[ofa-general] RDMA vs Shared Memory

[Majumder, Rajib]

Hello,

I was trying to find out fastest IPC where data source and sink run on
the same host on SLES 10, running ConnectX and OFED 1.3.0. It seems RDMA
is performing much better than shared memory. 

Mode 			Latency (microsecs)
Throughput 
-------			-----------------------------
----------------
SHM			20 (32 bytes)				17 Mbps 
RDMA			1.07 (32 bytes)				1.2 Gbps
SHM			70 (32k)
5.2 Gbps
RDMA			30 (32k)
8.5 Gbps

Can someone explain why RDMA's giving better performance on the same
host? Is it only kernel bypass and zcopy? 

Thanks

Rajib			   



==============================================================================
Please access the attached hyperlink for an important electronic communications disclaimer: 

http://www.credit-suisse.com/legal/en/disclaimer_email_ib.html
==============================================================================
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openfabrics.org/pipermail/general/attachments/20080502/438b29fb/attachment.html>