Professional Documents
Culture Documents
GO - NAST3019 - E01 - 1 GSM End-To-End Optimization Methods 72
GO - NAST3019 - E01 - 1 GSM End-To-End Optimization Methods 72
Methods
Objectives
End-to-end optimization is based on the user perception an
d targeted on the service type. It is conducted to trace the
whole process of the service from the terminal, the radio ac
cessing network, the core network, to the service platform.
And the purpose is to optimize the quality of service from th
e application layer to the bottom layer.
This course describes the solutions for the end-to-end opti
mization of WAP, MMS and the Mobile Streaming Media. It
will help you get deeper understanding about end-to-end o
ptimizaiton.
Contents
PS Domain Attach:
The main purpose is to create the route to the SGSN. At first, UE sen
ds the RRC connection request. After the RRC is connected, UE will s
end "Attach Request" message on the GMM layer via initial direct sig
naling. After the signaling connection between RNC and SGSN is cre
ated, SGSN will send "Attach Accept" message (containing the new
P-TMSI of UE) to UE. UE replies with "Attach Accept Complete" mess
age, and the attach process is completed.
UE may have been attached to the PS domain already when it is start
ed. It also may start attach when the service request is sent. So, the P
S Domain Attach may bot be tracked for certain service.
After the PS Domain Attach is completed, there will be a logical rout
e between the UE and SGSN, and the UE can be recognized by the S
GSN.
© ZTE Corporation. All rights reserved
PDP Context Activation
After activating the PDP context and acquiring the IP address, the UE will use the
acquired IP address to send "WSP Connect" message to the WAP gateway, which
contains the parameters about the UE's capabilities and the formats of supported
contents.The WAP gateway will send "Connect Reply" to the UE. If needed, it may
modify the capability parameters of the UE according to its condition. After receiv
ing "Connect Reply", UE will send "WTP Acknowledge" message to the WAP gate
way, and the 3-way handshake connection process is completed.
The UE sends "WSP GET" message (containing URL, i.e. the addre
ss of the website visited by the subscriber) to the WAP gateway.T
he WAP gateway translates the message into “HTTP GET” and
adds the parameter of content type accepted by the UE, and sen
ds the “HTTP GET” messgae to the corresponding SP. In
normal cases, the SP will send "Response" (with the status code
“200,OK”) to the gateway.WAP gateway will translate the mess
age into "WSP Reply".If the “WSP Reply” is too large, WTP will
divide the message into several segments of "WTP Segmented R
esult" and send the segments to the UE. After receiving the comp
lete Reply message, UE will send "WTP Acknowledge" message t
o the WAP gateway, and the "GET" process is completed.
The UE sends "WSP Post" message (containing URL, i.e. the address of the website
posted by the subscriber) to the WAP gateway.The WAP gateway translates the m
essage into "HTTP Post" message and sends it to the corresponding SP. In normal
cases, the SP will send "Response" (with the status code “200,OK”) to the gatew
ay.WAP gateway will translate the message into "WSP Reply".After receiving the R
eply message, UE will send "WTP Acknowledge" message to the WAP gateway, an
d the "POST" process is completed.
The above table shows the main causes for the denial of Attach.If
the cause is "GPRS services not allowed", it means the subscriber
cannot use the PS domain service.
"Protocol error, unspecified" menas the protocol is in error. Usual
ly, the problem is caused by poor radio. Sometimes, it is caused b
y that the SGSN can not recognize different subscribers with the
same P-TMSI.
© ZTE Corporation. All rights reserved
UE sends Attach request, SGSN doesn't respond
The IP address type of the subscriber is wrong - the IPv6 address is changed to IP
v4 address.
The APN set by the subscriber is wrong - APN is changed to CMWAP or CMNET.
The subscriber has set static IP address - The IP address is changed to be obtaine
d automatically.
Durig the connection process between the UE and the WAP Gate
way, the perfomance of the UE terminal and the responsing
capability of the WAP gateway are examined, no matter it is WSP
connection on WAP1.0 or 3 times TCP handshakes on WAP2.0.
If the status code is 1XX, 2XX, or 3XX, it means the process is nor
mal; If the status code is 4XX, it means the UE terminal has some
problem; If the status code is 5XX, it means the server has some
problem.
Status Code
Categories Definition
1XX Message prompt Message prompt
2XX Succeeded The request from the client is accepted by the server successfully
The client is required to take more actions to implement the
3XX Redirected request.
4XX Client error The client may have some problem
The server may have some problem, or it cannot respond to the
5XX Server error request.
413,Request entity too larg The requested webpage content is too large (e.g. wmv and mp3 files), and it is lar
53 0.03%
e ger than the acceptable range of the terminal.
502 Bad Gateway 21 0.01% Gateway error
404 Not Found 19 0.01% The webpage is not found.
400 Bad Request 10 0.01%
Connection Delay
The connection delay is the time between the UE sends the RRC connection r
equest and the connection to the WAP gateway is completed, which includes
the RRC conneciton setup delay, the Attach delay (Optional), the PDP context
activation delay and WAP gateway connection delay.
If any of the time is too long, the user perception will be directly affected.
Download/Upload Rate
The purpose of this index is to prevent that the download/upload delays are
different due to the webpage sizes are different.
The segments on WAP 1.0 fall into the following three types:
[WTP]GTR TTR with Flag bit "not last packet": After receiving this type of seg
ment, the gateway or terminal will not reply with the acknowledge message
(Ack).
[WTP]GTR TTR with Flag bit "last packet of packet group": After receiving this
type of segment, the gateway or terminal must reply with the acknowledge m
essage (Ack) or with the re-transmission request (Nack) for the packets in the
packet group.
[WTP]GTR TTR with Flag bit "last packet of message": It means the transmissi
on of the packet group is completed.After receiving this type of segment, the
gateway or terminal may reply with the acknowledge message.During the M
MS transmission process, the gateway may not reply with Ack message, it mu
st forward the MMS to the MMSC then.
Type1
Unnecessary service traffics and network loads are increased due to that the g
ateway re-transmits the packets in a short period.
In the following figure, the interval of re-transmission is only 1ms. It is sugges
ted to optimize the gateway performance.
Type2
The gateway stops the transmission of the packet group before it is completel
y transmitted, and the terminal doesn't receive the due packets for a long tim
e. The transmission times out.
In the following figure, the gateway compleled the transmission of packet[3].
But it didn't continue to transmit the other packets in 5 secons, so the transmi
ssion timed out.
From the analysis of the signaling file, we find that sometimes the re-transmission
is caused by wrong sequence packet . As shown in the following figure, the gate
way sent packet[3] at first, but not packet[1] which is the correct packet according
to the sequence. Then the gateway responded with "Reply".After the terminal sent
the request for re-transmission, the gateway stopped data transmission just after i
t completed the transmission of the two packets before packet[3], and the termin
al continued to wait for the trasmission of the needed packets, the timeout was ca
used by that.
Normally, after the gateway sends a complete packet group, the terminal should r
espond with "Ack" message.If the terminal has some problem and cannot respon
d promptly, the gateway will re-transmit the last packet in seconds, thus the time
on transmitting the data is prolonged, and it may lead to timeout.
In the following figure, after the gateway completed the transmission of packet[2
9] (the last packet in the packe group), the terminal didn't respond. So the gatewa
y re-transmits packet [29] 3 seconds later, and the terminal responds to it then.
Step1
Step2
Step3
Step4
Step 1 and Step4: RADIUS start and RADIUS. The involved networ
k elements in the PDP Context Activation and PDP Context Deacti
vation processes are GGSN and WAP Gateway.
Step2 : Setting up the connection on WSP layer. For the MMS se
rvice on WAP2.0 protocol stack, the corresponding step is the sta
ndard TCP handshakes for 3 times, i.e. Syn, Ack, Syn and Ack. Aft
er getting reliable WSP/TCP connection with WAP Gateway, the
UE is ready to send MMS to the MMSC.
Step3: Tranmission of the MMS layer packets.If the MMS is on W
AP2.0 protocol stack, The transmission of Ack message sent from
the UE is completed when the TCP layer shows "Fin".
Basically, Step 1, 2 and 4 of MMS MT flow are the same with that of
MMS MO flow. The difference is in Step3. When UR receives the PUS
H message, it can use two methods to get the MMS. One method is
automatic get, i.e. the UE automatically sends GET request to the M
MSC according to the URI in the PUSH message, and automatically d
ownload the MMS.This process is unknown to the user.The other me
thod is manual get, i.e. the user can click the link to download the M
MS according to the URI in the PUSH message.The basic processes o
f the two mwthods are the same.But, if manual get is adopted, when
the user has no interest in the MMS, he may not extract the MMS, or
he may delete the PUSH message directly. Besides, the user may extr
act the MMS later.
After receiving the MMS, the user needs to feed back to the MMSC t
o notify that.
The average delay between the calling end sending the MMS and the called
end receiving the MMS.Analysis of the main factors influencing the delay of
end-to-end MMS.Here we assume that both MO and MT processes can be
monitored by us.
The MMS delay mainly occurs in the following stages: (listed according to th
e duration of delay)
The terminal sends data to the gateway.
The gateway queries the E - DNS for the user information
Preparation for transmitting the MMS packets, including setting up connection between
NEs (terminal - gateway, gateway - MMSC) and setting up PDP activation.
The gateway sends data to the MMSC.
We need to get the delay on each stage and check the time loss on each NE
in the network (radio, corenetwork, WAP gateway) according to the message
flow.Besides, since the MMS delay is depending on the length of the MMS, s
o we also check the length of the MMS.
Unknown 10
situations, the terminal types that have PHILIPS9A9R / Obigo Browser 2.0 3
CoreTek_WAP 3
KBT_L220A_B1/LB7M401A/WAP2.0 Profile 3
From the statistics result, we can see that all the Mobile News tim
eouts are after the gateway sends Conf. It means that the gatewa
y can respond normally, and the tomeout is in the data transmissi
on process and the terminal receiving process.The reason for this
kind of failure may be that the network is too busy or the radio tr
ansmission is congested.
Based on the telephone interview with some subscribers, we get
to know that both the download rate and the success ratio in the
peak hours are lower than that of the routine hours.
So it is suggested to send the PUSH message during the period o
ther than peak hours.
The average downloading rates of different websites are quite close to each other,
it means the downloading bottleneck is not at the SP, but at the radio accessing n
etwor.
The steaming media data (video and audio) are mainly beared by
RTP/UDP, the static image and text can be beared by HTTP.Capbi
lity Exchange and Presentation Description can be encapsulated
via HTTP or RTSP.RTSP and SDP are used to set up and control th
e session. MIME describes the media type and RTP is the transmi
ssion protocol carried by the streaming media.
If the signalig from OPTIONS to PLAY is considered as successful playing, the succ
ess ratio is 69.68%, If the signaling from SETUP2 (set up radio bearer for the audi
o) to PLAY is considered as successful playing, the success ratio is 99.74%.We can
say that the successful playing is depending on the successful setup of wireless be
arer.According to the signalings, there is no specific status code to indicate the re
ason for the setup failure. If we want to find out the reason, we have to consider t
he factors from multiple aspects, such as the radio environment and the condition
s of core network devices.
Number of
1679 1679 1185 1173 1170 986 982
times
Mobile TV playing delay refers to the time between the start of options and the st
art of PLAY. The longest duration is 35.141s, and the shortest duration is 768ms.
The following figure shows the playing times of mobile TV with different delays. T
he horizontal coordinates are the playing delays, and the vertical coordinates are t
he playing times of mobile TV.Most of the palying delays are within 2 seconds.
Play duration refers to the time between the "Play" signaling and
"teardown" signaling.The shortest duration is 494ms, and the lon
gest duration is about 33 minutes.We can see that most of the pl
ay connection time are less than 1 minute.
Some terminals may report the model of the media player, some may report the manufacturer and model of the term
inal.
The above table shows the top10 of the media players or terminal models that are most frequently used to play TV pr
ograms. From the table, we can see that RealMedia Player is most frequently used software to play mobile TV, and it
has the highest ratio of successful playing.The handset Nokia5300 is used to play mobile TV for 81 times, and the rati
o of successful playing is 71.6%.