Hi All,
I'd like to understand the full end to end latency introduced by an Arista 7150s. Per the 7150 Data Sheet its roughly 350 nanoseconds on a 24port in cut-through mode.
Cut Through Switching ( Only works under specific scenarios )
Technically frames can be switched after receiving the first 6 bytes which should contain the destination MAC, however most of the switches (including Arista/Nexus) will wait for the first 54 bytes to be received before making the forwarding decision. Errors are usually detected within these bytes. Source - 40GB not always faster then 10GB
Store-and-Forward Switching
The entire packet is stored in a buffer and then transmitted out, depending on MTU size this can slow the packet down a lot.
Description | Latency |
---|---|
Read first 54 bytes@10GB | 43.2ns |
Arista 7150 | 325ns |
Serialisation back to wire | Anything here? |
Is the total 393.2 nanseconds or am i missing anything ?
I'd like to understand what latency if any is introduced going back to the wire?
How will interface speed affect this? ingress will be slower to read the first 54 bytes if less than 10GB
Thanks!
No comments:
Post a Comment