This chapter covers the following topics:
Layer 3 designs
Catalyst Supervisor and switch fabric redundancy
One goal of every network design is to provide users as much bandwidth as possible, as often as possible. Sounds easy, right? Network administrators often use terms such as highly available, redundant, scalable, and resilient to describe their network designs. To make good on the implied promises of these terms, administrators should adopt certain best practices.
Many long-time Cisco customers respond to the latest switching design best practices by saying "Just a few years ago I was told to switch where possible and route where I must, now I am told to route where possible and switch only where I must. What is the deal?" A best practice is only a best practice until your requirements or a technology change sufficiently to invalidate the best practice in favor of a new one. The term best practice does not indicate that only one way exists to accomplish a task, it only indicates that in the majority of circumstances experience demonstrates a certain solution to be a success. In the case of LAN switching, customers were told for years to implement Layer 2-only solutions to achieve high-speed Ethernet data transfer rates. With almost universal support of Transmission Control Protocol/Internet Protocol (TCP/IP) for applications and much improved hardware capabilities, these same customers are now being advised to implement Layer 3/4 solutions in those same networks.
The best network design is the one that meets the needs of its users. No one "correct" switched design exists, only proven design principles that should be incorporated where possible. Designs can differ based on a number of real-world factors including budgets, available existing hardware, application requirements, and implementation timelines. The key is to understand and weigh the pros and cons of each design principle against the overall goals for the design.