Hi, Derick Swanepoel. I'll reply inline between selected quotes below.
On Sep 6, 2013, at 9:16 AM, Derick Swanepoel wrote: ... > * Jumbo frames (9000) configured and working on target, initiator and Dell > PowerConnect 8132 switch (with hardware flow control enabled). There's an "aoe-sancheck" tool in the aoetools that you can use just to double check network things. But if you haven't tried it already, using a direct connection for comparison is a good sanity check. > * iperf benchmarks TCP performance at 9.87 Gb/s. > * aoetools-36, vblade-21 and kernel 3.10.7 with AoE v81 Great. The parallelism that v83 adds was implemented per target, so for a single target it will probably not affect your performance. By the way, are you using the v81 driver that's inside of the 3.10.7 kernel? If so, how did you know to use the aoetools v36? I sometimes worry that people won't find it when they notice aoe in the /lib/modules. (The driver at coraid.com comes bundled with the aoetools.) ... > I exported the RAID 60 as e0.0. Both read and write performance is > disappointing, though: [around 300MB/s] That's not really too surprising given your stack. (More below.) > dd, vblade and the aoe kernel threads are well below 100% CPU usage during > these tests. What am I missing here? I am surprised to find nothing to this effect in the vblade's README, and it should be added, but the vblade is supposed to be a completed piece of software, like TeX, that serves as a minimal example of how an AoE target can be implemented in software. It's great that the program is so small and easy to understand, but performance wasn't really a design goal, and that is occasionally surprising to first-time users. So an update to the README seems to be needed. We do sell AoE targets that are very fast at Coraid, but we have to work hard to get that performance. I don't have a lot of experience with the other non-Coraid AoE targets that are out there, but you might check whether one of them that's oriented more toward performance could be useful to you. That said, while checking the vblade README for the design goals, I noticed that it advertises a capacity for 16 outstanding commands. If you want to try some tuning, you could adjust Bufcount in dat.h and then make sure your settings in /proc are sufficient to allow the kernel to buffer 16 writes. (Read commands are small.) -- Ed Cashin ecas...@coraid.com ------------------------------------------------------------------------------ Learn the latest--Visual Studio 2012, SharePoint 2013, SQL 2012, more! Discover the easy way to master current and previous Microsoft technologies and advance your career. Get an incredible 1,500+ hours of step-by-step tutorial videos with LearnDevNow. Subscribe today and save! http://pubads.g.doubleclick.net/gampad/clk?id=58041391&iu=/4140/ostg.clktrk _______________________________________________ Aoetools-discuss mailing list Aoetools-discuss@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/aoetools-discuss
