I'm working with a client who has giant, campus-wde VLANs extended to every switch in creation. They have multiple /20 networks with possibly over 1000 PC clients and servers per vlan.
This is an old-style network and they have been dragging their feet on doing an IP re-addressing.
It is not their only problem, but pretty much every issue that gets brought up to support or consultings gets written off as due to large broadcast domains. This impedes getting other problems solved cause no one will look beyond the obvious "big vlans=your problem" line. I have already solved other issues that have been affecting them for TWO years because Cisco Certified professionals kept pointing fingers solely at broadcast domain issues. I encountered this with Nortel profesionals as well.
I'm looking at high CPU utilization on 4006s. The only two 4006s that don't have ultra-high CPU are running IOS 12.x instead of 7.1(2). I haven't dealt with 4006s running cat vs native IOS.
Do you see significant CPU level variances runngin CAT vs Nativve IOS?
How much CPU increase do people see when you add inline power modules? 1% 10% 20%? Does it vary based upon how many slots have inline, or how many inline ports?
With respect to the large broadcast domains. If I selectively block a gianormous VLAN from the trunk will the CPU hit from that VLAN be alleviated? I'm thinking of asking for a blackout window and removing the suspect VLANs and recording the CPU variances as proof to management of the exact damage each VLAN causes. The 4006s run at 50-60% utilization with near 0 traffic in the middle of the night.
I have read that CPU utilization on certain 4000/3500/2900 don't reflect CPU consumed by traffic, and I have quite a few where the monthly avg if 49% and the monthly peak is 50%. However the 4006s peak at 99% and the CLI becomes unresponsive regularly. We also have buffering problems--which I am sure bigVLANs is contributing to in a big way.
