Click to See Complete Forum and Search --> : Speed test on DSL line increases latency...how come ?
BobY777
09-12-2006, 07:45 AM
I was using Ping Plotter to check for latency on my new dsl line that I just got installed. With Ping Plotter going, and nothing else being done on my computer, I was averaging about 24 mS latency.
Then I ran a Speak Easy speed test to check my DSL speed. When the download ran, my latency jumped to about 40 mS.
But when the upload started, my latency jumped up to around 800 mS to 1000 mS. Sometimes higher.
Is it normal for the latency of a dsl line to jump up drastically when an upload starts? Or is my ISP having problems?
I also turned off my Ping Plotter and just did a continuous ping test to Yahoo.com. I then ran the speed test, and also noticed a considerable rise in ping times, especially when the upload speed portion of the test was running.
My ping times when using Ping Plotter or when doing continuous pings from the command line are great...around 24 mS...if I'm not trying to upload or download data...like on a speed test. But the ping times go to all heck when I do uploads, like on the speed test.
Thanks much for any help you can provide.
What is your contracted up and download bandwidth? What sort of QoS agreement do you have. Typically upload will be for less capacity as most users are contracted for ADSL.
You may find this article of interest, even though it refers to modems and ISDN.
http://rescomp.stanford.edu/~cheshire/rants/Latency.html
BobY777
09-12-2006, 09:19 AM
Hi M/Q:
My down BW throughput is around 2.78 Meg. Up is around 320 Kb throughput. Thats fine.
The problem is with latency when data flows...mainly upload data. If I ping my ISPs dns servers, with no data flowing I get around 46 mS. But on an upload, yesterday I got a little more then 1000 mS. Today it is around 490 mS. Seems to change on times of day, which suggests congestion.
My dsl light on the modem is quite stable, so sync seems real good. And BW is good. Problem is the darn latency.
I called a trouble ticket in today. I sure hope they can fix it. My Ping Plotter program showed 5 hops to their router. I was able to tell them the IP number of their device that made the latency jump like crazy.
That is your contracted Up and Down? It is pretty obvious that you are over reaching the upload side of the link. When there is no load up or down the only factor is the latency you observed. That does not change when you are sending one very small packet or many large packets, I call that actual latency. What comes into play is the timing. The ping packets have to get into a larger queue if you will, which gives you the appearance of increased latency, or what I call overall latency. You reduce the queue, the actual latency stays the same as that is a fixed entity, but the overall latency decreases. Read the link I sent you.
BobY777
09-12-2006, 08:46 PM
Earlier today, I did a continuous ping test on a broadband cable network, when I was doing a BW test.
The cable was 3.0 BW down and about 300 Kbs up.
This test had different results as compared to my dsl line. There was no noticable change in ping time when I ran the BW test and when I wasn't running the BW test on the cable.
This was Comcast Cable. They have got to be doing something different so that there is no change in ping times. Maybe they have a larger upstream pipe going to the Internet, as compared to my dsl ISP.
Something else pretty interesting...
I went to our office today to check for change in ping times on our fractional T1 line, when I ran the BW test. I ran the continuous ping and then ran the BW test. The increase in ping times when I ran the BW test, on our fractional T1 line was as bad as my dsl line. The fractional T1 line was from the same company that installed my dsl line yesterday.
It's interesting that the cable co. didn't have any change in ping times when the BW test was run as compared to when it wasn't running. But the fractional T1 line and my dsl line both had about the same increase in ping times (with the T1 and dsl being from the same ISP). Both circuits go back to the same CO. Interesting.