Cascade Limit of Gigabit Switches

[jeffland_98]

Does anyone know of a specification or even a rule of thumb for the number of gigabit switches (2960G-48TC specifically)that can be cascaded together before causing significant network performance problems? This would be a data only network, around 100 vlans, most large data movement done on a separate SAN.

What would be the limiting factor?

[dgahm]

How much traffic do you expect on the interswitch links? If each switch is sending 100mb/s towards the core, you probably don't want to exceed 5-7 switches or you will risk overloading the 1st link.

The cumulative latency would not really impact performance unless you have a very large number of cascaded switches. Switches use store and forward, meaning a whole frame is received before sending it on, but with gigabit this is really negligible at .012 millisecond per switch (1500 byte frame).

My biggest concern would be fault tolerance, and limiting the impact of a single switch or link failure. Can you afford to lose all the cascaded switches if the 1st switch goes down?

When in doubt do some research on best practices. Here is a link to the campus design SRND:

http://www.cisco.com/application/pdf/en/us/guest/netsol/ns432/c649/cdccont_0900aecd801a8a2d.pdf

Please rate helpful posts.

[jeffland_98]

We don't have baseline for the link traffic yet, still in design. Unlikely to be anywhere near 100 Mbps per switch. The customer was really wondering what his upper limit might be. Right now there are four switches with redundant links.

Thanks for the link to SRND. I hadn't seen that before.