eNodeB. However, why
does the UE not receive the release command from the eNodeB after the eNodeB
rejects the reestablishment request due to no contexts?
The result of internal message tracing on top three sites shows that six cells are set up
on one LBBPd3. In this case, three scheduling threads each correspond to two cells but
each thread can schedule only one cell in each TTI. Among the 2 ms On Duration, one
millisecond is used for scheduling system information (as the SIB2 broadcasting
period is 160 ms, shorter than the long DRX cycle 320 ms), and the other millisecond
is used for scheduling one cell's DCCH signaling. Therefore, some cells' DCCH
signaling cannot be scheduled in time. The wait time for RRC connection release at L3
is equal to the long DRX cycle plus 450 ms. If the DRX Inactivity Timer does not
start, there are only two to three scheduling chances. As a result, some RRC
connection release commands are not delivered.
For power-saving-preferred parameter settings, the On Duration Timer is set to 10 ms
and the problem will not occur. For performance-preferred parameters, the problem
will also not occur because the long DRX cycle is 40 ms, system information is
scheduled every four long DRX cycles, and two cells can be scheduled in the 2 ms On
Duration.
10. Conclusion
DRX parameters are not set as recommended by Huawei. The On Duration Timer is
too short, and some cells' DCCH signaling cannot be scheduled in time. After the UE
Inactivity Timer expires, some RRC connection release commands of the eNodeB are
not delivered in time and therefore they are not received by the UE. When the UE
sends a reestablishment request, the eNodeB rejects the request due to no contexts.
Problem Analysis
11. Identify the problem.
After DRX is enabled, the RRC connection establishment success rate decreases, the
average and maximum usage of the main control board CPU increases, and the
maximum number of online users decreases. (The changes in sensitive performance
indicators are not presented.) The problem is closely related to traffic volume, and it
disappears after the traffic volume decreases.
12. Analyze the impact of RRC connection reconfiguration signaling for DRX on the main
control board CPU usage.
With the growth in traffic volume after DRX is enabled, the number of RRC
connection reconfiguration messages for DRX significantly increases and the main
control board CPU usage also increases. These symptoms are indicated by the changes
of the main control board CPU usage and the value of L.Cdrx.Enter.Num. The
changes in sensitive performance indicators are not presented.
Theoretically, if the FddEnterDrxThd parameter is set to 300 and the
FddExitDrxThd parameter is set to 800, an RRC connection reconfiguration message
and a reconfiguration complete message are required each time a UE enters DRX
mode and the CPU usage significantly increases. Under these parameter settings, the
UE starts measurement after accessing the network. In the period defined by the
DataAmountStatTimer parameter, if the percentage of TTIs with data scheduling
exceeds 30%, an RRC connection reconfiguration message is delivered to instruct the
UE to enter DRX mode. As the measurement period is far longer than the access delay,
this message cannot be combined with the reconfiguration message in the access
procedure and must be delivered separately.
After the main control board CPU is overloaded, a large number of RRC connection
establishment failures occur. The reason is that the processing delay is prolonged due
to CPU overload and the Msg4 is probably delivered too late. As a result, the RRC
connection requests are rejected or even discarded in flow control. These symptoms
are indicated by the peak usage of the main control board CPU, the number of RRC
connection establishment failures, and the number of RRC connection requests
discarded in flow control (L.RRC.ConnReq.Msg.disc.FlowCtrl).
A large number of RRC connection establishment failures results in a large number of
Msg3 for attempts. This further increases the main control board CPU usage, triggers
more intensive flow control, and decreases the RRC connection establishment success
rate. This vicious cycle leads to a high CPU usage and a large number of RRC
connection establishment failures during peak hours.
13. Locate root causes.
After the FddEnterDrxThd and FddExitDrxThd parameters are set to the
recommended value 1000, the RRC connection reconfiguration signaling for DRX is
integrated into the first reconfiguration message in the access procedure and will not
consume additional CPU resource. This is confirmed by the air interface signaling
tracing result. The changes in sensitive performance indicators are not presented.
Provided that the traffic volume remains unchanged and the value of
L.Cdrx.Enter.Num is in the same order of magnitude, there is not a significant increase
in the main control board CPU usage after those two parameter values are changed. As
a result, there are not a large number of RRC connection establishment failures or
RRC connection requests discarded in flow control. These symptoms verify that RRC
connection reconfiguration signaling for DRX is the cause of the problem.
The changes in sensitive performance indicators are not presented; these indicators
include the peak usage of the main control board CPU, number of RRC connection
establishment failures, number of RRC connection requests discarded in flow control,
maximum number of online users, and RRC connection establishment success rate.
14. Conclusion
After DRX is enabled, a large number of RRC connection reconfiguration messages
are added on the air interface and the main control board CPU is overloaded. Msg 4 is
delivered too late, and flow control is triggered, leading to a large number of RRC
connection establishment failures and RRC connection requests discarded. Repeat
access attempts after access failures further increase the CPU usage and intensify flow
control until the traffic volume decreases.
The solution is to set the FddEnterDrxThd and FddExitDrxThd parameters to the
recommended value 1000.
Problem Analysis
15. Simply analyze the problem.
At 21:37:31, fourteen uplink voice packets are lost in TTI 20215215, according to the
eNodeB PDCP tracing result.
The uplink scheduling tracing result shows that the eNodeB performs scheduling
according to the SR of a UE in TTI 20214878 at 21:37:31. After the scheduling, this
UE sends a BSR indicating that the buffer is not empty. However, the eNodeB does
not perform uplink scheduling in time until the UE sends again the SRI 330 ms later,
that is, in TTI 20215208. As the PDCP packet discard timer is set to 100 ms, the
eNodeB detects packet losses in TTI 20215215.
The cause of voice packet losses is that the eNodeB does not promptly perform
scheduling based on BSRs.
16. Analyze the voice problem of the LBBPc.
The eNodeB detects the UE's uplink BSRs every 8 ms (which is defined by the BSR
Check Timer. If a BSR indicates that the buffer is not empty, there is a chance of
scheduling. After DRX is enabled, the eNodeB does not schedule a UE if it detects that
the UE is in the sleep state and the BSR check time does not fall in the On Duration.
The eNodeB schedules the UE until the BSR Check Timer expires and the UE sends
an SR. As the delay is long, the PDCP packet discard timer expires and VoLTE packet
loss occurs.
For the LBBPd and UBBPd, the scheduling algorithm is more complex and accurate.
The BSR Check Timer is set to 2 ms (as shown in the following figure), and the non-
empty buffer status can be detected in the active time. As a result, UEs can be
scheduled more promptly, provided that the number of UEs served by the LBBPd and
UBBPd is the same as that served by the LBBPc.
With the increase in the number of users, the calculation workload will be reduced to
ensure system stability. Specifically, the BSR Check Timer will be automatically
prolonged to reduce the frequency of updating scheduling chances, as shown in the
following table, where N indicates the number of users served by a BBP.
BSR Check Timer LBBPc LBBPd UBBPd
17. Conclusion
On the live network, the On Duration Timer and the DRX Inactivity Timer are set to
small values, leading to concentrated scheduling requirements. As a result, some UEs
cannot be scheduled in time, voice packets are lost, and voice delays are prolonged.
In addition, the parameter settings recommended for the live network only consider
power saving for a single UE without considering performance, delay, and capacity.
The parameter settings should be optimized for the entire system.
To ensure timely scheduling when the buffered data cannot be scheduled at a time,
Huawei recommends that the DRX Inactivity Timer be set to 80 ms. The DRX
parameters are not adjusted during web browsing and data download and therefore
power consumption will not increase. The DRX parameters are adjusted only when the
UE is performing a VoLTE service. In this case, as the active time increases, UE power
consumption will slightly increase. The lab test result shows that Huawei-
recommended DRX parameter settings can effectively reduce the number of VoLTE
packet losses.
Problem Analysis
11. Identify the problem in the IOT lab.
Based on the parameter settings used on live networks, a test is performed in the IOT
lab to compare the MOSs under Huawei-recommended and Apple-recommended
parameter settings. The test result is as follows:
For Samsung Galaxy S4, S5, S6, and Note 3 with coding rates of 23.85 bit/s and 12.65
bit/s, the MOSs decrease by 0.1 to 0.3 under Apple-recommended DRX parameter
settings. The decreases at the coding rate of 23.85 bit/s are greater than those at the
coding rate of 12.65 kbit/s.
For Huawei Mate 7 with a coding rate of 23.85 bit/s, the MOS also decreases by 0.1 to
0.15 under Apple-recommended parameter settings.
For Apple iPhone 6s with a coding rate of 12.65 bit/s, the MOS does not obviously
decrease under Apple-recommended parameter settings.
UEs Coding Rate Parameter MOS E2E_Delay_AVG (ms)
All terminals experience greater voice packet delay variations under Apple-
recommended parameter settings than under Huawei-recommended parameter settings.
The voice packet delay variations for iPhone 6s are in the range of about 0–40 ms
under Apple-recommended parameter settings while those for Samsung terminals are
in the range of 0–80 ms.
The result of comparison between UEs' TX and RX voice waveforms shows that delay
variations lead to jitter-buffer packet losses and MOS decreases.
Case 1: MOS = 3.4
From 15:33:29 to 15:33:40, both calling and called parties do not experience uplink or
downlink voice packet losses on the air interface.
The waveform analysis result shows that about 886 ms later after the beginning,
additional frames introduced in the jitter buffer lead to a waveform phase shift, which
affects the MOS. The first figure below shows a low-MOS waveform and the second
shows a high-MOS waveform.
The QXDM data shows that a long delay of 80 ms appears in the 44th conversational
frame.
From 15:35:18 to 15:35:29, both calling and called parties do not experience uplink or
downlink voice packet losses on the air interface.
The waveform analysis result shows that about 744 ms later after the beginning,
additional frames introduced in the jitter buffer lead to a waveform phase shift, which
affects the MOS. The first figure below shows a low-MOS waveform and the second
shows a high-MOS waveform.
The QXDM data shows that a long delay of 80 ms appears in the 37th conversational
frame. (This delay is equal to 29 ms plus 51 ms, where 29 ms is the delay of the RTCP
packet with a PDU size of 55.)
However, only one 103-byte voice packet can be scheduled in subframe 8 of frame
368 according to the uplink grant. The reason is that the SR does not contain any
traffic volume information about the terminal and therefore the uplink scheduler
adaptively determines the to-be-scheduled traffic volume (133) according to the last
scheduled traffic volume (113).
After the scheduling, the UE still has data in the buffer to transmit and therefore sends
a BSR instead of resending an SR. However, the Apple-recommended On Duration is
only 4 ms, and the data can be scheduled only in the first subframe of the next On
Duration. As a result, two voice packets are scheduled 40 ms later.
Considering both uplink and downlink delay variations, a total of 0–80 ms variation is
introduced from end to end under Apple-recommended parameter settings.
Difference 40–80 40
The following figure shows the distribution and of MOSs and the effect after the
optimization.
The following figures compare the delay variations before and after optimization.
Under Apple-recommended DRX settings and Huawei optimization solution, the
uplink delay variations for the local Note 3 decrease from 0–40 ms to 15–25 ms and
the downlink delay variations of 99% of samples for the peer Note 3 are controlled in
the range of 0–40 ms, equivalent to the optimization effect of Apple.