On Thu, Jul 8, 2010 at 9:53 AM, Kevin Grittner <Kevin.Grittner@xxxxxxxxxxxx> wrote:
Eliot Gable <egable+pgsql-performance@xxxxxxxxx> wrote:FWIW, I concur that the description so far suggests that this server
> For about $2k - $3k, you can get a server that will do upwards of
> 300 MB/sec, assuming the bulk of that cost goes to a good
> hardware-based RAID controller with a battery backed-up cache and
> some good 15k RPM SAS drives.
either doesn't have a good RAID controller card with battery backed-
up (BBU) cache, or that it isn't configured properly.
On another note, it is also entirely possible that just re-writing your queries will completely solve your problem and make your performance bottleneck go away. Sometimes throwing hardware at a problem is not the best (or cheapest) solution. Personally, I would never throw hardware at a problem until I am certain that I have everything else optimized as much as possible. One of the stored procedures I recently wrote in pl/pgsql was originally chewing up my entire development box's processing capabilities at just 20 transactions per second. It's a pretty wimpy box, so I was not really expecting a lot out of it. However, after spending several weeks optimizing my queries, I now have it doing twice as much work at 120 transactions per second on the same box. So, if I had thrown hardware at the problem, I would have spent 12 times more on hardware than I need to spend now for the same level of performance.
If you can post some of your queries, there are a lot of bright people on this discussion list that can probably help you solve your bottleneck without spending a ton of money on new hardware. Obviously, there is no guarantee -- you might already be as optimized as you can get in your queries, but I doubt it. Even after spending months tweaking my queries, I am still finding things here and there where I can get a bit more performance out of them.
--
Eliot Gable
