On Tue, Mar 9, 2010 at 2:52 PM, Jonathan Ellis <jbel...@gmail.com> wrote: > By "reads" do you mean what stress.py counts (rows) or rows * columns? > If it is rows, then you are still actually reading more columns/s in > case 2.
Well, unless I'm mistaking, that's the same in my example as I give in both case to stress.py the option '-c 1' which tells it to retrieve only one column each time even in the case where I have 100 columns by row. >> And it don't go up significantly after >> that. Turns out this seems to be a GC problem. Indeed, the info log (I'm >> running trunk from today, but I first saw the problem on an older version of >> trunk) show every few seconds lines like: >> GC for ConcurrentMarkSweep: 4599 ms, 57247304 reclaimed leaving >> 1033481216 used; max is 1211498496 > > First, use the 0.6 branch, not trunk. we're breaking stuff over there. Fair enough, I will do the test with 0.6. But again, I saw such behavior with a trunk from like 3 weeks ago. Just to say that I don't believe it to be something broke recently. But I admit, I should have tried with 0.6 and I will do it. > What happens if you give the jvm 50% more ram? A quick test doesn't show the problem with 50% more ram, at least not in a short time frame. But I'm still not convinced there is no problem, I saw pretty weird performance with bigger columns. Let me try to come up with a more compelling test against 0.6. I'll keep you posted, even if I'm wrong :) > Are you using a 64-bit JVM? yep -- Sylvain