Hello all,
My website has been having issues with our new Linux/PostgreSQL
server being somewhat slow. I have done tests using Apache Benchmark
and for pages that do not connect to Postgres, the speeds are much
faster (334 requests/second v. 1-2 requests/second), so it seems that
Postgres is what's causing the problem and not Apache. I did some
reserach, and it seems that the bottleneck is in fact the hard
drives! Here's an excerpt from vmstat:
procs -----------memory---------- ---swap-- -----io---- --system--
-----cpu------
r b swpd free buff cache si so bi bo in cs us
sy id wa st
1 1 140 24780 166636 575144 0 0 0 3900 1462 3299 1
4 49 48 0
0 1 140 24780 166636 575144 0 0 0 3828 1455 3391 0
4 48 48 0
1 1 140 24780 166636 575144 0 0 0 2440 960 2033 0
3 48 48 0
0 1 140 24780 166636 575144 0 0 0 2552 1001 2131 0
2 50 49 0
0 1 140 24780 166636 575144 0 0 0 3188 1233 2755 0
3 49 48 0
0 1 140 24780 166636 575144 0 0 0 2048 868 1812 0
2 49 49 0
0 1 140 24780 166636 575144 0 0 0 2720 1094 2386 0
3 49 49 0
As you can see, almost 50% of the CPU is waiting on I/O. This doesn't
seem like it should be happening, however, since we are using a RAID
1 setup (160+160). We have 1GB ram, and have upped shared_buffers to
13000 and work_mem to 8096. What would cause the computer to only use
such a small percentage of the CPU, with more than half of it waiting
on I/O requests?
Thanks a lot
Jason
