Re: [PATCH] Load balancing problem in 2.6.2-mm1

[Nick Piggin <piggin@cyberone.com.au>]

Martin J. Bligh wrote:

>>From a later email ....
>>
>>Hopefully just tending to round down more would damp it better.
>>*imbalance = (*imbalance + SCHED_LOAD_SCALE/2) >> SCHED_LOAD_SHIFT;
>>Or even remove the addition all together.
>>
>
>I'd side with just removing the addition alltogether ...
>
>

By that stage though, it has already passed through the
imbalance_pct filter, and with the higher precision and
averaging of previous loads, it might take a while to
get there.

>>>Moreover, as Rick pointed out, it's particularly futile over idle cpus ;-)
>>>
>>I don't follow...
>>
>
>If CPU 7 has 1 task, and cpu 8 has 0 tasks, there's an imbalance of 1.
>There is no point whatsoever in bouncing that task back and forth
>between cpu 7 and 8 - it just makes things slower, and trashes the cache.
>There's *no* fairness issue here.
>
>

Right, it should not be moved. I think Anton is seeing a problem
with active balancing, and not so much a problem with the imbalance
calculation though.

>If CPU 8 has 2 tasks, and cpu 1 has 1 task, there's an imbalance of 1.
>*If* that imbalance persists (and it probably won't, given tasks being
>created, destroyed, and blocking for IO), we may want to rotate that 
>to 1 vs 2, and then back to 2 vs 1, etc. in the interests of fairness,
>even though it's slower throughput overall.
>

Yes, although as long as it's node local and happens a couple of
times a second you should be pretty hard pressed noticing the
difference.