Cisco Support Community
cancel
Showing results for 
Search instead for 
Did you mean: 
Announcements

Welcome to Cisco Support Community. We would love to have your feedback.

For an introduction to the new site, click here. If you'd prefer to explore, try our test area to get started. And see here for current known issues.

Silver

QoS design question

Hi, I have a design question regarding QoS when having a subscriber over a L2 MAN.

Here are some details:

CCM version 4.0(2)sr2a

Publisher + Subscriber at location 1

Subscriber + IPCC at location 2

Connection between location, fully layer 2 transparant Ethernet MAN, bandwidth 10Mbps. Both location have two 2851 routers using split SRTS and performing vlan routing. Also, both sites have Catalyst 4000/4500 series switches as core switch, 3500/2550 series access switches. The MAN is connected to the switches in a separated vlan.

Currently running smoothly without QoS, but need to implement as bandwidth consumption will increase.

Want to use the following classes:

ICCS/SQL replication (1.5Mbps guaranteed between subscriber and publisher to be in a supported configuration)

Call signaling, 64kbps guaranteed

RTP streams, 1,5Mbps for appr. 20 simultanious G.711 calls between locations.

Internet related traffic, 2Mbps

Business related traffic, all the rest

Now, I do not need the access-list definitions, as I allready have those figured out. What I need to know is what is the best way to implement. Any advice is welcome ;-)

Thanks in advance,

Leo

2 REPLIES

Re: QoS design question

Hi there Leo,

The best thing to do is mark the traffic as soon as pissible from the source with a value, EF, AF21, AF31, etc. Then throughout the rest of the infrastructure you can use class maps and policy maps to define how each class of traffic is handled. The reason you would do it this way is because if you had access-list specifying traffic and you wanted to add some traffic to a class, then you would have to add it all through the infrastructure, whereas, if you were just adding a valu, then you would not have to change anything else.

For the traffic you then create a policy map specifying how much you want each class to have, ie., 1.5 mb for SQL, 64 for signalling, etc.

For the RTP streams, if you have enough bandwith to cater for all of the calls, then youjust give it priority, if not, then look at using RSVP and limiting the number of flows:-

http://www.cisco.com/univercd/cc/td/doc/product/software/ios113ed/113ed_cr/np1_c/1crsvp.htm

Regards,

LH

Please rate all posts

Re: QoS design question

Hi Leo,

you need to implement queueing on every interface, where an overload situation could occur.

CBWFQ can be used in the Cisco 2800 routers. They offer the greatest flexibility with respect to QoS compared to your switches.

As I understand your L2 MAN connection is between two of your switches. In switches generally QoS is much more hardware dependant than in single CPU routers, where basically everything is done in software.

So the big question is: which switches have the MAN connection and which QoS capabilities do they offer?

As there are many different 3500 it is hard to tell which exact model you have and what can be done. The other is a Catalyst 2950?

Generally you could use COS to classify, but there might be no possibility to enforce your suggested data rates due to a lack of QoS capabilities.

So for a more precise answer we would need exact switch models/modules and software release.

Regards, Martin

110
Views
0
Helpful
2
Replies