0

On a new db server I only have 10 disks to play with,

The usage is about a booking every 3-5 seconds, so not high volume, I know compromises have to be made, but my initial thoughts are -

DISK 1 & 2 - RAID 1 - OS
DISKS 3,4,5,6 - RAID 10 - Data, Indexes & TempDB
DISKS 7,8,9,10 - RAID 10 - Logs & Backup

Full backups will take place when there is virtually no traffic on the website so not bothered about the contention with the logs.

disk 3-10 - 8kb NTFS unit allocation size
disk 3-10 - 64kb Disk Stripe size

does this seems to be sensible, any other considerations I have omitted ?

thanks

Chris Wood
  • 133
  • 5

2 Answers2

0

Nearly there, I'd do;

  • 1 & 2 - OS - R1
  • 3 - 6 - Data etc - R10
  • 7 & 8 - Logs - R1
  • 9 & 10 - Backup - R1

You don't need much space for logs, so I'd split the logs and backup - otherwise fine, stripe and block's fine too.

anon
  • 404
  • 1
  • 5
  • 15
Chopper3
  • 101,299
  • 9
  • 108
  • 239
  • I would put the tempdb (DB) on the OS discs. If that one starts being used (IO wise) the memory is too small - tempdb tries to avoid actually WRITING to the disc. – TomTom Jun 24 '12 at 17:14
0

You'll want 64k blocks across the board. SQL does almost all of it's IO in 64k blocks.

mrdenny
  • 27,174
  • 4
  • 41
  • 69