- This topic has 20 replies, 6 voices, and was last updated 1 year, 7 months ago by
.
-
Posts
-
Since you are saying layer 2, I interpret that to mean you are seeing Ethernet II frames with a bad CRC being dropped.
That should be the case on FABRIC’s (or any) ethernet switches with a store and forward architecture. Note they should confirm to IEEE 802.3 section 3.4 “The contents of invalid MAC frames shall not be passed to the LLC or MAC Control sublayers” meaning those frames would be dropped and not forwarded. This could open up an interesting research question for you, how would a new FCS scheme coexist or be incrementally deployed in a backwards-compatible manner?
Interestingly, the transponders FABRIC uses on its optical network do not process ethernet frames at the traditional LLC or MAC layer in this manner and as such they should happily and blindly forward corrupted frames.
-
This reply was modified 1 year, 7 months ago by
. Reason: poor grammar attributed to insufficient caffeination
Arash,
Have you tried to do this with a physical machines outside of FABRIC. It would be interesting to see if any switches forward frames with errors.
Some switches based on a cut-through architecture (currently, I think these are somewhat rare vs store and forward architecture?) do pass some frames with errors. On those architectures typically only the first 100-ish bytes are read and the packet is already being sent out the egress port before it’s been completely read in. As such there isn’t time to evaluate the FCS and discard the frame as it’s already been sent.
Arash, can you clarify that by checksum you mean ip header or Ethernet frame?
For IP header you can try L2PTP. For Ethernet, as Dale pointed out, you may have to consider optical layer solutions.
-
This reply was modified 1 year, 7 months ago by
- You must be logged in to reply to this topic.