Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

Switches are routers are generally benchmarked in packets per second instead data transfer. In particular, Cisco benchmarks packets per second (PPS) at 64-Byte sized packets.

For example a switch is benchmarked at 50 mpps (Million Packets per Second) with 64-byte packets. Does anyone have any data on how this scales at different packet sizes?

For example would 128 byte packets go at 25 mpps? I imagine it will not maintain a strict ratio. Does anyone have any actual data on this?

share|improve this question
add comment

3 Answers

Dependence is more complicated(Cisco Nexus). This is for Nexus Cisco Switches: enter image description here enter image description here

enter image description here

share|improve this answer
    
Say what? (extra characters) –  EEAA Feb 28 '11 at 19:38
    
@ErikA, I think he means that it's not as simple as double the packet size = half the number of packets; though he picked a lousy example as that model's stats are somewhat close (and the logarithmic graphs are very hard to read). –  Chris S Feb 28 '11 at 21:41
add comment

Any good switch will provide two measures of performance at minimum:

  • Packets Per Second - This is usually measured with relatively small packets, as that's a worst case scenario. Packets of 64-bytes seems to be a relatively standard number.
  • Switch Fabric - This is the other end of the spectrum, with ideal packets (usually the MTU max), just how much data can the whole switch push. Any high end switch should have Full Switching Fabric (meaning every port can send and receive it's maximum data rate simultaneously).

Knowing how the switch will behaving in-between these two extremes may be useful if you know your particular data load, but if usually sufficient information to make a good estimate of performance in "common" applications. Also, it's common to rate the worst case latency, though this (and other measures) are less standardized than the above two.

To more directly answer the question. If a switch has full switching fabric, and can do 50 MPPS at 64-bytes, it's reasonable to assume it will do about 25MPPS @ 128B. The larger the packets the more efficient they will be, so the trade-off isn't linear, but will be somewhat close on good switches. The graphs in the other answer show this relationship.

share|improve this answer
add comment

It is a temporal thing. The larger the packet (more time on wire) the lower the PPS, and the higher the BPS.

share|improve this answer
    
While this is somewhat true, if the switch fabric can't handle the throughput load, it'll impact both PPS and BPS. It's not an explicitly simple trade-off. –  Chris S Feb 28 '11 at 20:04
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.