dual 4506 vs. single 4507

[dave_miller]

Does anyone have any overiding opinions on whether it is better to utilise a single C4507 with dual Sup IV's in HA and cold spare chassis on a VoIP enabled site or to go with two C4506's each with single Sup IV's and set up in high availability between the two of them.

This is to form the core of the network - traffic then is out to edge 4506's or stacked 3750's (any opinions on that as well ?)

[Kevin Dorrell]

If I understand you right, in the first case you would have a single HA core chassis and a single uplink from each access switch, and in the second case you would have two 4506's, and two uplinks from each access layer chassis.

I would go for the second option, without a doubt. It would provide redundancy for your uplinks as well as for the chassis. In the event of a failure, the switchover would be faster, especially if you used fast Spanning Tree or uplinkfast.

OK, the chassis is usually pretty reliable on the 4500 series, but nevertheless, imagine the hassle of switching over to you cold standby chassis.

But then I've never been a fan of complicated failover mechanisms such as between the two supervisors. IMHO, failover mechanisms should be as simple as possible. I have seen failovers that were less reliable than the equipment they were supposed to protect. We had a RAID controller once that was less reliable than the underlying disks.

Just a tip: make sure your power supplies are big enough to operate in fully redundant mode, and feed each one from a different supply circuit.

Kevin Dorrell

Luxembourg

[dave_miller]

You understand perfectly .. and you've hit on our possible thinking also, full resiliency vs. simplicity (I guess cost is also a factor, but a small one).

In the dual 4506 setup I assume that there is no loss of connectivity if one of the SupIV's were to fail - ie. we would not loose any 'live' VoIP calls to the edge switches .. or do we have to rely on the ST convergence and loose connectivity for a few seconds. Or (as I've not setup like this before) do they both work in tandem and load balance giving no visible outage during a failuire ..

Thanks for your reply .. much appreciated ..

[jroyster]

Things are just so much more reliable with two core switches in my opinion.

You are correct...there would be a loss of connectivity for a few seconds that would not drop the call. There are lots of ways to accomplish this...the most control is to use layer3 links to the access or distribution switches. It allows you to tune your routing protocols to whatever failover/convergence you like.

If that isn't feasible then you can use HSRP and tune the timers way down to achieve fast failover if a supervisor/gateway fails.

Not to mention any kind of maintenance/changes are much less impactful. I really can't see any pro to a single switch except ease of initial configuration and cost.

hope this helps