Criticism of the Catalyst 4500

From Internetworkpro

This page or section contains architectural information Please note that this page or section contains information on the architecture of a product or software and may be inaccurate, confidential or unsupported by the vendor. Any stated capabilities or future roadmap information may be dependent on software and hardware modules.

However, the device also has a number of limitations that quickly cause problems unless understood and catered for. This article attempts to articulate them.

The biggest mistake anyone makes with the 4500 is to assume that it can do a 6500's job. The moment this assumption is made, scalablity becomes a large factor. Cisco never position the 4500 in a datacentre environment, nor do they promote its use in anything but low density aggregation scenarios.

[edit] Architecture

To understand any true objection towards the 4500, one must first understand the architecture. The below shows how the supervisor connects in to each of the line cards:

The basic principal is a supervisor with 6x1Gbps of traces connecting each line card. These traces are then shared amongst the ports. For example, if we have a 48 port 10/100/1000 line card (such as the 4548) then the allocation may look like this:

The line cards in themselves have no queuing ability, nor do they possess the ability to buffer. On ingress, a port merely copies its ingress data on to its local trace. As shown above, should a group of ports receive more than 1Gbps traffic, any excess will be instantly dropped.

A further problem to this design, aside from the blocking architecture, is the reliance on the supervisor for all intelligence. Everything including queuing, buffering, routing, switching, policing and more is done on the supervisor. Engineering challenges make it difficult to extend or add functionality due to lack of physical space.

When compared to the 3750-E series switches and also the Catalyst 6500, the 4500 comes up as a very poor choice.