determine cause of latency

Unanswered Question
Oct 2nd, 2007

newbie question time...

I see high latency on ping replies to Internet sites - varies between 17ms and 500ms+. T1 vendor has run tests and says everything's fine, where do I start? How can I check the processor usage on my router?

thanks

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 3.7 (3 ratings)
Loading.
Paolo Bevilacqua Tue, 10/02/2007 - 12:11

Hi,


is not a problem with your router, it's normal on the internet and is due to queuing at circuits that may not be under your ISP control.

On the other hand if the delay get accumulated on your ISP circuits, it's his responsability to contain it to reasonable level. Since this in practice involves costly upgrades, ISP may then leave things as they are.


Hope this helps, please rate post if it does!

rblankers Tue, 10/02/2007 - 12:22

thanks for the reply. the latency is much lower at night when nobody is here using the t1, so I figured it may be related to the volume of traffic we're sending during business hours.

Paolo Bevilacqua Tue, 10/02/2007 - 13:21

Quite possible, do show interface to check or even better run something like PRTG where you ca observer the traffic in real time and over time.

Joseph W. Doherty Tue, 10/02/2007 - 18:44

On Cisco routers, you normally can see processor usage by using "show proc cpu" command. However, more likely variable latency is caused by transient congestion on your T-1.


If you do a "show interface", you'll see the load on the interface, both in and out.


If using a Cisco router, you might check that fair-queue is enabled on your T-1 interface. Also ask whether your ISP is using it or FIFO on their side of the link. If they are using FIFO, ask if they would change to fair-queue.

rblankers Wed, 10/03/2007 - 14:02

proc usage on the router (cisco 2600) appears ok, 25% or less. All interfaces are using FIFO. I ran ping tests overnight using PRTG and PingPlotter - the traffic spikes significantly during the day (8am-4:30pm) but stays below 50ms during off peak. We don't have a baseline to refer to, but is this normal? thanks again guys.

Joseph W. Doherty Wed, 10/03/2007 - 15:13

Your nightly 50 ms results sound about right for general Internet access and could be considered your baseline.


Activation of WFQ may help since it keeps one flow from hogging all the bandwidth. Again, for full effect, it needs to be active on both ends feeding your link.

rblankers Thu, 10/04/2007 - 12:10

so, I was apparently looking in the wrong place on the interface stats... bandwidth is maxed out - 2 bound T1's and input is almost constantly pegged at 3Mbps. Now I'll be chasing down those connections that are eating it up. Thanks for your help guys.

Actions

This Discussion