How to evaluate Ethernet performance?

Question

I need to think about performance limits of 100 Mbps Ethernet (including scenarios with ~ 100 endpoints on the same subnet), and I wonder how best to evaluate network bandwidth. Are there any rules of thumb?

The reason I ask is because I am working on some envelope level performance calculations about performance limitations, so I don't have to be incredibly accurate. I just did not do this exercise before and was hoping to get some idea from those who have. Mark Brackett's answer (starting at 1/26) matches what I'm looking for.

Answer 1

If you use switches (and to be honest who is not these days), then I found 80% of the capacity of a reasonable estimate. This is typically around 90% due to TCP overhead, but 80% allow for random retransmissions.

If this is a single collision domain (hubs), then you will probably be about 30% with moderate activity at these 100 nodes. But that would be pretty variable based on the generated traffic. And anyone who puts 100 nodes in one CD these days will surely be taken off - so I don’t think that you actually came across these IRLs.

: , , . , , .

Answer 2

WANem

WANem - , /, / .

, , , . sourceforge.

: WANem -

Answer 3

Opnet . - Opnet IT Guru Academic edition. , opnet .

Answer 4

100 . ( ), . Fast Ethernet (100 /) 10 () . . . . ( ).

Answer 5

While working on the title of his post, “How Can I Evaluate Ethernet Performance”, see this wiki link; http://en.wikipedia.org/wiki/Ethernet_frame#Maximum_throughput