High Availability with Redundant switch fabrics

[Ryan17]

I'm curious peoples opinions on when a single redundant fabric switch (Nexus 7004 for arguments sake) is appropriate.

(Nexus 7Ks are expensive)

Examples of when a operational situations where a single switch would cause an outage?

any recommendations on how to approach assessing the cost/risk associated with these design decisions?

Thank you

Ryan

[Philip D'Ath]

I like dual supervisors systems like the 4507 and Nexus 7k.  I like the ability to do hitless software upgrades.  I find there tends to be more things you can do with zero (and I mean zero) disruption.

Dual supervisor systems also tend to have better "port to port" bandwidth, as it only has to go across a back plane, rather than channel groups of ports.

Dual chassis designs do cover you against entire chassis failures - but those are extremely rare.  Far more rare than the need to do software upgrades.

How could you get a chassis failure?  Perhaps environmental damage from an overhead water pipe erupting (but no one runs water pipes over their computer room do they?).  Perhaps a massive electrical failure causing the unit to destroy it self.

Like I say, whole chassis failures are pretty rare.

I tend to lean towards dual supervisor solutions that dual chassis solutions myself.

[devils_advocate]

Cost/Risk is entirely based on your business.

If you worked for a company selling shoes, 10 seconds of downtime might equate to £500.

If you worked for a bank, 10 seconds downtime might equate to millions.

The cost/risk is different in both cases.

10 seconds of downtime may be acceptable for the shoe company but it won't be for the bank.

If you invest in high end Cisco switches, my experience tells me that rarely have downtime that is not caused by a human action. 

A redundant Nexus Switch might cost the shoe company say £10k.
Over the next 10 years they may get 10 seconds of downtime just once or twice which is a cost to the business of £500-£1000.
In the case, the cost vs risk may weigh more towards the risk being work taking against having a resilience switch.

Its Apples and Oranges, Cost vs Risk is down to the individual business.