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.

New Member

Redundant Supervisors in a Dual-Core Catalyst 6500 network

Read the PDF doc called:

Designing a Campus Network for High Availability.

http://www.cisco.com/application/pdf/en/us/guest/netsol/ns432/c649/cdccont_0900aecd801a8a2d.pdf

It essentially recommends that dual supervisors only be used in SPOF points - not in dual-distribution or dual-core switches.

The rationale is that a layer three routing protocol convergence (EIGRP for instance) only takes 60-200ms, but a supervisor failover using NSF and SSO takes up to 3 seconds.

In practice, I almost always see dual supervisors recommended in the core (and that's what I've recommended as well).

I can see keeping dual supervisors in the core because the customer may place a higher value on redundant hardware capability than an up to three second convergence, but I'd like to hear other rationales why dual supervisors should be used in a redundant core given the definite speed advantage of routing protco convergence over convergence following NSF/SSO failover.

Thanks,

3 REPLIES
Silver

Re: Redundant Supervisors in a Dual-Core Catalyst 6500 network

Supervisory redundancy does not always provided take over in case of active sup's failure. In most cases it will do if the failure is obvious. However, very occasionally the subtle failure can occur like failure in CAFE, forwarding asic, which can be detected only with off-line diagnostic, which switch cannot detect during run-time but the failure is severe enough to screw up the whole switch and still prevent take-over.

Re: Redundant Supervisors in a Dual-Core Catalyst 6500 network

The added complexity of a dual supervisor is not always desirable and in a design, we also have to look at the cost. Both these arguments are not favouring the use of dual SUP's in a configuration that is already redundant.

I would say that a dual enclosure with single SUP's is the best redundancy solution available.

Regards,

Leo

Silver

Re: Redundant Supervisors in a Dual-Core Catalyst 6500 network

In a completely clean textbook design where the cores are mirrors of each other and only provide transit between distribuition layers, NO, dual sups are not needed and will only slow convergence as the document indicates.

However, in some live environments this is not always the case, where the textbook does not reflect reality. One core switch can have a configuration or connection that the other core cannot backup. As much as people try to keep their cores clean and symetrical, a fast growing and underfunded network can quickly get messy with one-offs and bandaids that get forgotten about a few months later. In situations like these, having dual sups in the core provides the peace of mind and assurance that 3 seconds can buy (if you can afford it).

Please rate this post.

Thanks,

Brad

181
Views
4
Helpful
3
Replies