From: Martin J. B. <mb...@ar...> - 2003-04-19 15:20:24
|
BTW, thinking more about the CPU load stuff, I'm confused as to why the node_balance stuff helps JBB so much ... the load average is less than 1 per cpu IIRC. Can you try doing the wakeup patch, but only throwing tasks around *within* a node instead of across nodes, and see if that makes a difference? If not, I guess we need to stare at the schedstat data again. I think the case when the busy node rebalance is broken (as far as I'm concerned) is when the load is low. If I have 4 tasks on one node (1 per cpu) and none on the other, that's just fine by me. That's what I said I was going to fix last night. However, I realised that sucks for other people ... we need a better metric here. Erich, I presume you want more perfect balancing across nodes for mem bandwidth concerns? Ie on a 4 node, 2 cpu per node system (which is what I thought yours was), you don't want 2/2/0/0 tasks for each node, you really, really want 1/1/1/1? Or is 2/2/0/0 just as good? M. |