On Wed, 2014-08-27 at 14:55 -0400, Andrew Stitcher wrote: > ... > Of this only applies under the exact circumstances of these tests, until > we have more data points. I'd suspect it must also depend on the message > size too. >
Just to reply to myself... Keeping the credit window at around the MTU size is going to give you the most efficient networking transfers. if looks like we get maximum throughput with at least 20 messages in flight at once. That is something over 1KiB and will pack them into few transfer units. For regular ethernet the MTU is 1500, and for loopback on Linux it is 64KiB. So I'd guess that when the message is big enough it makes less difference. And it looks like once you can hit around 1KiB of content for each packet then you get small performance gain after that. Andrew
