Big Switch, Little Switch
Having deliberated on it for a fair while, finally the decision is made to adopt a couple of big cats and move towards a much more centralised blade based switching infrastructure.
The original plan was to operate on a similar principle to RAID but with switches - i.e. lots of little switches, so minimising the effect of an individual failiure and making them small enough and cheap enough to be able to simply swap out on fail.
This has worked fine for years, utilising cisco 3550-48 switches for distribution, cisco 3508 aggregation switches (pictures), and individual 2950-Ts for each rack, however we are now hitting performance issues with the load on the 35xx Layer3 switches and VTP problems with the 29xx Layer 2 switches leading to a rethink.
The biggest problem facing a move towards the large blade based 55xx/65xx switches is cost - buying 2 (must have a spare!) 7.5k backplanes and enough £500 each 10/100 5225R blades is a lot harder to swallow than simply adding another £350 ws-2950 as more servers are added.
And to really take full opportunity of the multiple Cisco Catalyst 5513’s we now have, all the racks really should be structured-cabled back to a central point…
So cancel everything, for the next month or so another 14 kilometers of LS0H cable need to be run under the floor, 24 lengths to each of 32 racks…
