Cisco Support Community
Showing results for 
Search instead for 
Did you mean: 

Welcome to Cisco Support Community. We would love to have your feedback.

For an introduction to the new site, click here. If you'd prefer to explore, try our test area to get started. And see here for current known issues.

New Member

Non-Blocking Architecture

Cisco cat 2950G-48 is a Non-Blocking switch.But if I cascade TWO 2950G-48 switches with a gigabit fiber cable , will the architecture between 2 switches still remain non-blocking or would it be a blocking architecture.???Cisco site or DOCs do not mention this specific concern.Any help in this regard would be appreciated.

New Member

Re: Non-Blocking Architecture

Here's my thoughts: The specs for the 2950G-48 say it has a 16.6 GB per second matrix, but a 6.8 GB per second MAXIMUM forwarding bandwidth. ( I assume thats because each port can operate full duplex) If all 48 ports had 100 MB per second going thru them, that would be roughly 4.8 GB, not quite the full 6.8 forwarding rate. But as you cascade switches, you're doubling the amount of the ports in the switching matrix, however the backplane bandwidth (switching fabric & forwarding rate ) remains the same. In addition, you are adding a choke point , that is, the Gigabit Fiber link, so the "switching fabric" is now way less than max of any one switch. (Data on port 1 on switch 1 might have to be sent to port 48 on switch 2 and traverse the fiber...)

So, I think once you have more than 1 switch it's no longer guaranteed to be non-blocking.

Cisco, any thoughts on this?

New Member

Re: Non-Blocking Architecture


The definition of a non-blocking switch is that it can pass the sum of the data offered on ingress to the egress ports without packet loss. So, if I have 48 ports sending full duplex traffic between them ie port 1-port2, port 3-port 4, then the 2950G is non-blocking.

The question you ask is based upon the fact that you have a single GE link between switches. This doesn't change the non-blocking nature of the switches, but introduces a network bottleneck as traffic aggregates on a single port betwen the two switches. This isn't a switch issue but a fact of life that traffic aggregates within networks.

So, it would still be correct to say the switch fabric is non-blocking but the switched network is inherently blocking, hence why you need robust QOS ;-)