QoS on mismatched links

[JACQUES DU PLESSIS]

Hi Guys,

We have a 500m link at HQ and one of the branch sites have a 2m link, both connecting to a ME network. We run both voice and data to the site, however they are complaining about quality. QoS markings are OK, but we can see that they start complaining when their link gets to about 1.8m (SP wants us to reserve 0.2m). So obviously the HQ link is sending at a much higher rate and then the branch site gets congested. I have a question about how we should be doing this. Currently I have this at HQ:

policy-map egress

class voice

police 80m 8m conform-action transmit exceed-action drop (there are other sites as well so we need about 80m)

class branch-in-question

police 1.7m 0.1m conform-action transmit exceed-action drop (this is for their data traffic).

The problem here is that voice can go to for example 1m and data to 1.5m, which over-utilizes the link and traffic drops.

Does anyone have any alternatives for this. We are using a 4500 at HQ and 3750x at the branch (with srr-queue bandwidth limit to get it to 1.8).

I know I can create a voice class for each site and rate limit exactly (for example voice 0.6m and data 1.2m), but that sort of waists bandwidth in terms of the data class.

Thanks!

Jacques

[Joseph W. Doherty]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

If you're going to use policers, which is far from ideal, you can police none VoIP trying to "reserve" sufficient bandwidth for VoIP.

A better approach, on the 3750X, would be to run the egress interface at 10 Mbps, and then SRR bandwidth limit for 20%.  Then prioritize VoIP traffic.

For a better approach on the HQ side, you need something more WAN oriented than a 4500.  The lease expensive approach, if you have available copper ports on the 4500, would be to run a WAN router in parallel to the 4500.  For 2 Mbps, even an 800 series would do.  With a WAN router, you could shape for 2 Mbps and priortize VoIP.  You could also FQ other traffic.  (You could do some similar with some small switches, but they won't support FQ.)

[JACQUES DU PLESSIS]

Hi, yes that is what we have done for the 3750x. And I agree about using policers, it is just that I can only do shaping on the tx-queues on the 4500 and that will not work for this situation. Unfortunately the speed at HQ is 500m, not 2m, so I am not sure a small router will work (and this could actually be actual throughput since we replicate to DR over the same link and have other branches on it as well).

But if there is no better (software) configuration option than we have currently and hardware is the only option, I guess it is the only way. Getting the budget will be the biggest challenge there, although I might be underestimating the smaller routers?

Thanks

Jacques

[Joseph W. Doherty]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

A (really) small router, suitable for 2 Mbps, cannot deal with your 500 Mbps.  That's why I noted it should be installed in parallel (not in line).

Basically, only egress traffic, to the branch with the 2 Mbps link, would transit the small router.  (How you would wire this up, and configure it, migth be very "Rube Goldberg", but it would minimize the CAPEX.)

Of course, an ISR that can handle the 500 Mbps could be placed in line, but the CAPEX shoots way, way up.