Subnet manager and multicast RDMA
Replacing infiniband-general mail list (which I wasn't aware even existed, let alone a subscriber to) with linux-rdma.
Quoted 10 lines
I went through the source code of mckey utility (available e.g. here:
https://github.com/ofiwg/librdmacm/blob/master/examples/mckey.c ). There is an interesting comment at the line 524:/*
* Pause to give SM chance to configure switches. We don't want to
* handle reliability issue in this simple test program. */
sleep(3);
Quoted 4 lines
If we remove sleep function call, mckey does not work properly anymore (receiver side does not get all messages that were sent). Does anyone know how long does it take for SM to „configure switches“, and what is the proper way to „handle reliability issue“?
The SM time depends on the size of the fabric and how busy the SM is. I can't guess on the expected time; maybe someone who works closer with the SM will be able to guess. My comment about reliability was that the test assumes no dropped packets. If you wanted reliable multicast, you would need to add some sort of retry/acknowledgement protocol.
- Sean
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html