Re: High volume inserts - more disks or more CPUs?

Hi!

> -----Ursprüngliche Nachricht-----
> Von: pgsql-general-owner(at)postgresql(dot)org
> [mailto:pgsql-general-owner(at)postgresql(dot)org] Im Auftrag von
> Guy Rouillier
> Gesendet: Montag, 13. Dezember 2004 07:17
> An: PostgreSQL General
> Betreff: [GENERAL] High volume inserts - more disks or more CPUs?

> (1) Would we be better off with more CPUs and fewer disks or
> fewer CPUs and more disks?

From my experience, it's generally a good idea to have as many disks as possible - CPU is secondary. Having enough RAM so that at least the frequently accessed parts of your db data including the indexes fit completely into memory is also a good idea.

> (3) If we go with more disks, should we attempt to split
> tables and indexes onto different drives (i.e., tablespaces),
> or just put all the disks in hardware RAID5 and use a single
> tablespace?

RAID5 is not an optimum choice for a database; switch to RAID0+1 if you can afford the disk space lost - this yields much better insert performance than RAID5, as there's no parity calculation involved. There's another performance gain to be achieved by moving the WAL-files to another RAID-set than the database files; splitting tablespaces across RAID-sets usually won't do much for you in terms of performance, but might be convenient when you think about scaling in size.

Kind regards

Markus