data throughput measurement on application layer

[punto]

Hello,
I would like to open a new subject , surely interested for all of us > data throughput measurements on application layer.

Please, share which SW tools and methods you use to measure 2G/3G data throughput measurements , and respond time of your networks ... could someone share some report and some advices and know how about measuring while making http, wap, email, DL UL ftp, ping ... ?

Lets share ... :)
P:)

[Lesmana]

Very Good Topic....

The VIP Members should response.....
I mean the super engineer...Hassen...

Hopefully Hassen is the right person for this topic..

thanks

[wolverine]

I use the following tools in my company..

For application layer throughput monitoring..

Netpersec & DU meter

For latency measurements..

ping command from the DOS prompt & Perfping

For throughput measurements using UDP there is a very good utility called UDP flood that you can run on a test server hanging off your GGSN. This way you can measure throughput with worrying about TCP peculiarities..

[Mr.David]

ethereal is the usefull one

[hassen]

we use business objects for gprs statistical according to counters and formula
ping to know time responce
The use of WAP to access the various servers of the operators
create a session at SGSN and from drive test make downloading in order to measure
throughput .

[punto]

Thanks Hassen and all :)

@all :)
Do you use only ping volume of the data 32 bytes ... to be assured that you have information how customer see response of your network ... or , you have done pinging with bigger then 32 bytes of volume of data sent by this command ... maybe, to be closer to simulate some service access , like http ? Logically I think, we should discard first PING result in making precise statistics of consecutive PINGs, as at start TBF takes time to establish ... and we are looking for average of all resuts made by some sequence (except that first longer one ... ) Do you agree ?

[wolverine]

I would suggest varying the ping size in order to uncover any potential bottlenecks you might have. Perfping has a nice interface that allows you to do that and will also start off with a small ping packet and keep increasing it.

[punto]

Thanks M :)
what is range of ping size/volume you set in Perfping ... from minimum xx bytes , to how mucho Max yy bytes ? Can you share with us install of these nice tools ? Correct me if I am wrong , but I am not sure they were offered by someone before ... If is not a hassle to you ... we know how to return smile and thanks ...

Also, here is link to many SW tools from the similar family of tools, like Perfping ... useful, scroll down ..
http://3d2f.com/pcgi/search.cgi?s=23737&t=2&w=2
P:)

[da_manole]

we use business objects for gprs statistical according to counters and formula
ping to know time responce
The use of WAP to access the various servers of the operators
create a session at SGSN and from drive test make downloading in order to measure
throughput .

Actually Zte BSC use rollcall polling way to process ping test packets which is different from traffic packet. And the way increases the ping packets' delay..

"Because ping is only one test tool with which we test the link connected or not, so we use rollcall
polling way. With rollcall polling way, we can decrease the ping packets’ processing load and the
influence of normal traffic packets.
However we do not the rollcall polling way to process traffic packets. In one word, the delay of ping
packets does not mean that the true delay of traffic packets."

[RF engineer]

Any one can help me what is the acceptable RTT(latency) for both GPRS and EDGE .

[adedia]

Netpersec were accurate considerig it look at the application level...
But in my opinion ethereal are better.. since you can drill down inside. Provided you know how to use it correctly

[eritelecom]

hello all,

can anyone tell me why in HSDPA log files we see some times that data throughput is higher than RLC throughput ?
logically, we must have RLC throughput usually higher than application throughput !!

[zloy_qwer]

hello all,

can anyone tell me why in HSDPA log files we see some times that data throughput is higher than RLC throughput ?
logically, we must have RLC throughput usually higher than application throughput !!


Application throughput sometime can be higher, than possible maximum throughput because of buffer of Windows. Try to play with parameter TCP_Window_size (you can change it using soft Dr TCP/IP)

[punto]

*has anyone serial or so .... for SW on this link http://www.pingplotter.com/download.html ?

* Can someone tell more about UDPflood ?
BR,
P:)

[budo8002]

use free edition... it's enough

[punto]

http://www.colasoft.com/ping_tool/

This is free and nice SW , 3D diagram and changeable ping parameters !

p.s. is not cocacola software :), just adding some humor to our engineering world !

[Rooster[BG]]

Hello,

In my company we use 32B and 1460B with 1 and 5 seconds delay between pings. for EDGE I am using "Delay between request and reply" option and for GPRS - "Delay between requests". I am using Nemo Outdoor for the measurements and Ethreal for IP traces.
You can use FTP UL/DL of binary files to check the throughput on application layer.


Any one can help me what is the acceptable RTT(latency) for both GPRS and EDGE .

Ping RTT depends on the network parameters. For example, if you change the initial coding scheme from MSC3 to MCS6 you can get better results.

[da_manole]

it's very good and it's freeware
http://www.wireshark.org/download.html

[gprastomo]

Any one can help me what is the acceptable RTT(latency) for both GPRS and EDGE .

For 3G/HSDPA < 80ms, maybe for GPRS little bit more

br

[RF engineer]

Thank you for your response

[hemaldoshi14]

Acceptable Value of RTT is:
For GPRS, it should be less than 500 msec
For EDGE, it should be less than 250 msec


Any one can help me what is the acceptable RTT(latency) for both GPRS and EDGE .

[ghamiz]

boss can you please shed some more light on that, i am facing the same problem in which the application layer throughput is greater than RLC throughput and am confused :S

[ahmedrad]

hey guys,

I use a tool that we developed that does all kinds of tests and is really insightful .. we normally test the following

1- HTTP page load time
2- time till first object appears on the page
3- time for the DNS lookup
4- time for the TCP establishment
5- ping RTT
6- ping loss
7- HTTP download throughput

we are planning on adding FTP throughput and e-mail delivery time also :)

[titoe80]

Application throughput sometime can be higher, than possible maximum throughput because of buffer of Windows. Try to play with parameter TCP_Window_size (you can change it using soft Dr TCP/IP)

Dear,

Could you explain more detail? How buffer of Windows makes application throughput sometime can be higher than possible maximum throughput?

thanks

[paranoja]

we use sigos for testing various e2e traffic cases performance.
it emulates the cellphone and you can put the units over a different coverage areas.
for example if you have 5 MSC's and 10 BSC you position the units so your calls/web browsing/sms goes through every network element.

pretty good tool...