Trying to identify single point of failure on 4507

Unanswered Question
Sep 28th, 2010

Hello!

I am trying to figure out if by having dual power supplies and dual supervisor engines on my 4507, do I still have a single point of failure on this switch?

I guess a whole blade could die, but I am trying to identify another componant that if it failed, would be catostrophic to anything connected to that switch.

Thanks!

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (2 ratings)
Loading.
Collin Clark Tue, 09/28/2010 - 12:35

To add to Jason's post. A single chassis will almost always have a single point of failure somewhere. The only one I know of that where everything is N+1 is the Nexus 7000 switch (maybe the ASR router). But as Jason pointed out, the single points of failures in a chassis are unlikely (eg line card, chassis bus). What we do is put in redundant switches where 99.999% connectivity is required. Where that isn't required we just bought spares and swapped in case of a failure.

pener1963 Tue, 09/28/2010 - 12:42

Thanks for the great informartion!

In my case the 4507 represents the core switch of my network. I have it covered with 24x7 SmartNet. I am looking for this to last me another year and a half. Then I would like to go with a single or dual 6500s. I have about 250 users right now, but we are growing quickly.

Perhaps buying a spare chassis is an alernative, and will let me sleep at night.

Again many thanks...

Collin Clark Tue, 09/28/2010 - 12:44

If it were me, I would look at the 6500 running VSS. Imagine two 6500's running as one! No single points of failure as a virtualized system!

pener1963 Tue, 09/28/2010 - 12:49

WOW! I have never heard of VSS. I will definately investigate. Sounds complicated but worth a try!

Collin Clark Tue, 09/28/2010 - 12:50

It's easy, about six commands. Just check the configuration guide if you decide to go that way.

Jason Masker Tue, 09/28/2010 - 13:04

I was referring to the reduction in spanning tree complexity that can be accomplished with virtual port channels alone.

Jason Masker Tue, 09/28/2010 - 12:51

Yes, VSS can simplify things quite a bit by eliminating much of the reliance on spanning tree, etc. The Nexus equipment does much of the same. If buying new today, definitely investigate these options.

Collin Clark Tue, 09/28/2010 - 12:55

Actually the NEXUS does not do any multiple chassis virtualization, but it will support TRILL in the future which will address STP issues.

Jason Masker Tue, 09/28/2010 - 12:48

I would agree. A chassis is cheap, but the hours of sleep are precious.

We have run a dual core design on the 6500 platform for a number of years now and it has worked very nice for us. It is a very nice configuration that offers all the redundancy with no equipment sitting on a shelf. We're looking at Nexus, but even then I tend to prefer a design with two 7ks to a design that relies only on the N+1 redundancy of the Nexus.

In our closets we are running 4500s with no smartnet, taking advantage of the lifetime warranty. For those, we keep a spare of everything on the shelf.

Jason Masker Tue, 09/28/2010 - 12:29

You are very well protected in this configuration. The most likely thing that could happen that would be a significant impact would be a line card failure due to the number of ports per line card. This can happen, but it does not happen all that often. Cisco actually promotes that the mean time between failures for their 4500 line cards is up to 6 times that of supervisors, fans, power supplies, etc. There is not much to the chassis but technically it must be considered a point of failure. The chassis is probably the least likely to fail. Fans and power supplies can and will fail, but they are redundant in the 4500.

It all depends on how comfortable you are with the risk and the trade-off between expensive smartnet versus spare equipment on the shelf, etc. I have ordered a spare chassis before for equipment that was critical enough. All that said, the 4500 does a really really good job of providing good internal redundancy for applications like access closets where it doesn't make sense to have completely redundant equipment.

If you contact a channel partner and sign a non-disclosure agreement, Cisco may be willing to share some of their actual mean time between failure statistics to help with your decision making process.

Actions

This Discussion