 jimkyleBtrieve GuyPremium
join:2002-10-20
Oklahoma City, OK
kudos:2
 ·AT&T Southwest
| Need help interpreting throughput graph
Wireshark test |
My connection is rated at 1536; I've avoided upgrading to any higher speed since getting ADSL back in 2002, because I have the grandfathered 5-true-static-IP plan and don't want to lose those true static IPs. This makes my normal download speed from any site that supports that speed come out to approximately 160 KBytes/second, but lately I've been seeing a maximum of around 40 and it often drops to single digits.
As I read this wireshark graph, it seems to be holding steady at the rated speed, but dropping significantly for a few seconds about every 11 seconds during the test download (from the site suggested in the FAQ). Anyone have ideas of what might be causing this, or the drop in speed during normal operation?
--
Jim Kyle |
|
 d_lBarsoomPremium,MVM
join:2002-12-08
Reno, NV
kudos:7 | Jim, I think you are reading it right. I don't really know what causes those periodic slowdowns (every 11 secs). We've seen them before at times and usually they are minor compared to the overall problem at hand. They may even be due to something in the RWIN tweaking if you are running XP.
I think you should post in Direct: » AT&T Direct and let them look at your line.
There seems to be a new type of throughput problem cropping up here between yours and the other thread.
Good Luck.
Dave
--
TCE Weather |
|
|
|
jimkyleBtrieve GuyPremium
join:2002-10-20
Oklahoma City, OK
kudos:2 Reviews:
·AT&T Southwest
| said by d_l:I think you should post in Direct: »AT&T Direct and let them look at your line.
There seems to be a new type of throughput problem cropping up here between yours and the other thread. I've done so, and will pass back anything I learn there.
I'm wondering if it might be some form of throttling, connected with the other thread here about discontinuing DSL in OKC. I've not gotten any notification of such action, although some of the Uverse sales folk that are going door to door have said that will be happening. However, back in the 70s (!!!) a Cox Cable salesman assured me they would have internet connection within a year. It finally arrived a decade or more later, and by that time I had abandoned Cox in favor of DirecTV...
--
Jim Kyle |
|
d_lBarsoomPremium,MVM
join:2002-12-08
Reno, NV
kudos:7 | reply to jimkyle
Jim, what modem model are you using? I'm curious about the possibility of interactions of certain modems with some AT&T gateway settings.
Whenever throughput speeds are reduced from the normal levels, you could say that throttling exists whether it is intentional or not. A misconfigured profile at the gateway should produce a throttled graph like #4 in the FAQ with some sharp speed ups and slowdowns due to bufferig. Your graph doesn't have any of the speed ups like in the other thread (which are even minimal compared to #4).
I'm wondering if the Speedstreams (5100b, 4100, 4100b), that were used for most of the graphs in the FAQ, might have a different response to possible throttling at the gateway because of their ACK buffers.
--
TCE Weather |
|
jimkyleBtrieve GuyPremium
join:2002-10-20
Oklahoma City, OK
kudos:2 Reviews:
·AT&T Southwest
| It's a 4100, 5-light model.
That almost constant, absolutely flat line on my graph surprised me. I was expecting to see a lot more fluctuation. Next time I'm doing a significant download (such as a new distribution of several hundred MB) I'll do another graph to see if there's any major difference.
--
Jim Kyle |
|
jimkyleBtrieve GuyPremium
join:2002-10-20
Oklahoma City, OK
kudos:2 Reviews:
·AT&T Southwest
| reply to d_l
Second test graph |
--
Jim Kyle |
|
d_lBarsoomPremium,MVM
join:2002-12-08
Reno, NV
kudos:7 | reply to jimkyle
Now that speed fluctuation looks more like an exhausted router graph. These sort of slow downs occur more on weekday evenings and not on Sat. mornings.
--
TCE Weather |
|
jimkyleBtrieve GuyPremium
join:2002-10-20
Oklahoma City, OK
kudos:2 Reviews:
·AT&T Southwest
| I also noticed on the wireshark displays that the RWIN (which adjusts dynamically on Linux) was around 25000 on the capture that made this graph, while on a later capture in the same download (which is not yet completed), it dropped to around 2500. This might also be part of the problem...
--
Jim Kyle |
|
