Re: hardware for a server - Mailing list pgsql-general

From A B
Subject Re: hardware for a server
Date
Msg-id dbbf25901003150922u62e7c5ablf077ef4c38f98aa0@mail.gmail.com
Whole thread Raw
In response to Re: hardware for a server  (Greg Smith <greg@2ndquadrant.com>)
Responses Re: hardware for a server  (Greg Smith <greg@2ndquadrant.com>)
List pgsql-general
> Don't put SAS drives on a 3ware controller.  They say that works now, but
> they haven't really gotten it right yet--their controllers are still only
> good with SATA drives.

How bad will it be with SAS drives? Is there so little performance
gain witn 3ware+SAS?


Scott Marlowe stated in earlier reply that Seagates ES.2 disks are not
very good, which would leave the SAS Cheetah discs with the LSI card.
The LSI card is down to 128 MB memory.
But LSI+SAS is still the clear winner over 3ware with  512 MB and SAS/SATA?


> In your situation, I'd probably get a pair of TB drives for the OS and
> "other data", just to get them all out of the way on one place to fight with
> each other, then use whatever budget is leftover to get the best performing
> drives you can to create a 6-disk RAID10 for the database + xlog.  If all
> six of those can only be a smaller drive instead after that, that's not such
> a bad combination--you can always grab a larger capacity drive as your spare
> and then put it anywhere in the array in an emergency.

>Don't mix SAS and SATA; vendors will tell you it works, but it's extremely painful when it doesn't, and that happens
sometimes.

Does that also forbid the case when you create two raid arrays, let
say a raid-1 with only SATA discs (huge discs) and a  raid-10 with
only SAS drives? (as your example with the 2/6 split)
There are internal SATA controllers so I don't have to bother the Raid
card with a pair of SATA drives, but I'd prefer to use the BBU  for
all the drives. Data loss is not funny.

pgsql-general by date:

Previous
From: Vick Khera
Date:
Subject: Re: libpq: compatibility with server versions
Next
From: Alvaro Herrera
Date:
Subject: Re: Postgres 8.4 Segfault on xpath