Advantages of 9600 Dual Supervisor

[lcaruso]

With two 9600 chassis both using a single Sup compared to using dual Sup, what are the advantages?

I mean with a single Sup, if one Sup unit fails, the other chassis takes over, correct?

Is the forklift upgrade/reboot time the main advantage then where you upgrade the standby Sup on both chassis then cutover/swap roles for only a few seconds of downtime?

Two Sups is quite expensive. There must be some compelling reasons to justify costs. 

Thanks in advance. 

[MHM Cisco World]

not only upgrade downtime BUT the NSF is important for Core SW

[Joseph W. Doherty · Accepted solution]

The principle advantage of have dual sups in a VSS/StackWise pair is with single sups (per chassis) if you lose a sup you lose half your capacity until a failed sup is replaced.

"Two Sups is quite expensive."

Yup, but there is a "third" option.  Have one chassis with single sup, other with dual sups.  If sup fails in latter, no outage.  If sup fails in former, move standby sup from dual sup chassis to replaced failed sup in single sup chassis (i.e. failure time in minutes).

"There must be some compelling reasons to justify costs."

True, but same question and answers might apply to why you chose to use a 9606 pair.  For example, a 9410 would provide the same number of line card slots, i.e. possible less expensive than a pair of 9606s.  So, if 9410 less expensive, I suspect you had reasons to justify the extra cost for a 9606 pair.

[lcaruso]

Thanks much for your post. One reason I am looking at the 9600 was the 9400 only supports 10G line cards unless I missed something. Another reason is the client needs 10 year lifetime for their investment and the 9400 line is two years older than the 9600 line. Your response was perfectly informative, and I was seeking assurance I had not missed anything else for justification. Thanks again. 

[Joseph W. Doherty]

"One reason I am looking at the 9600 was the 9400 only supports 10G line cards unless I missed something."

That appears to be the case.  Of course, with 480 Gbps per slot, no reason Cisco couldn't provide 25, 40 or even 100 Gbps ports.  (BTW, not pushing the 94xx over the 96xx, the latter appears to have much more bandwidth capacity [also BTW, I understood the 9400 series designed to replace the 4500 series while the 9600 series designed to replace the 6{5/8}00 series.])

"Another reason is the client needs 10 year lifetime for their investment and the 9400 line is two years older than the 9600 line."

I haven't looked, but beyond just being two years older, has Cisco provided a guaranteed service life?  Even if Cisco has, there's a difference in guaranteeing replacement parts and/or software bug fixes, vs. new hardware enhancements and/or IOS innovations.  If your customer is just looking for parts and/or bug fixes, then just going by Cisco lifecycle commitments is a pretty safe approach.

[Leo Laohoo]

I can no longer recall how many 4500, 6500 & or 6807/6880 I have "touched". 

But I can never recall RMAing any supervisor card. 

However, this was a time when Cisco's quality control was never in doubt.  

[lcaruso]

Good point, things have changed. For this client, being able to do upgrades without the ISSU concern, is a strong reason for the added expense. Still not clear on whether or not the following described elsewhere is ISSU or some other procedure: one can forklift upgrade the standby sups on chassis with no service degradation or interruption and then switchover to them in seconds of downtime and then do the other pair. Is there a way to manually do this without ISSU?

[Joseph W. Doherty]

Yup, on the 4500s and 6500s, over the years, I vaguely recall maybe one sup failure.  On the other hand, having worked on 6513s hosting 11 96 port line cards (yup over a thousand ports), with VoIP phones and PCs in a call center, a sup failure taking out all those ports was not considered an acceptable risk, regardless of MTBF calculations, etc.  (Heck, MTBF on the chassis, itself, was still a concern.  I.e. The forgoing wasn't the "best" redundancy option, technically, but was a considered reasonable for its cost.  Later on, that business was probably one of the few that tried the 6500 IA approach.)

[MHM Cisco World]

for use one chassis with single Sup and other with dual Sup, that not work. 
you need Quad Sup 

for the point of single Sup if it failed I still check the sinlge and multi homed to override this problem and as @Joseph W. Doherty  mention if the Sup is down all Chassis is not forward traffic, even if Sup is responsible of control plane Still the data plane need Sup Why ? because the Sup contain the fabric that internal interconnect all line card. 

[Joseph W. Doherty]

"for use one chassis with single Sup and other with dual Sup, that not work. 
you need Quad Sup"

Are you sure?

Consider, if you start with quad sups, and one fails (particularly the standby in the chassis), does 9606 pair continue to function?  (Answer, of course is yes, i.e. quad setup will run with just 3 sups, or even with a failed sup in each chassis.)

Possibly, you might not be able to initially "setup" using just 3 sups, and if not, you can still have the 3rd sup sitting nearby for immediate replacement, then you only lose out if the one chassis, that would have had the 3rd sup in it fail, you lose the auto switchover.  (Another option for replacement, might be some form of expediated replacement service guarantee, e.g. same day or a few hours.)

BTW, I recall in the 6500 series, with VSS supporting quad sups, Cisco "frowned" installing only 3, although technically, I believe there was no limitation actually doing that, just, of course, you lose sup redundancy in the chassis with only a single sup.

Assuming you do have 3 sups, and assuming you can start with such, why even bother with installing the 3rd sup?  One reason, as already described, on chassis would have the 2nd sup in it for additional redundancy (for just that chassis), and it further (and on-going) confirms the sup is functional.

[MHM Cisco World]

the Quad-Supervisor with ONE ICS 

[Joseph W. Doherty]

Sorry, it's (to me) unclear what's the point you're making.

Although your posting show 3 sups, I believe the point is if the primary for the pair's master and active sup fails, its standby sup takes over but master for the pair goes first to the secondary chassis and then reverts back.

[timoffet]

This is an old thread but it's important to know that both single/dual and dual/dual are valid designs. 

Now having a single supervisor in one chassis is not very useful if that is the one that fails.  Maybe that was the point that was trying to be made.

In a quad SUP SWV configuration, if a supervisor fails in a chassis, the SWV will fail over to the standby.  Half the combined forwarding capacity will be available until the warm standby (AKA ICS - In Chassis Standby) boots.  That is about 5 min.

Then full capacity is available again.  In mission critical environments where full capacity is a key concern, QUAD SUP makes total sense.  It is a very subjective decision though.

Tim