Re: [PERFORM] Hardware suggestions for high performance 8.3

2008-06-27 Thread Joshua D. Drake
Greg Smith wrote: On Thu, 26 Jun 2008, Henrik wrote: I've seen some concerns about buying database performance hardware from DELL. Are there at least some of the RAID cards that work well with Linux or should I stay clear of DELL permanently? People seem to be doing OK if the RAID card is

Re: [PERFORM] Hardware suggestions for high performance 8.3

2008-06-26 Thread Henrik
I've seen some concerns about buying database performance hardware from DELL. Are there at least some of the RAID cards that work well with Linux or should I stay clear of DELL permanently? Thanks! //Henke 25 jun 2008 kl. 17.45 skrev Greg Smith: On Wed, 25 Jun 2008, Henrik wrote: Would

Re: [PERFORM] Hardware suggestions for high performance 8.3

2008-06-26 Thread Scott Marlowe
On Thu, Jun 26, 2008 at 10:47 PM, Greg Smith [EMAIL PROTECTED] wrote: On Thu, 26 Jun 2008, Henrik wrote: I've seen some concerns about buying database performance hardware from DELL. Are there at least some of the RAID cards that work well with Linux or should I stay clear of DELL

[PERFORM] Hardware suggestions for high performance 8.3

2008-06-25 Thread Henrik
Hi list, We have a database with lots of small simultaneous writes and reads (millions every day) and are looking at buying a good hardware for this. What are your suggestions. What we are currently looking at is. Dual Quad Core Intel 8 - 12 GB RAM 10 disks total. 4 x 146 GB SAS disk in

Re: [PERFORM] Hardware suggestions for high performance 8.3

2008-06-25 Thread Claus Guttesen
We have a database with lots of small simultaneous writes and reads (millions every day) and are looking at buying a good hardware for this. What are your suggestions. What we are currently looking at is. Dual Quad Core Intel 8 - 12 GB RAM 10 disks total. 4 x 146 GB SAS disk in RAID 1+0

Re: [PERFORM] Hardware suggestions for high performance 8.3

2008-06-25 Thread Matthew Wakeling
On Wed, 25 Jun 2008, Henrik wrote: What are your suggestions. What we are currently looking at is. Dual Quad Core Intel 8 - 12 GB RAM More RAM would be helpful. It's not that expensive, compared to the rest of your system. 10 disks total. 4 x 146 GB SAS disk in RAID 1+0 for database 6 x

Re: [PERFORM] Hardware suggestions for high performance 8.3

2008-06-25 Thread Henrik
25 jun 2008 kl. 12.56 skrev Claus Guttesen: We have a database with lots of small simultaneous writes and reads (millions every day) and are looking at buying a good hardware for this. What are your suggestions. What we are currently looking at is. Dual Quad Core Intel 8 - 12 GB RAM 10

Re: [PERFORM] Hardware suggestions for high performance 8.3

2008-06-25 Thread Henrik
25 jun 2008 kl. 13.15 skrev Matthew Wakeling: On Wed, 25 Jun 2008, Henrik wrote: What are your suggestions. What we are currently looking at is. Dual Quad Core Intel 8 - 12 GB RAM More RAM would be helpful. It's not that expensive, compared to the rest of your system. True, as long as

Re: [PERFORM] Hardware suggestions for high performance 8.3

2008-06-25 Thread Claus Guttesen
If you have a good RAID controller with BBU cache, then there's no point splitting the discs into two sets. You're only creating an opportunity to under-utilise the system. I'd get ten identical discs and put them in a single array, probably RAID 10. OK, thats good to know. Really want to

Re: [PERFORM] Hardware suggestions for high performance 8.3

2008-06-25 Thread Matthew Wakeling
On Wed, 25 Jun 2008, Henrik wrote: Would you turn off fsync if you had a controller with BBU? =) No, certainly not. Fsync is what makes the data move from the volatile OS cache to the non-volatile disc system. It'll just be a lot quicker on a controller with a BBU cache, because it won't

Re: [PERFORM] Hardware suggestions for high performance 8.3

2008-06-25 Thread Greg Smith
On Wed, 25 Jun 2008, Henrik wrote: 4 x 146 GB SAS disk in RAID 1+0 for database 6 x 750 GB SATA disks in RAID 1+0 or RAID 5 for OS and transactions logs. The transaction logs are not that big, and there's very little value to striping them across even two disks. You should just get more SAS