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

[Anton Blanchard <anton@samba.org>]

Hi,

> This patch allows for this_load to set max_load, which if I understand
> the logic properly is correct.  It then adds a check to imbalance to make
> sure a negative number hasn't been coerced into a large positive number.
> With this patch applied, the algorithm is *much* more conservative ...
> maybe *too* conservative but that's for another round of testing ...

Good stuff, I just gave the patch a spin and things seem a little
calmer. However Im still seeing a lot of balancing going on within a
node.

Setup:

2 threads per cpu.
2 nodes of 16 threads each.

I ran a single "yes > /dev/null"

And it looks like that process is bouncing around the entire node.
Below is a 2 second average.

Anton

cpu    user  system    idle             cpu    user  system    idle

node 0:
cpu0      2       0      99             cpu1      9       0      91
cpu2      1       0      99             cpu3      8       0      92
cpu4      3       0      97             cpu5     10       0      90
cpu6      2       0      98             cpu7     10       0      90
cpu8      2       0      98             cpu9      9       0      90
cpu10     3       0      96             cpu11     9       0      90
cpu12     2       0      98             cpu13    10       0      90
cpu14     2       1      97             cpu15    10       1      89

node 1:
cpu16     0       0     100             cpu17     0       0     100
cpu18     0       0     101             cpu19     0       0     100
cpu20     0       0     100             cpu21     0       0     101
cpu22     0       0     101             cpu23     0       0     100
cpu24     0       0     100             cpu25     0       0     100
cpu26     0       0     100             cpu27     0       0     100
cpu28     0       0     101             cpu29     0       0     100
cpu30     0       0     100             cpu31     0       0     100