bergstpg
asked on
Track alignment in Symantec Storage Foundation for Windows 2000 and 2003 - Hosting SQL Server 2000 and 2005 Databases
It's recommended to align disk after a Windows installation to drastically improve I/O performance within SQL Server 2000 or 2005. There are several articles throughout the web on the benefits to making this adjustment, so I won't get into it here. Bottom line - proper alignment can give 40% improved performance for SQL Server - on Windows 2000 and 2003.
Symantec automatically aligns the disk within Storage Foundation - however - it sets the offset to be 96 sectors(49152 bytes) for Hitachi Disk - even though Hitachi specifically suggests 64 sectors (34k or 32768 bytes). For EMC disk SFW sets it to 128 sectors (64k or 65536 bytes) - which is the standard for most other disk - including non-Dynamic disk.
Are we the only ones seeing this? Has any one done any performance tests when set to 96 sectors versus 128 or 64? Unfortuately this would be a hack since it isn't a configurable option within SFW. General thoughts?
Symantec automatically aligns the disk within Storage Foundation - however - it sets the offset to be 96 sectors(49152 bytes) for Hitachi Disk - even though Hitachi specifically suggests 64 sectors (34k or 32768 bytes). For EMC disk SFW sets it to 128 sectors (64k or 65536 bytes) - which is the standard for most other disk - including non-Dynamic disk.
Are we the only ones seeing this? Has any one done any performance tests when set to 96 sectors versus 128 or 64? Unfortuately this would be a hack since it isn't a configurable option within SFW. General thoughts?
Last Comment
ASKER
The block size is 512k - standard block size.
That doesn't sound right. EMC CLARiiON standard block size is 64K, and from what you've posted, HDS block size will be similar. 512K is very large indeed for block size at the RAID level, but it depends on the array.
I found this Symantec support article that describes how to configure alignment offset: http://seer.entsupport.symantec.com/docs/292492.htm
You may need this as well: http://seer.entsupport.symantec.com/docs/291720.htm
Looking at the first support article, perhaps the alignment offset was modified at some stage for your HDS array?
I found this Symantec support article that describes how to configure alignment offset: http://seer.entsupport.symantec.com/docs/292492.htm
You may need this as well: http://seer.entsupport.symantec.com/docs/291720.htm
Looking at the first support article, perhaps the alignment offset was modified at some stage for your HDS array?
ASKER
Sorry 512bytes. My bad.
BTW - that's sector size. Block size is the number of sectors written to each disk before moving to the next disk in a striped RAID schema (exception is RAID 3).
ASKER
Back to the alignment issue.
Hitachi is suppose to be set to 128 sectors - according to their documentation. However some of our arrays are being read by the SFW as Open-V systems when they are really Hitachi Lightning 9990v and HP XP 24000 arrays - but are being offset at 96 sectors. We are currently having Veritas look at the issue and still believe 128 or 64 sector alignment is what we need for best performance.
Hitachi is suppose to be set to 128 sectors - according to their documentation. However some of our arrays are being read by the SFW as Open-V systems when they are really Hitachi Lightning 9990v and HP XP 24000 arrays - but are being offset at 96 sectors. We are currently having Veritas look at the issue and still believe 128 or 64 sector alignment is what we need for best performance.
ASKER CERTIFIED SOLUTION
THIS SOLUTION IS ONLY AVAILABLE TO MEMBERS.
View this solution by signing up for a free trial.
Members can start a 7-Day free trial and enjoy unlimited access to the platform.
Microsoft Legacy OS
The Microsoft Legacy Operating System topic includes legacy versions of Microsoft operating systems prior to Windows 2000: All versions of MS-DOS and other versions developed for specific manufacturers and Windows 3/3.1, Windows 95 and Windows 98, plus any other Windows-related versions, and Windows Mobile.
55K
Questions
--
Followers
--
Top Experts
Get a personalized solution from industry experts
TRUSTED BY
I suspect that the reason the offset is different for HDS is perhaps the array has been configured with non-standard block sizes for that specific RAID group?
Incidentally, if your I/O profile is highly sequential in nature, you can get some worthwhile performance gains by offsetting a complete stripe, rather than just the first block. Doesn't apply to databases as a rule, I'm afraid.