You are on page 1of 17

GSM BSS Network PS KPI (RTT Delay) Optimization Manual INTERNAL

Product Name Confidentiality Level


GSM BSS INTERNAL
Product Version
Total 31 pages

GSM BSS Network PS KPI (RTT Delay)


Optimization Manual

(For internal use only)

Prepared by GSM&UMTS Network Date


Performance Research 2008-12-20
Department

Reviewed by Date

Reviewed by Date

Granted by Date

Huawei Technologies Co., Ltd.

All rights reserved

2008-11-01 All rights reserved Page 1 of 17

PDF created with pdfFactory trial version www.pdffactory.com


GSM BSS Network PS KPI (RTT Delay) Optimization Manual INTERNAL

Revision Record

Date Version Description Author


2008-12-20 1.0 Draft completed Geng Haijian

2008-12-25 1.0 Revised according to review comments Geng Haijian

2008-11-01 All rights reserved Page 2 of 17

PDF created with pdfFactory trial version www.pdffactory.com


GSM BSS Network PS KPI (RTT Delay) Optimization Manual INTERNAL

Contents

1 Overview ................................................................................................................................... 7
1.1 Definition ................................................................................................................................................... 7
1.2 Ping Process Analysis ................................................................................................................................. 7
1.3 Introduction to Delay in the GPRS/EGPRS System ..................................................................................... 9

2 Analysis and Optimization of the RTT Delay ................................................................... 10


2.1 TBF Establishment and Release Delay Optimization ................................................................................. 13
2.2 Data Transmission Delay Analysis ............................................................................................................ 14
2.3 System Processing Delay .......................................................................................................................... 16
2.4 Transmission Delay between Network Elements ....................................................................................... 16

2008-11-01 All rights reserved Page 3 of 17

PDF created with pdfFactory trial version www.pdffactory.com


GSM BSS Network PS KPI (RTT Delay) Optimization Manual INTERNAL

Figures

Figure 1 Execution result of the Ping command ............................................................................................... 8

Figure 2 Capturing Ping service packets........................................................................................................... 8

Figure 3 Ping flow when the TBF establishment or release flow is not optimized............................................ 10

Figure 4 Message tracing at the GB interface ................................................................................................. 11

Figure 5 Comparison of interface tracking (left, corresponding to line 24 in Figure 4) and Ethereal tracking
(right) ............................................................................................................................................................ 11
Figure 6 Um interface access flow (the first Uplink ACK message contains the TLLI for settling the conflict) 12

Figure 7 Um interface tracking the second Ping data ...................................................................................... 12

Figure 8 TEMS interface tracks the block which is not received by the network side when the uplink sends
blocks with BSN number 4 and 5 during the second Ping ................................................................................ 13
Figure 9 Packing a Ping packet at each protocol layer .................................................................................... 15

2008-11-01 All rights reserved Page 4 of 17

PDF created with pdfFactory trial version www.pdffactory.com


GSM BSS Network PS KPI (RTT Delay) Optimization Manual INTERNAL

Abstract and Abbreviation


Keywords
Ping, RTT delay

Abstract
Ping is used to test the RTT delay that affects the utilization rate of bandwidth. This document
describes the basic process of the Ping service and the method for reducing the RTT delay.

Acronyms and Abbreviations


Acronym and Abbreviation Full Spelling
GPRS General Packet Radio Service
EGPRS Enhanced GPRS
MS Mobile Station
CQT Call Quality Test
DT Drive Test

RTT Round Trip Time

ICMP Internet Control Message Protocol

PCU Packet Control Unit

BSS Base Station System

BSC Base Station Control

BTS Base Transceiver Station


TBF Temporary Block Flow
RLC Radio Link Control
MAC Media Access Control

MCS Modulation and Coding Scheme

References
[1] TCP/IP Detail Volume 1, by W.Richard Stevens, in April, 2008
[2] GPRS Network Technology, prepared by Motorola Engineering Institute, on June 1, 2005

2008-11-01 All rights reserved Page 5 of 17

PDF created with pdfFactory trial version www.pdffactory.com


GSM BSS Network PS KPI (RTT Delay) Optimization Manual INTERNAL

[3] GSM PS Wireless Performance Baseline (field version), prepared by the GSM Solution
Testing Department, on September 3, 2008

2008-11-01 All rights reserved Page 6 of 17

PDF created with pdfFactory trial version www.pdffactory.com


GSM BSS Network PS KPI (RTT Delay) Optimization Manual INTERNAL

1 Overview

This document describes the role of the RTT delay, each occurrence of the RTT delay in the
GPRS network, and methods of reducing the RTT delay.
The fault identifying tools in this document mainly refer to TEMS and Ethereal/Wireshark.
These tools can be used to trace or browse the information at the NE side. To analyze the RTT
delay, you need to collect the following data: signaling traced by the TEMS, packets captured
by the Ethereal/Wireshark at the MS side, signaling traced on the BSS PS Um interface, and
signaling traced on the BSS PS Gb PTP.
The description in this document is based on the condition that no download or upload
problem exists. In other words, no such fault as channel fault, link out of synchronization, or
packet loss on the interface, occurs.

1.1 Definition
The RTT delay refers to the time elapsed for a packet to the destination and back again. What
impact does the RTT delay have on the system? Why do telecom operators need to assess the
RTT delay?
In the case that the RTT delay is long, the confirmation of the data at the application layer
takes a long time. Therefore, at the slow startup phase, windows are displayed after a long
delay, and the time to reach the congestion threshold is prolonged; at the congestion control
phase, the increase in windows size is also slow. In this case, when the RTT delay is long, the
bandwidth utilization rate is low.

1.2 Ping Process Analysis


The RTT delay is often tested through the Ping command. The Ping service is based on the
ICMP protocol. The client sends an ICMP echo request packet to the server. After receiving
the request, the server encapsulates the data that is contained in the request into an ICPM
response packet and sends it back to the client. The following figure shows the result of the
Ping command:

2008-11-01 All rights reserved Page 7 of 17

PDF created with pdfFactory trial version www.pdffactory.com


GSM BSS Network PS KPI (RTT Delay) Optimization Manual INTERNAL

Figure 1 Execution result of the Ping command

In the Ping command, -n indicates the number of Ping times, and -l indicates the size of a
Ping packet. After the Ping command is run, the RTT delay of each Ping packet will be
displayed. After the execution of the Ping command is complete, the minimum, maximum,
and average delay of the Ping packet will be displayed. In the Windows operating system, the
time precision is 1 ms.
Figure 2 shows the data traced by the WireShark.

Figure 2 Capturing Ping service packets

8 indicates the echo requests, and 0 indicates the echo response times.

Identify a process, and allow the terminal to perform multiple Ping operations at the same time.
Number every Ping request packet. Give the corresponding Ping echo packet a same serial number to
calculate the delay

When the default settings of the Windows operating system are used, the interval for sending
two Ping requests is 1 second, as indicated by the interval between the two packets in Figure 2.
If the delay of the Ping echo response packet is larger than 1 second, the next Ping request is
sent immediately after the echo response packet is received. In addition, if an echo response
packet is not received within 5 seconds, a message indicating request time out is returned.
This means that the peer is not reachable. The peer is not certainly unreachable. A possible
cause is that the peer network firewall blocks Ping packets.

2008-11-01 All rights reserved Page 8 of 17

PDF created with pdfFactory trial version www.pdffactory.com


GSM BSS Network PS KPI (RTT Delay) Optimization Manual INTERNAL

1.3 Introduction to Delay in the GPRS/EGPRS System


Generally, when calculating delay, factors such as the network load and changes made in the
intermediate network to data packet authority are not considered. The delay in the GPRS or
EGPRS system includes the system processing delay, transmission delay between network
elements, data transfer delay, and TBF establishment and release delay. The system processing
delay depends on the hardware processing capability of the system. The transmission delay
between network elements is related to the interface transmission mode. The data transfer
delay is related to the size of the Ping packet and the encoding rate employed. The TBF
establishment and release delay is related to the access method (for example, one-phase
access or two-phase access) and the optimization methods related to TBF establishment or
release at the network side.

2008-11-01 All rights reserved Page 9 of 17

PDF created with pdfFactory trial version www.pdffactory.com


GSM BSS Network PS KPI (RTT Delay) Optimization Manual INTERNAL

2 Analysis and Optimization of the RTT


Delay

Figure 3 shows the transmission process of a Ping packet in the case that no function is
enabled.

Figure 3 Ping flow when the TBF establishment or release flow is not optimized

Network layer at MS PCU Network layer at


the client layer RLC/MAC layer RLC/MAC layer the server
ICMP echo request
Uplink TBF
establishment request

Assigning uplink resource

RLC data blocks that carry


the ICMP echo request ICMP echo request

ICMP echo response


RTT delay
Assign downlink resources
Assignment success
confirmation

RLC data blocks that carry


the ICMP echo request

ICMP echo response

As shown in Figure 3, the Ping flows involves the establishment of the downlink TBF,
transmission of uplink and downlink data blocks, and data assembly at the MS side and client
side.
The Ping delay is analyzed as follows:
1. Check whether delay is incurred at the core network side (data is traced on the Gb
interface). If the delay is incurred, contact Huawei technical support engineers of the
core network for help. Otherwise, go to the next step.
2. Compare the delay of the first Ping packet and that of consecutive Ping packets, and
determine which one is longer.
l If the delay of only the first Ping packet is long, check whether the TBF establishment
flow is normal.

2008-11-01 All rights reserved Page 10 of 17

PDF created with pdfFactory trial version www.pdffactory.com


GSM BSS Network PS KPI (RTT Delay) Optimization Manual INTERNAL

l If the delay of consecutive Ping packets is long, check whether the TBF
establishment/release optimization flow is started, whether the parameter settings are
improper, and whether the deferred release times out due to faults in data transmission.
l If the delay of certain packets among the consecutively sent Ping packets is long, check
factors such as the encoding rate, channel quantity, and error blocks in data transmission.
Take the Ping test shown in Figure 1 as an example. The data traced on the Gb interface
shows that no large delay is incurred in the core network.

Figure 4 Data traced on the Gb interface

The delay of the first Ping packet is 767 ms, a little longer than the usual value. As shown by
the signaling, downlink pre-establishment is not started (Downlink pre-establishment should
be started when the first Uplink ACK message is sent. The downlink pre-establishment is
started if the downlink assignment message is sent after the first Uplink ACK message is sent);
one uplink channel is used and uplink and downlink encoding rates are relatively low (this
will be described in the process of the second Ping packet).
Among the consecutive Ping packets that are sent, the delay of certain Ping packets is long.
Based on the process of the second Ping packet shown in Figure 1, you can locate the data of
this Ping on the Gb interface. Figure 5 shows the two data blocks corresponding to the
process of the second Ping packet.

Figure 5 Comparison between the data traced on the Gb interface (on the left side, corresponding
to line 24 in Figure 4) and that traced by using the Ethereal (on the right side)

2008-11-01 All rights reserved Page 11 of 17

PDF created with pdfFactory trial version www.pdffactory.com


GSM BSS Network PS KPI (RTT Delay) Optimization Manual INTERNAL

If no long delay is incurred on the Gb interface, the delay is mainly caused by factors on the
Um interface. If a lot of services are being processed in the existing network, you must locate
the data of the MS and check the first Uplink ACK message that is used to settle the conflict,
as shown in Figure 6.

Figure 6 Um interface access flow (the first Uplink ACK message contains the TLLI for settling
the conflict)

This is one-phase access. Find the last immediate assignment message and you can see that
the uplink TFI in the message is 10. Find the next downlink assignment message and you can
see that the MS is addressed based on uplink TFI 10 and that the downlink TFI of the MS is
13. Locate the data blocks transmitted on the Um interface from the Gb interface. You can
identify the specific data blocks according to the contents of the entire LLC PDU on the Gb
interface and the contents of the RLC data blocks on the Um interface, as shown in Figure 7.

Figure 7 Data related to the second Ping operation traced on the Um interface

As shown in Figure 7, the uplink and downlink encoding rates are relatively low; the uplink
uses only one channel; the first sending of uplink data blocks with BSN number 4 and 5 are
not received by the network side.
The uplink encoding rate is relatively low because the default encoding rate is MCS2, which
is not adjusted in time. Based on the configuration, the uplink encoding rate is adjusted
according to the downlink encoding rate. In this case, it can be inferred that the downlink
encoding rate is not adjusted. The default downlink encoding rate is MCS6; however MCS2 is
used instead. This is because the sublink cannot be bound. Only one channel is used on the
uplink because it is still in one-phase access and the time to obtain two uplink timeslots is not
reached. The uplink data blocks with BSN number 4 and 5 are not received by the network
side. However, the quality of the Um interface is good. The possible cause is that faults occur
on the link and channel on the uplink G-Abis interface.
Such a long delay would not occur even if two blocks are lost, as the data blocks should be
continuously dispatched in the delay release state in the case that the channel is not

2008-11-01 All rights reserved Page 12 of 17

PDF created with pdfFactory trial version www.pdffactory.com


GSM BSS Network PS KPI (RTT Delay) Optimization Manual INTERNAL

multiplexed by other MSs. Figure 8 shows the message traced by using the TEMS on the Um
interface (in Mode Report—PH PDCH Block Header DL Report).

Figure 8 Tracing the data blocks with BSN number 4 and 5 that are not received by the network
side by using the TEMS on the Um interface

If many downlink blocks are not received by the MS, it is probable that the transmission link
is faulty.
Step 2 Analysis shows that the fault is probably caused by the link quality. Check whether an alarm
indicating loss of synchronization on the link is generated and whether the BTS clock is
synchronized with the BSC clock, and check the bit error rate on the G-Abis interface. In the
laboratory environment, the BTS clock is confirmed to be in free-run state. The file attached
provides the Ping operation related data traced the configuration data. Based on the preceding
analysis, you can try to analyze the long delay in the fourth Ping operation.

2.1 TBF Establishment and Release Delay Optimization


1. Uplink TBF establishment delay optimization
An uplink TBF can be established under two situations: one-phase access and two-phase
access. In one-phase access, it takes about 193 ms from the sending of a channel request to the
assignment of uplink resources. In two-phase access, it takes about 452 ms from the sending
of a channel request to the assignment of uplink resources. The access method is determined
by the MS. However, the network side can specify the access as two-phase access, which,
however, is not recommended. Check whether the network side specifies two-phase access by
entering the super user mode and choosing Configure BSC Attributes > Software
Parameters > Force MS to Phase2 Function Switch. The switch should be set to Not
Force.

2008-11-01 All rights reserved Page 13 of 17

PDF created with pdfFactory trial version www.pdffactory.com


GSM BSS Network PS KPI (RTT Delay) Optimization Manual INTERNAL

In two-phase access, the most time-consuming part is the process of assigning single
block—> the two-phase access request of the MS on the single block. You can modify the
value of Delay Blocks of Uplink Single Block Assignment from 9 to 6 by entering the super
user mode and choosing Configure BSC Attributes > Software Parameters >Delay Blocks
of Uplink Single Block Assignment. In this way, the delay can be reduced by about 60 ms.

Delay Blocks of Uplink Single Block Assignment considers the delay incurred during the transmission
of the immediate assignment from the PCU to the MS, including the transmission delay and the delay
caused by the queuing of the immediate assignment message at the BTS side. In the laboratory, no fault
occurs when Delay Blocks of Uplink Single Block Assignment is set to 6. In the actual network,
however, if the number of paging and access requests is great, setting Delay Blocks of Uplink Single
Block Assignment to 6 may cause the problem that the MS receives the assignment message at a time
later than the time of the assigned block. In this case, the channel request must be re-sent. Therefore, set
Delay Blocks of Uplink Single Block Assignment to 6 only in the case of competition of the RTT delay
performance with peer vendors.

2. Uplink TBF release delay optimization


After all uplink data blocks are sent, the release procedure of uplink TBF can be started. But
after the uplink TBF is released, it takes a long time to establish the downlink TBF. In
addition, for consecutive Ping packets, each Ping operation requires the re-establishment of an
uplink TBF, which is quite time-consuming. Therefore, the release of an uplink TBF is very
often delayed.
Usually, the uplink extension function is enabled. To enable this function, choose Configure
Cell Attributes > GPRS Attributes > PS Network Optimization Parameters > Inactive
Delay of Extended Uplink TFB(ms), and set Inactive Delay of Extended Uplink TFB(ms)
to a proper value. The recommended value is 2000 ms.

The uplink extension function should be supported by the MS.


As the interval for two Ping packets is 1 second, Inactive Delay of Extended Uplink TFB(ms) should
be set to a value that is larger than 1 second.

3. Downlink TBF establishment delay optimization


Normally, the downlink TBF establishment starts when the downlink data reaches the PCU.
But for common PS services, the uplink TBF is established first, and downlink data is
available. Therefore, the product provides the downlink pre-establishment function. You can
enable this function by entering super user mode, choosing Configure BSC Attributes >
Software Parameters > Support advance downlink TBF establishment, and setting
Support advance downlink TBF establishment to Optimize with ALL TLLI. In this case,
after the uplink TBF is established, the establishment of the downlink TBF is immediately
triggered.
4. Downlink TBF release delay optimization
The release of the downlink TBF should also be delayed to avoid repeated establishments of
the downlink TBF when consecutive Ping packets are sent. To enable this function, choose
Configure Cell Attributes > GPRS Attributes > PS Network Optimization Parameters >
Release Delay of Downlink TFB(ms) and set Release Delay of Downlink TFB(ms) to a
value larger than 1000 ms. The recommended value is 2400 ms.

2.2 Data Transmission Delay Analysis


Figure 9 shows how a 32-byte Ping packet is encapsulated at each protocol layer.

2008-11-01 All rights reserved Page 14 of 17

PDF created with pdfFactory trial version www.pdffactory.com


GSM BSS Network PS KPI (RTT Delay) Optimization Manual INTERNAL

Figure 9 Ping packet encapsulated at each protocol layer

As shown in Figure 9, the Ping packet is encapsulated at each layer, and when it reaches the
RLC layer, it contains 73 bytes. Table 1 lists the sizes of an RLC data block in different
encoding rates.

Table 1 Payload of LLC layer data in the RLC data blocks in different encoding rates in EGPRS

Channel Coding Scheme EGPRS RLC Data Unit Size (N2) (octets) Family
MCS-1 22 C

MCS-2 28 B
MCS-3 37 A
MCS-4 44 C

MCS-5 56 B

MCS-6 74 A
MCS-7 2x56 B
MCS-8 2x68 A
MCS-9 2x74 A

Table 2 Payload of LLC layer data in the RLC data blocks in different encoding rates in GPRS

Channel Coding RLC Data Block Size Number of RLC Data Block
Scheme Without Spare Bits (N2) Spare Bits Size (octets)
(octets)
CS-1 22 0 22

CS-2 32 7 32 7/8
CS-3 38 3 38 3/8

CS-4 52 7 52 7/8

2008-11-01 All rights reserved Page 15 of 17

PDF created with pdfFactory trial version www.pdffactory.com


GSM BSS Network PS KPI (RTT Delay) Optimization Manual INTERNAL

As listed in Table 1 and Table 2, the MCS6 coding scheme is required to send a Ping data
block within 20 ms in the case of a single uplink channel. Thus, when you test the Ping delay,
it is recommended that you use MCS6 as the default uplink or downlink coding scheme. For
GPRS, two channels are required to send a Ping packet within 20ms when the MCS3 or
MCS4 coding scheme is adopted. Therefore, when the Ping operation is initiated for the first
time, the product regards the service as neutral by default. For an MS with the multislot class
10, three downlink channels and two uplink channels are assigned. For script compilation, if
the Ping operation is started immediately after download services, it is recommended that you
set the first bit of DSP Control Table 2 to ON. To enable this function, enter the super user
mode, choose Configure BSC Attributes > Software Parameters >DSP Control Table 2,
and set the first bit of DSP Control Table 2 to ON.
If error blocks exist, the Ping delay may burst. This is because if the sent data is not
confirmed by the peer, and the unconfirmed data blocks are not re-sent; in this case, the local
end waits a long time for the peer to confirm the data blocks before re-sending the blocks. To
solve this problem, a new function, PACK data block retransmission, is developed at the
network side. When error blocks exist (especially in DT Ping tests), you need to check
whether the current version supports the PACK data block retransmission function.

2.3 System Processing Delay


The processing delay for different BTS hardware versions is different. For the BTS3012, the
delay of the new double-transceiver BTS3012 is 40 ms shorter than that of the old
double-transceiver BTS3012.
In addition, different MSs also have different delay. The delay of a Nokia N95 is 40 ms
shorter than that of a Sony Ericsson K790.

2.4 Transmission Delay between Network Elements


The transmission delay between network elements is related to the transmission mode
between network elements, mainly the transmission mode on the G-Abis interface
(TDM/HDLC/IP). The uplink extension function, downlink release delay function, built-in
PCU, and new double transceiver should fit into the transmission mode. Table 3 describes
results of Ping test in the scenario where the wireless environment is good and the network
load is not considered.

Table 3 Results of CQT Ping Test

Consecutive
First Pinged
Pinged
EDGE/GPRS Device 32-Byte
32-Byte Packet
Packet (ms)
(ms)
Built-in PCU; Abis TDM; K790 678 217
External PCU; Abis TDM; K790 1061 —
EDGE
Built-in PCU; Abis TDM; N95 — 163
Built-in PCU; Abis IP 503 169

2008-11-01 All rights reserved Page 16 of 17

PDF created with pdfFactory trial version www.pdffactory.com


GSM BSS Network PS KPI (RTT Delay) Optimization Manual INTERNAL

Consecutive
First Pinged
Pinged
EDGE/GPRS Device 32-Byte
32-Byte Packet
Packet (ms)
(ms)

Built-in PCU; Abis IP(synchronization


392.56 97
optimized)

Built-in PCU; Abis TDM; K790 704 246


External PCU; AbisTDM; K790 782 336
GPRS
Built-in PCU; Abis TDM; N95 565.53 196
Built-in PCU; Abis IP 337 171

3 Appendix

The file attached provides the data traced in the Ping test.

Ping Test Data


Traced.rar

2008-11-01 All rights reserved Page 17 of 17

PDF created with pdfFactory trial version www.pdffactory.com

You might also like