Connections from Brie to Oracle DBs on chevre and gruyere

[Ken Y. Clark]

On Mon, 29 Oct 2001, Xiaokang Pan wrote:

> Date: Mon, 29 Oct 2001 17:09:35 -0500 (EST)
> From: Xiaokang Pan <pan at brie.cshl.org>
> Reply-To: gramene at brie.cshl.org
> To: gramene at brie.cshl.org
> Subject: Re: Connections from Brie to Oracle DBs on chevre and gruyere
>
>
>     Please notice, currently, there are 9 people working on chevre but
> only 1 people working on gruyere. I think the work load is one of the main
> reasons which make it much slower when fetching from chevre than from
> gruyere.
>
>     Xiaokang

Absolutely!  Just look at the load average on chevre:

[chevre@~:]$ uptime
  5:12pm  up 35 day(s),  3:25,  16 users,  load average: 1.93, 1.91, 1.74

top points out some more interesting info:

last pid:  1660;  load averages:  2.03,  1.97,  1.80 17:15:40
128 processes: 125 sleeping, 1 running, 2 on cpu
CPU states:  4.3% idle, 88.7% user,  7.1% kernel,  0.0% iowait,  0.0% swap
Memory: 2048M real, 1342M free, 43M swap in use, 2007M swap free

  PID USERNAME THR PRI NICE  SIZE   RES STATE   TIME    CPU COMMAND
20206 groracle   1  10    0   12M 3560K cpu3   99.4H 48.47% tnslsnr
 1588 x-pan      1  10    0   38M   36M run    21:46 43.22% perl
[snip]

Wow.  Just those two processes seem to be eating most of the cycles.

Chevre is obviously much overworked.  I notice a *huge* difference
moving from my little Intel-based, single-processor server to the Sun
box.  I think we have a lot more progress to make on balancing the
various loads asked of our hardware.  Probably our Oracle box should
be very beefy and should not be asked to do much else.

ky

-- 
# -----------------------------
# k e n            Ken Y. Clark
# r   y      kclark at logsoft.com
# a l c            972-542-7923
# -----------------------------