0

This is a followup to a previous question I posted: Debugging Cisco 2621 Routers

I've discovered that my router starts becoming unresponsive during offsite uploads of large datafiles. I have two questions:

1) Is there a way to throttle the connection between my cluster and the offsite server, so the spikes don't overwhelm the router?

2) Is this the best solution to the problem? I would rather have a general solution - I'm a little disappointed that the router becomes unresponsive during these spikes, I'd much rather it know to throttle connections which cause it to hang on its own.

1
  • Isn't the 2621 EOL now? I'm not even certain that you can keep them up to date on IOS anymore. Jul 23, 2009 at 14:07

3 Answers 3

3

You can rate-limit traffic that matches a certain ACL. You can create an ACL that will match the traffic by matching on source and destination addresses, ports etc. then configure something like:

...
class-map match-all off_site
 match access-group 1
!
!
policy-map limit_off_site
 class off_site
   police rate 6000000 bps
!
interface Ethernet0/0
 service-policy input limit_off_site
!
1

The correct answer is likely the following:

Scale your router to the proper size for the connection(s) that are going through it.

For instance, you don't throw a 26xx (or 28xx) series Cisco at a GigE connection.

Since you don't mention the volume of traffic you need to handle, it's hard to give you a good recommendation for router size to use.

Throttling bandwidth, while a workable solution, does not let you take full advantage of your infrastructure.

2
  • Thanks for the response - the traffic which is hanging the router is anything above 6 mbps, but we see spikes up to 24 mpbs. What is the appropriate sized router for this kid of traffic? We may have to hold off on upgrading hardware due to budgetary constraints - until then, do you have any recommendations on throttling bandwidth? We very rarely see any 'real traffic' at 6 mpbs, at this point we've only correlated it with our offsite backups (which are large, btwn. 1-10 G).
    – Harry
    Jun 30, 2009 at 22:17
  • You could try switching to a 2850-series. They have a fair bit more oomph than the xx21's, whether it be a 2650 or 2850. Jul 4, 2009 at 21:11
0

I wouldn't expect 6+ Mbps to peg a 2621, otherwise you could easily kill one with just two WIC-2T cards. You noted a correlation to CPU usage in the other post, so you might want to look at the following:

  1. Do you have CEF enabled? Look for "ip cef" in your config.

  2. Do you have any ACLs defined on the affected interfaces? They will cause IOS to process switch some packets, meaning the CPU has to deal with them (though CEF mitigates this to an extent).

  3. Are you running a crypto feature set and terminating encrypted tunnels on the router? The 2621 has no hardware crypto acceleration, so the CPU has to do all the transforms.

You must log in to answer this question.