3G Troubleshooting: End to End Slow Throughput Performance

[yoronjo]

Would like to seek everyone's thoughts on the case i'm working on:

Issue: Slow HSDPA Throughput

Initial Findings:

1. Tested HSDPA in 3 different Node Bs (IBS) in 3 different respective RNCs.
2. RSCP and Ec/Io measurements are excellent
3. No concerns with HO since i'm inside IBS
4. No drops, no abnormal RRC releases or whatever. no disconnects.
---
4. I'm getting slow throughput when viewing 1 youtube window, roughly 400kbps
5. when viewing multiple youtube videos simultaneously, throughput increases up to 1.6 Mbps with more videos opened

I seriously have no idea how to achieve the high throughput i'm getting at multiple videos when i'm opening just 1 video.


thoughts? anyone? thanks a lot

[siankirene]

Would like to seek everyone's thoughts on the case i'm working on:

Issue: Slow HSDPA Throughput

Initial Findings:

1. Tested HSDPA in 3 different Node Bs (IBS) in 3 different respective RNCs.
2. RSCP and Ec/Io measurements are excellent
3. No concerns with HO since i'm inside IBS
4. No drops, no abnormal RRC releases or whatever. no disconnects.
---
4. I'm getting slow throughput when viewing 1 youtube window, roughly 400kbps
5. when viewing multiple youtube videos simultaneously, throughput increases up to 1.6 Mbps with more videos opened

I seriously have no idea how to achieve the high throughput i'm getting at multiple videos when i'm opening just 1 video.


thoughts? anyone? thanks a lot
hi yorojo,
that true....if you open many streaming video in youtube you'll get more high speed, i think is like a trigger to improve trhgput.

if you open only 1 video, thrgput small...this maybe because your only have 1 request (UL) to network and if you open many (maybe 4~5 video), every video will send the request (UL) at the same time and every video will get DL.....so you got high thrgput because your tools will sum thgrput from all video.

i think like that,


correct me if i wrong....ty:)

[samirzubair]

In the RNC450 the HW consists of A2SU,GTPU and NIS cards, whereas in the RNC2600 these cards are replaced by the NPS card

[yoronjo]

Hi Guys,

we're doing some end to end tests, and it seems that the concern falls in the core network. We're still doing some tests, and hopefully after the testing period we can come up with some conclusions and recommendations as to what is causing the trouble.

@siankerene,

i think you have a point there, more requests for downlink speed in the uplink will result to more provisioning of downlink speed by the server if the server is capable. I'll check on that point also.

Thanks a lot for the inputs. really helpful, cheers :)

[yoronjo]

Hi Guys,

we're doing some end to end tests, and it seems that the concern falls in the core network. We're still doing some tests, and hopefully after the testing period we can come up with some conclusions and recommendations as to what is causing the trouble.

@siankerene,

i think you have a point there, more requests for downlink speed in the uplink will result to more provisioning of downlink speed by the server if the server is capable. I'll check on that point also.

Thanks a lot for the inputs. really helpful, cheers :)



update

did initial troubleshooting, and we have bypassed one of the bandwidth managers our postpaid and prepaid are passing through (they aren't being managed since the bwm manages other types of subscribers, but the thing is indirectly the element apparently affects the throughput of the 3g postpaid and prepaid subscribers).

throughput improved. streaming increased by 400-600 kbps. :)

[triple_patte]

When you're having significant throughput improvement by using more streams then you're having issue with your end to end round trip time. In acknowlegde based protocol such TCP, roundtrip time will affect overall throughput. In this situation, you're having more throughput by using multiple connection to the server because actually you're pipelining the connection. To verify this, try to check your round trip time by pinging your server. Then try throwing UDP packet from your server to your ue/client (since UDP does not need ACK). If the perceived througput is far larger compared your initial test (accessing 1 video stream) then most likely your e2e delay is your problem, not the rf.

[giegala]

Have you tried to benchmark your network with some other servers, that youtube. Google uses distributed servers and provide you with throughput they want, but not what is acheivable.

[riverstreamer]

As I can conclude from your different tests, you don't need to search in Rado/RAN area for your problem.
Search in Core (QoS services) to find restrictions for services, connections and delays.