Docstoc

GPRS Drive Testing quick guide

Document Sample
GPRS Drive Testing quick guide Powered By Docstoc
					                                                                                   Open
                                                                                   INSTRUCTION                                                       1 (1)
Prepared (also subject responsible if other)                                       No.

EAB/ZG/NI Robert Blom                                                              EAB/G-04:000245 Uen
Approved                                                       Checked             Date                  Rev           Reference

EAB/ZG/NI [Johannes Lindvall]                                                      2004-02-10            A




                                               Quick guide to GPRS drivetesting with TEMS
                                               Investigation

                                               Abstract

                                               This document is intended to be a quick guide on how to efficiently test and
                                               monitor performance of a GPRS network with TEMS Investigation.



                                               Contents
                                               1    Introduction .......................................................................................... 2
                                               2    Test Applications ................................................................................. 3
                                                    2.1    Attach / Detach ......................................................................... 3
                                                    2.2    Activate / De-activate PDP context ........................................... 3
                                                    2.3    Ping .......................................................................................... 3
                                                    2.4    HTTP Load ............................................................................... 4
                                                    2.5    FTP Get/Put.............................................................................. 4
                                               3    Measurements ...................................................................................... 5
                                                    3.1    Pre-requisites ........................................................................... 5
                                                    3.2    GPRS Functionality Tests......................................................... 5
                                                    3.2.1  Setup examples........................................................................ 6
                                                    3.2.2  Analysis examples .................................................................... 7
                                                    3.3    GPRS Radio Network Test ..................................................... 11
                                                    3.3.1  Setup example........................................................................ 11
                                                    3.3.2  Analysis examples .................................................................. 11
                                                    3.4    General Tips & Tricks ............................................................. 20
                                               4    Overall Procedures ............................................................................ 21
                                               5    References.......................................................................................... 21
                                                                          Open
                                                                          INSTRUCTION                                      2 (2)
Prepared (also subject responsible if other)                              No.

EAB/ZG/NI Robert Blom                                                     EAB/G-04:000245 Uen
Approved                                                    Checked       Date            Rev       Reference

EAB/ZG/NI [Johannes Lindvall]                                             2004-02-10      A



                      1                        Introduction
                                               This document describes a procedure for measuring GPRS performance
                                               using TEMS Investigation 4, where two different aspects are covered.

                                               1. How to do continuous testing of basic GPRS core functionality, such as
                                                  Attach & PDP Context activation.

                                               2. How to measure radio network quality as experienced by a GPRS user.

                                               The main goal of the document is to describe a procedure for the second part:
                                               I.e. to present a methodology for finding problems related to the radio
                                               network, giving a few examples on possible solutions and also to give
                                               examples on how to identify problems that are not related to the radio
                                               network.

                                               In chapter two the different applications that can be tested from in TEMS
                                               Investigation is outlined and the main usage of the test applications are
                                               described.

                                               In chapter three examples on how to set-up and analyze measurements are
                                               given and chapter four is a short summary of the overall measurement
                                               procedures.
                                                                           Open
                                                                           INSTRUCTION                                    3 (3)
Prepared (also subject responsible if other)                               No.

EAB/ZG/NI Robert Blom                                                      EAB/G-04:000245 Uen
Approved                                                     Checked       Date            Rev       Reference

EAB/ZG/NI [Johannes Lindvall]                                              2004-02-10      A



                      2                        Test Applications
                      2.1                      Attach / Detach

                                               Attach procedure is when the MS is activating mobility management for
                                               GPRS, i.e. going from GMM Idle state. When this is done the SGSN node will
                                               continuously handle location information for the mobile, knowing the routing
                                               area that the mobile belongs and in some cases also the specific cell (when
                                               the MS is in Ready state).

                                               The main purpose of GPRS Attach tests is to test the functionality of the
                                               GPRS core network. The procedure is so short that radio environment has
                                               limited impact on performance.

                                               These tests are best performed with automated GPRS Attach and GPRS
                                               Detach commands in the Command sequence handler.

                      2.2                      Activate / De-activate PDP context

                                               PDP Context activation / de-activation is session management procedures
                                               used when the MS communicates with the GGSN node and receives a user
                                               IP address (at activation). It is only after the PDP context activation is
                                               performed that any real user data can be sent. Prior to this only GPRS
                                               Mobility Management messages and SMS over GPRS will be sent (when MS
                                               is attached).

                                               The main purpose of Activate / De-activate PDP context is to test the
                                               functionality of the GPRS core network. The procedure is so short that radio
                                               environment has limited impact on performance.

                                               These tests are best performed with automated dial up and hang up
                                               commands (which is the “computer” commands generating PDP context
                                               activation) in the Command sequence handler.

                      2.3                      Ping

                                               Ping is an Internet control message that is sent to a server and a response
                                               back is expected.

                                               The main purpose of the ping is to check accessibility to a specific server and
                                               the roundtrip time to that server (which is a quality indicator of the network).
                                               The procedure is so short that radio environment has limited impact on
                                               performance. The time to transfer the ping message over the air interface is
                                               normally the main contributor, which makes the result BSS implementation
                                               dependant (thus different between vendors).

                                               These tests are best performed with automated ping commands in the
                                               Command sequence handler.
                                                                           Open
                                                                           INSTRUCTION                                    4 (4)
Prepared (also subject responsible if other)                               No.

EAB/ZG/NI Robert Blom                                                      EAB/G-04:000245 Uen
Approved                                                     Checked       Date            Rev       Reference

EAB/ZG/NI [Johannes Lindvall]                                              2004-02-10      A

                      2.4                      HTTP Load

                                               HTTP load corresponds to a user browsing the web and downloading a web
                                               page. This includes signaling through the GPRS core network and IP network
                                               between MS and HTTP server.

                                               It is important to remember that the actual design of the web page will impact
                                               the measured performance. Number of objects in a webpage, what servers
                                               these objects resides on and used HTTP version are some factors
                                               contributing to how efficient the download will be.

                                               HTTP load is thus useful in order to investigate performance of a specific web
                                               page or to do repetitive tests of downloading a standardized/default webpage.
                                               Most web pages will however not provide sufficiently stable throughput to be
                                               useful for radio network evaluation (note that a web page could also be
                                               designed to give a stable throughput suitable similar to FTP).

                                               These tests are best performed with automated HTTP load commands in the
                                               Command sequence handler.

                      2.5                      FTP Get/Put

                                               FTP Get and Put are commands for downloading and uploading data to FTP
                                               servers. This includes signaling through the GPRS core network and IP
                                               network between MS and FTP server.

                                               FTP Get and Put is normally used to initiate data transfer in downlink and
                                               uplink direction. This data transfer uses TCP and will try to maximize the
                                               throughput from the beginning to the end of the file transfer. It is thus a good
                                               choice when evaluating the radio network performance. There will be a small
                                               set-up time in the beginning of FTP Get and Put applications due to:
                                               connection to server, providing username and password, setting directory etc.

                                               These tests are best performed with automated FTP Get and Put commands
                                               in the Command sequence handler.
                                                                          Open
                                                                          INSTRUCTION                                    5 (5)
Prepared (also subject responsible if other)                              No.

EAB/ZG/NI Robert Blom                                                     EAB/G-04:000245 Uen
Approved                                                     Checked      Date            Rev       Reference

EAB/ZG/NI [Johannes Lindvall]                                             2004-02-10      A



                      3                        Measurements
                      3.1                      Pre-requisites

                                               There are a number of pre-requisites for successfully completing a
                                               measurement and a small checklist of these is provided below. For a more
                                               complete description of connection and measurement set-up procedures,
                                               please refer to the TEMS Investigation manual.

                                               ! SIM card with a GPRS subscription

                                               ! TEMS Mobile and a GPS if geo-positioned data is wanted

                                               ! Data account that is correctly configured in the MS
                                                 This includes e.g. which APN that the MS will connect to. Different data
                                                 accounts will have different CID numbers.

                                               ! The mobile should be defined as a modem in windows
                                                 Verify that an updated modem driver exists for the used TEMS model.
                                                 Use cable connection from PC to mobile.

                                               ! Create a dial-up connection in windows
                                                 Different data accounts can be selected with different phone numbers
                                                 when doing dial up procedure, i.e. *99***1# refers to CID 1 (account 1),
                                                 whereas X*99***2# refers to CID 2.

                                               ! Test the applications in advance
                                                 It normally saves time if the applications that are to be used is tested from
                                                 a LAN connection in advance, e.g. verifying FTP servers, user IDs,
                                                 passwords, ping response time etc. This could otherwise be a time
                                                 consuming procedure to do when drivetesting.

                      3.2                      GPRS Functionality Tests

                                               These tests are performed to test the functionality of a GPRS network, i.e. the
                                               actual radio environment is not really interesting.

                                               Tests falling under this category is GPRS Attach / Detach, PDP context
                                               activation / de-activation tests for the GPRS core network and Ping for the
                                               general IP network perceived.

                                               Typical measures are success rate, cause values of rejects and time taken to
                                               complete a procedure.
                                                                             Open
                                                                             INSTRUCTION                                  6 (6)
Prepared (also subject responsible if other)                                 No.

EAB/ZG/NI Robert Blom                                                        EAB/G-04:000245 Uen
Approved                                                     Checked         Date          Rev       Reference

EAB/ZG/NI [Johannes Lindvall]                                                2004-02-10    A

                      3.2.1                    Setup examples

                                               An example of a command sequence for performing multiple Attach and
                                               Activate PDP Context tests is shown in Figure 3-1




                                               Figure 3-1   Command sequence example of Attach & PDP Activation tests

                                               An example of a command sequence for performing repetitive Ping tests is
                                               shown in Figure 3-2.




                                               Figure 3-2   Command sequence example of ping test

                                               The Ping command has a number of arguments

                                               -   Size of the ping packet
                                               -   Number of consecutive pings
                                               -   Timeout time (indicates when ping is seen as unsuccessful)
                                               -   Time between consecutive pings

                                               The setting of these parameters will affect the measured performance. E.g. a
                                               large packet will take long time to transfer over the air interface. Short time
                                               between consecutive pings will increase the probability that the MS will be in
                                               Packet Transfer mode when next ping is to be sent (which will improve
                                               performance) etc.
                                                                           Open
                                                                           INSTRUCTION                                        7 (7)
Prepared (also subject responsible if other)                               No.

EAB/ZG/NI Robert Blom                                                      EAB/G-04:000245 Uen
Approved                                                     Checked       Date             Rev       Reference

EAB/ZG/NI [Johannes Lindvall]                                              2004-02-10       A

                      3.2.2                    Analysis examples

                                               The first step of the functionality analysis is normally to look at the success
                                               rate of the procedures.

                                               For Attach and Activate PDP context procedures this can be quickly
                                               calculated by creating a report over relevant events, i.e. including GPRS
                                               Attach / Attach Failures and GPRS PDP Context Activation / Activation
                                               Failure. By comparing number of successes with failures the success rate can
                                               be calculated.

                                               This success rate should be very high, as a correctly configured MS (with
                                               correct data for that MS in the SGSN) almost never fails. Figure 3-3 shows a
                                               TEMS report example.




                                               Figure 3-3   Report example counting different events

                                               The Attach failure rate should be calculated as the number of Attach Failures
                                               compared to the sum of the Attach and the Attach Failures (same procedure
                                               applies for the Activate PDP Context) – in the above example this is 0 %,
                                               corresponding to a 100% success rate.

                                               Comparing the actual number of Attach to the number of GPRS PDP Context
                                               activation messages is difficult, as there might be session interruptions before
                                               the next step in the command sequence.

                                               When the success rate (or failure rate) has been calculated, the time to
                                               complete the procedure can be calculated to further analyze the performance.

                                               In order to analyze the time distribution of these procedures a text export can
                                               be done, where the information elements Attach time and PDP Context time
                                               are exported.

                                               A selection of the messages, Attach Complete and Activate PDP Context
                                               Accept in the resulting text files, will give one time sample per event. The
                                               selection can be done in any suitable program such as: Access, any text
                                               viewer/editor capable of handling large / multiple text files, Perl etc.
                                                                                               Open
                                                                                               INSTRUCTION                                    8 (8)
Prepared (also subject responsible if other)                                                   No.

EAB/ZG/NI Robert Blom                                                                          EAB/G-04:000245 Uen
Approved                                                                      Checked          Date                  Rev          Reference

EAB/ZG/NI [Johannes Lindvall]                                                                  2004-02-10            A



                                                                              Activate PDP Context Time Distribution

                                                                    250

                                                                    200



                                                 No of samples
                                                                    150

                                                                    100

                                                                     50

                                                                      0
                                                                              0

                                                                                    0

                                                                                           0

                                                                                                0

                                                                                                        0

                                                                                                             0
                                                                      50




                                                                                                                    00

                                                                                                                            50

                                                                                                                                  00
                                                                            20

                                                                                  35

                                                                                         50

                                                                                               65

                                                                                                      80

                                                                                                            95
                                                                                                                 11

                                                                                                                           12

                                                                                                                                 14
                                                                                                      Time [ms]

                                               Figure 3-4                    Activate PDP Context Time distribution graph

                                               An example of a time distribution graph of Activate PDP Context is shown in
                                               Figure 3-4. Most samples of the response time are in the 400 – 500 ms range.
                                               Only a few are longer then 1 second and the longest time is around 1.4
                                               seconds. The above example shows good performance for Activate PDP
                                               context as the success rate is high and the response time is low.


                                                                            GPRS Attach Time (Request - Complete)

                                                                    300
                                                                    250
                                                    No of samples




                                                                    200
                                                                    150
                                                                    100
                                                                    50
                                                                     0
                                                                          1.7 1.8 1.9   2 2.1 2.2 2.3 2.4 2.7 2.9     3 3.2 3.3 3.4 4.2 8.3
                                                                                                 Time [0.1 s bins]


                                               Figure 3-5                    GPRS Attach Time distribution graph

                                               An example of a time distribution graph of GPRS Attach is shown in Figure
                                               3-5. The time to complete the GPRS Attach procedure are fairly long, most
                                               samples are in the 1.9 to 2.1 second range. The reason for the relatively long
                                               attach time can be found if signaling during the Attach procedure is analyzed
                                               in the Layer 3 messages window in TEMS Investigation.
                                                                             Open
                                                                             INSTRUCTION                                       9 (9)
Prepared (also subject responsible if other)                                 No.

EAB/ZG/NI Robert Blom                                                        EAB/G-04:000245 Uen
Approved                                                      Checked        Date             Rev        Reference

EAB/ZG/NI [Johannes Lindvall]                                                2004-02-10       A

                                               The Attach procedures shown in Figure 3-5 are started with the Attach
                                               Request being sent from the MS to the SGSN. The SGSN then starts
                                               authentication procedure and sends Authentication and Ciphering request
                                               message and MS answers with Authentication and Ciphering response. The
                                               Attach procedure is then completed by receiving Attach Accept from the
                                               network followed by Attach Complete from the mobile.

                                               The time between the different messages are a few hundred milliseconds, but
                                               the total time ends up being around two seconds. A few samples of even
                                               longer Attach time exist and these are due to a delay in the reception of the
                                               Authentication Request message.

                                               Note that the network used in the example above could be configured to use
                                               authentication less often, which would reduce the average Attach time.

                                               Ping success rate (or failure rate, sometimes denoted the packet loss rate)
                                               can be calculated with a TEMS Investigation report by adding the events Ping
                                               Response, Ping Success and Ping Timeout, as can be seen in Figure 3-6




                                               Figure 3-6   Ping response report

                                               If a Ping delay threshold is also selected in the report, then a cumulative
                                               distribution graph will be generated for the time distribution, se examples in
                                               Figure 3-7 - Figure 3-9.

                                               In these figures it is evident that different set-up of ping measurements will
                                               give different results. I.e. it is important to use similar settings for any type of
                                               comparisons of ping measurements.

                                               With a 32 byte ping message and 100 ms delay between sequential pings the
                                               normal range of ping response is 600 – 700 ms, as can be seen in Figure 3-7.
                                               The result of increasing the time between sequential pings to 5 seconds is
                                               that ping response in many cases takes more than 1 second, see Figure 3-8.
                                               One contribution to this is that the closure of TBFs is normally delayed
                                               somewhat to have an associated signaling channel open for a short time span
                                               after data transfer (which improves set-up time of next TBF).

                                               Larger packets also gives a longer ping response time, as can be seen in
                                               Figure 3-9.
                                                                           Open
                                                                           INSTRUCTION                                    10 (10)
Prepared (also subject responsible if other)                               No.

EAB/ZG/NI Robert Blom                                                      EAB/G-04:000245 Uen
Approved                                                     Checked       Date            Rev       Reference

EAB/ZG/NI [Johannes Lindvall]                                              2004-02-10      A




                                               Figure 3-7   Ping delay distribution, size 32 byte 100 ms between ping




                                               Figure 3-8   Ping delay distribution, size 32 byte 5 s between ping




                                               Figure 3-9   Ping delay distribution, size 1460 byte 100 ms between ping
                                                                           Open
                                                                           INSTRUCTION                                  11 (11)
Prepared (also subject responsible if other)                               No.

EAB/ZG/NI Robert Blom                                                      EAB/G-04:000245 Uen
Approved                                                     Checked       Date            Rev       Reference

EAB/ZG/NI [Johannes Lindvall]                                              2004-02-10      A

                      3.3                      GPRS Radio Network Test

                                               These tests are performed in order to test the quality of the radio network and
                                               performance of radio network functionality such as cell reselection. Actual
                                               application used is not primarily of interest but should provide as stable
                                               throughput as possible. This simplifies the analysis of the radio network.

                                               The normal test for this is with FTP download (or upload), as this will provide
                                               a stable throughput. Note that HTTP also could be used to initiate long and
                                               stable downloads similar to FTP if the HTTP object is suitable.

                                               Typical measures are RLC throughput, cell reselection radio link outage time
                                               etc.

                      3.3.1                    Setup example

                                               An example of a command sequence for performing a FTP Get test is shown
                                               in Figure 3-10. In this example a very large file is downloaded where the time
                                               to download is longer than the planned testing time. Repetition of the
                                               procedure can be used when smaller files are downloaded.




                                               Figure 3-10 Command sequence example of FTP Download

                      3.3.2                    Analysis examples

                                               Three different kinds of performance indicators are important when the GPRS
                                               radio network is analyzed with TEMS Investigation.

                                               1. General performance indicators
                                                  These performance indicators should ideally be affected by as many
                                                  possible problems as possible in order to quickly identify performance
                                                  degradation.

                                               2. Application characteristics performance indicators
                                                  These performance indicators should give guidance on how application
                                                  behavior is affecting the measured radio network performance, i.e. trying
                                                  to identify areas where measured radio network performance is limited by
                                                  the application layer.
                                                                           Open
                                                                           INSTRUCTION                                  12 (12)
Prepared (also subject responsible if other)                               No.

EAB/ZG/NI Robert Blom                                                      EAB/G-04:000245 Uen
Approved                                                     Checked       Date            Rev       Reference

EAB/ZG/NI [Johannes Lindvall]                                              2004-02-10      A

                                               3. Specific performance indicators
                                                  These performance indicators should be as specific as possible in order to
                                                  quickly pinpoint found problems.

                      3.3.2.1                  General Performance Indicators

                                               The most widely used general problem indicator is the RLC throughput. It will
                                               be affected by both capacity and radio link quality problems.

                                               It is important to know what RLC throughput should be considered good and
                                               what should be considered bad. The used FTP server should be able to
                                               provide a stable and high data stream to the mobile under stable radio
                                               conditions, thus it is important to in advance test this.




                                               Figure 3-11 Stable FTP Download

                                               In the above graph the used FTP server provides a stable and high RLC
                                               throughput (>50 kbit/s) over four timeslots. If no quality and capacity problems
                                               exist in a network the RLC throughput should be around this figure for a four
                                               slot mobile, see reference [1].

                                               In the above figure it is evident that the RLC throughput is the performance
                                               indicator with the smoothest variation in time. This is a result of that
                                               throughput measurements always are time averages and the calculations will
                                               only be based on “completed” blocks / frames / packets on the various
                                               protocol layers. RLC throughput will thus have a “smoother” variation due to
                                               the fact that RLC data blocks are considerably smaller then LLC frames or IP
                                               packets.

                                               Other general problem indicators are: Areas with no GPRS coverage and
                                               abnormal releases of TBFs.
                                                                           Open
                                                                           INSTRUCTION                                   13 (13)
Prepared (also subject responsible if other)                               No.

EAB/ZG/NI Robert Blom                                                      EAB/G-04:000245 Uen
Approved                                                     Checked       Date            Rev        Reference

EAB/ZG/NI [Johannes Lindvall]                                              2004-02-10      A

                      3.3.2.2                  Application characteristics Performance Indicator

                                               The application characteristics performance indicator available in TEMS is the
                                               Application Throughput. It will “follow” RLC throughput in the sense that when
                                               capacity or radio link performance limits what RLC throughput can be
                                               achieved on, then the application throughput will also be degraded.

                                               In some situations the application layer will limit the measured throughput on
                                               lower layers (RLC/LLC etc). This is due to that the application layer will not
                                               feed data fast enough to the BSS to keep the radio link saturated. A typical
                                               example of this is when one IP packet is lost somewhere between the FTP
                                               server and the GPRS MS as in Figure 3-12.




                                               Figure 3-12 Lost IP packet causing lower RLC throughput

                                               In the above figure an IP packet is lost on the downlink and the TCP layer in
                                               the computer will ask for retransmission of that packet.

                                               The application throughput in TEMS is measured on top of the TCP layer
                                               (WinSock), i.e. data is only included in calculations when it is in the correct
                                               sequence. On measured application layer throughput, a lost packet leads to
                                               that application throughput goes down to zero until the missing packet is
                                               received. Then a lot of data is transmitted through WinSock, which gives a
                                               “spike” in the application throughput.

                                               Due to the missing packet TCP will send packets at a lower rate and try to
                                               build up transmission speed. This will be seen as a low but slowly increasing
                                               application and RLC throughput.

                                               This example shows low RLC throughput that is not caused by any problems
                                               in the radio network but on application layer behavior. More information on
                                               application behavior can be found in reference [2].
                                                                           Open
                                                                           INSTRUCTION                                 14 (14)
Prepared (also subject responsible if other)                               No.

EAB/ZG/NI Robert Blom                                                      EAB/G-04:000245 Uen
Approved                                                     Checked       Date           Rev       Reference

EAB/ZG/NI [Johannes Lindvall]                                              2004-02-10     A

                                               One way to become more familiar with the application layer characteristics is
                                               to use a packet sniffer tool such as Ethereal or WinDump and collect all IP
                                               data received. It will then be easy, although time consuming, to identify lost
                                               packets, retransmission etc. on the application layer.

                      3.3.2.3                  Specific Performance Indicators

                                               The potential problems limiting the performance for a GPRS mobile can be
                                               grouped into a number of problem areas from the mobile point of view.

                                               1. Low cell capacity available
                                               2. Sharing of resources
                                               3. PCU capacity limitations
                                               4. Bad radio link quality
                                               5. Cell re-selection problems

                                               In this chapter these problem areas are explained briefly and useful
                                               performance indicators for respective area are listed.

                                               Low cell capacity available

                                               This is when the available cell resources are not sufficient to give maximum
                                               bandwidth possible to a specific user. It is normally caused by CS and PS
                                               users competing over the same resources. The impact on the experienced
                                               performance depends on network configuration, prioritization between CS and
                                               PS and how high the competition is.

                                               Medium competition for resources

                                                  CS and PS users will compete for shared resources, i.e. on-demand
                                                  PDCHs. The capacity available for PS is limited but competition is not so
                                                  strong that there are long outages of GPRS service. This is normally seen
                                                  in cells with high CS traffic. The impact on the measured performance
                                                  depends on how scarce these resources are, but indicators are:

                                                  -   Low RLC Throughput
                                                  -   Low number of used timeslots
                                                  -   Packet PDCH releases and Packet TBF releases
                                                                          Open
                                                                          INSTRUCTION                                   15 (15)
Prepared (also subject responsible if other)                              No.

EAB/ZG/NI Robert Blom                                                     EAB/G-04:000245 Uen
Approved                                                     Checked      Date            Rev       Reference

EAB/ZG/NI [Johannes Lindvall]                                             2004-02-10      A




                                                         1




                                               Figure 3-13 Low cell capacity example

                                               Performance indicators for low cell capacity are shown in the above figure.
                                               MS is only assigned two TS when entering a cell due to high amount of CS
                                               traffic. Additional CS traffic causes the release of used resources, seen here
                                               at 1 as a Packet PDCH Release message. The result of the low cell capacity
                                               is seen in all performance indicators: Low RLC throughput, low number of
                                               used DL timeslots and Packet PDCH Release messages.

                                               High competition for resources

                                                  If there are no dedicated PDCHs and no idle TCHs that can be used as
                                                  on-demand PDCHs, it will not be possible to assign resources to a GPRS
                                                  mobile when it requests resources in the Channel request message. This
                                                  will lead to GPRS service outage in that cell.

                                                  In addition to performance indicators listed for medium competition
                                                  example above the following might be seen.

                                                  -   Channel Request without answer or PS Immediate Assignment Reject
                                                      message as a response to channel requests.

                                               Sharing of resources

                                               If low amount of cell resources exist, sharing between users occur and lower
                                               throughput will be the result. This is not the only reason for sharing as the
                                               amount of sharing can be one part of the GPRS network design in the same
                                               way as some congestion is planned for in GSM (GoS).
                                                                           Open
                                                                           INSTRUCTION                                   16 (16)
Prepared (also subject responsible if other)                               No.

EAB/ZG/NI Robert Blom                                                      EAB/G-04:000245 Uen
Approved                                                     Checked       Date            Rev       Reference

EAB/ZG/NI [Johannes Lindvall]                                              2004-02-10      A

                                               Thus it is impossible to say if it is too much sharing or not without knowing
                                               what quality the network is planned for. In TEMS it is very easy to identify
                                               sharing between users with the information element:
                                               -   Low RLC Throughput
                                               -   Other data on used TS (RLC data to other users on a shared timeslot).




                                                                 1


                                               Figure 3-14 Sharing of resources example

                                               Performance indicator for other data sent on timeslots used by the GPRS
                                               mobile is shown in the above figure. Percentage of the TS that is utilized by
                                               other users is shown in the information element Other data (%) as indicated
                                               by 1.

                                               PCU capacity limitations

                                               The reason could also be that there is a limitation in the number of available
                                               GSL devices in the PCU so that no PDCH can be created despite that free
                                               timeslots exist in the cell. This will give low number of used timeslots and low
                                               RLC throughput, although no release of PDCH or TBF should be seen.

                                               -   Low RLC Throughput
                                               -   Low number of assigned timeslots
                                               -   NO Packet PDCH Releases or Packet TBF Releases seen.

                                               If low number of timeslots is assigned for a number of cells throughout a
                                               drivetest then PCU capacity problems can be suspected and should be
                                               investigated.

                                               Bad radio link quality

                                               The impact of bad radio link quality depends on the severity of the radio link
                                               quality problems.

                                               For minor to medium quality problems the result will be that a lower coding
                                               scheme will be used. If it is not possible to change to a lower CS, either due
                                               to the usage of fixed CS or that the MS is already using CS-1, then the Block
                                               Error Rate will increase. As well as for CS the C/I is excellent to look at to
                                               evaluate radio link quality, i.e. good performance indicators are:
                                                                            Open
                                                                            INSTRUCTION                                17 (17)
Prepared (also subject responsible if other)                                No.

EAB/ZG/NI Robert Blom                                                       EAB/G-04:000245 Uen
Approved                                                     Checked        Date            Rev        Reference

EAB/ZG/NI [Johannes Lindvall]                                               2004-02-10       A

                                               -   Low measured C/I
                                               -   CS usage for data
                                                   High proportion of CS-1 indicates bad radio environment high proportion
                                                   of high CS indicates good radio environment.
                                               -   High Block Error rate
                                               -   Signal strength (RxLev)
                                                   Used to evaluate if quality problems are caused by low signal strength or
                                                   high interference.




                                               Figure 3-15 Bad radio link quality example

                                               The impact of the bad radio link quality on the performance indicators is
                                               shown in the above figure. RLC throughput is low, C/I is low, BER and BLER
                                               are high. Only CS-2 is used for data as link adaptation is not used, if link
                                               adaptation was used BLER would be lower, but CS-1 Usage would be higher.

                                               Severe radio link quality problems might lead to problems to: decode system
                                               information in the serving/neighboring cells, released TBFs etc. The following
                                               could be indicators of severe radio link problems:

                                               -   Slow update of SS, C1 & C2 parameters for neighboring cell.
                                               -   Failed cell reselection attempts to a particular cell
                                               -   Released TBF that was pre-ceded by bad radio quality
                                                                            Open
                                                                            INSTRUCTION                                   18 (18)
Prepared (also subject responsible if other)                                No.

EAB/ZG/NI Robert Blom                                                       EAB/G-04:000245 Uen
Approved                                                     Checked        Date            Rev        Reference

EAB/ZG/NI [Johannes Lindvall]                                               2004-02-10      A

                                               Cell re-selections

                                               Mobility is one of the key aspects of GSM and GPRS, which makes cell re-
                                               selections an integral part of any GPRS network. The impact on performance
                                               due to cell reselections depends on a many factors.

                                               Different applications will have different sensitivity to mobility problems, i.e. a
                                               user downloading small wap pages have very short transfers (in the range of
                                               seconds) and the probability of a cell reselection occurring during this time will
                                               be low. A user that is stationary will also seldom experience cell reselections,
                                               irrespective of the used applications.

                                               The users that will experience mobility problems are users doing long data
                                               transfers, while moving around in the network. Drivetesting a GPRS network
                                               with TEMS can normally be seen as a worst case for mobility problems.

                                               Cell reselections will in general be seen as a temporary outage of the link
                                               between the MS and the application server. In many cases the used
                                               applications will not suffer too much from these short interruptions in the data
                                               transfer. Protocols such as TCP will ensure that lost packets will be re-sent.

                                               However when cell re-selections becomes to long or there are to many within
                                               a short timeframe then the user performance can be degraded more than
                                               wished for. Measures to look out for is then:

                                               -   Cell re-selection outage time
                                               -   Cell re-selection frequency

                                               The cell re-selection outage time can be measured on different layers, e.g. on
                                               radio link level or on application layer level. The following messages are
                                               important when analyzing cell re-selection outage time on the radio link.

                                               Packet Downlink Ack/Nack in the old cell
                                                     Normally the last message sent in the old cell

                                               System Information Type 13 in the new cell
                                                     Contains system information specific for GPRS that must be read from
                                                     the new cell before requesting a packet data channel. This is normally
                                                     one of the main contributors to the radio link outage time as the time
                                                     between sending different SI 13 messages on the BCCH can be in the
                                                     range of seconds.

                                               Channel Request in the new cell
                                                    A Channel Request is normally sent quickly after SI 13 has been read
                                                    in the new cell.

                                               Immediate Assignment in the new cell
                                                    An Immediate Assignment to a PDCH is normally received quickly
                                                    after a channel request. This can be seen as the re-establishment of
                                                    the radio link, as there now can be signaling between BSS and MS.
                                                                            Open
                                                                            INSTRUCTION                                   19 (19)
Prepared (also subject responsible if other)                                No.

EAB/ZG/NI Robert Blom                                                       EAB/G-04:000245 Uen
Approved                                                     Checked        Date            Rev        Reference

EAB/ZG/NI [Johannes Lindvall]                                               2004-02-10      A

                                               Packet Uplink Ack/Nack in the new cell
                                                     The Packet Uplink Ack/Nack is normally received quickly after the
                                                     Immediate assignment indicating that data transfer has occurred on
                                                     the uplink, i.e. a verification that radio link is established and working.

                                                      Thus the time from the last Packet Downlink Ack/Nack in the old cell to
                                                      the first Packet Uplink Ack/Nack in the new cell can be seen as a good
                                                      measure of the radio link cell reselection outage time.

                                               Packet Downlink Assignment in the new cell
                                                     When the used test application is a downlink transfer, it can not really
                                                     be considered as re-established until a TBF in the downlink exists. The
                                                     Packet Downlink Assignment indicates that a new downlink TBF has
                                                     been established.

                                               Packet Downlink Ack/Nack in the new cell
                                                     Packet Downlink Ack/Nack message normally indicate that data has
                                                     been transferred over the radio link, i.e. the time from the last Packet
                                                     Downlink Ack/Nack message in the old cell to the first in the new cell
                                                     could be seen as a closer measure to the user experience outage
                                                     time.

                                                      This measure will however be difficult to analyze, as it is the BSS that
                                                      controls when TBF should be initiated on the downlink. This could be
                                                      done before actual user data arrives to improve the BSS performance.




                                               Figure 3-16 Cell re-selection example
                                               Signaling at cell re-selection is shown in the figure above. The time from the
                                               last Packet Downlink Ack/Nack in the old cell to the Immediate Assignment in
                                               the new cells is 2.2 second. Normal values for intra routing area update cell
                                               re-selections are from 2 to 4 seconds.
                                                                                           Open
                                                                                           INSTRUCTION                               20 (20)
Prepared (also subject responsible if other)                                               No.

EAB/ZG/NI Robert Blom                                                                      EAB/G-04:000245 Uen
Approved                                                               Checked             Date                Rev      Reference

EAB/ZG/NI [Johannes Lindvall]                                                              2004-02-10          A

                                               The different contributions to the cell re-selection outage time can be
                                               analyzed by exporting data and calculating the time difference between
                                               messages in the cell re-selection procedure.

                                                                                     Cell Reselection Time
                                                              20
                                                              18                                                        Packet DL
                                                                                                                        Assignment
                                                              16
                                                   Time [s]   14                                                        Immediate
                                                              12                                                        Assignment
                                                              10
                                                               8                                                        Channel
                                                                                                                        Request
                                                               6
                                                               4                                                        SI_13
                                                               2
                                                               0
                                                                   1

                                                                       3

                                                                                 5

                                                                                       7

                                                                                                 9

                                                                                                     11

                                                                                                          13

                                                                                                                   15
                                               Figure 3-17 Cell re-selection outage time distribution example
                                               In the figure above it can be seen that the a number of cell reselections have
                                               very long outage time (> 4 seconds), the reason for this should be
                                               investigated. It can also be seen that time to read System Information 13
                                               message is the main contributor to the total radio link outage time.

                      3.4                      General Tips & Tricks

                                               •       Try to introduce waiting time between commands in the command
                                                       sequence if unexplainable failures occur (remember to wait on the same
                                                       MS as commands are sent to!).

                                               •       Set the TEMS mobile to operate in GSM only mode. This means that it will
                                                       not attach at power-on but when needed – with this setting log files will
                                                       always start with attach & PDP context activation (and network will have
                                                       one less user doing GMM signaling when not needed).
                                                                           Open
                                                                           INSTRUCTION                                   21 (21)
Prepared (also subject responsible if other)                               No.

EAB/ZG/NI Robert Blom                                                      EAB/G-04:000245 Uen
Approved                                                     Checked       Date            Rev       Reference

EAB/ZG/NI [Johannes Lindvall]                                              2004-02-10      A



                      4                        Overall Procedures
                                               The reason for doing GPRS performance measurements can vary and two
                                               important reasons are: GPRS functionality tests and GPRS radio network
                                               tests.

                                               GPRS functionality tests should primarily be done to verify normal behavior
                                               and to do functionality comparisons (over time, parameter changes, vendors
                                               etc.)

                                               -   A correctly defined MS together with correct configuration of core nodes
                                                   should have very high Attach & Activate PDP Context success rate

                                               -   Actual time measured for the procedures will depend on network
                                                   configuration and general vendor performance

                                               These tests should preferably be done under controlled radio conditions
                                               without limitations in the radio network.

                                               The tests should be performed every now and then to verify “normal”
                                               performance levels, procedure completion time, do comparisons at
                                               system/functionality upgrades etc.

                                               GPRS radio network tests should be performed continuously.

                                               It is important to have established what is “normal” performance to use as a
                                               comparative measure. The test set-up of the radio network performance
                                               measurements should thus be controlled now and then with FTP transfers
                                               under ideal radio network conditions. This ensures that the application layer is
                                               not varying too much, which will make the analysis of the radio network
                                               performance much easier.

                      5                        References
                                               GPRS Measurements in the TEMS Products                              [1]

                                               Application-level Data Service Measurements in                      [2]
                                               TEMS Investigation GSM

				
DOCUMENT INFO
Shared By:
Stats:
views:440
posted:11/4/2012
language:Latin
pages:21
Description: GPRS, Drive Testing,dt,tems, quick guide