Click here to monitor SSC
SQLServerCentral is supported by Red Gate Software Ltd.
 
Log in  ::  Register  ::  Not logged in
 
 
 
        
Home       Members    Calendar    Who's On


Add to briefcase ««12

How will SSDs change SQL Server storage arrays? Expand / Collapse
Author
Message
Posted Wednesday, December 8, 2010 8:57 AM


SSC-Dedicated

SSC-DedicatedSSC-DedicatedSSC-DedicatedSSC-DedicatedSSC-DedicatedSSC-DedicatedSSC-DedicatedSSC-DedicatedSSC-DedicatedSSC-DedicatedSSC-DedicatedSSC-DedicatedSSC-Dedicated

Group: Administrators
Last Login: Today @ 3:31 PM
Points: 33,051, Visits: 15,159
I think the current generation of FusionIO, OCZ, Intel, Patriot, etc SSDs might not have enough of a track record to determine if they are really enterprise stable. There just haven't been enough people using them for a long enough time to determine how well they stand up. Note that I'm not saying you shouldn't use them, but be careful. There are people that have had great luck, using them for over a year in busy systems, and other people that have seen them burn out.

I would make doubly sure my backups are running, copied, and safe, but then go ahead with testing and see how they perform. I've had an Intel in my laptop for about 8 months and it works great.







Follow me on Twitter: @way0utwest

Forum Etiquette: How to post data/code on a forum to get the best help
Post #1031951
Posted Wednesday, December 8, 2010 11:26 AM
Forum Newbie

Forum NewbieForum NewbieForum NewbieForum NewbieForum NewbieForum NewbieForum NewbieForum Newbie

Group: General Forum Members
Last Login: Monday, December 20, 2010 8:33 AM
Points: 9, Visits: 81
Once again, FLASH based tech is absolutely fabulous.

But, it does physically wear out over time - that’s the physics of the pup. And there is nothing inherent to FLASH that will warn of imminent failure, so cover your *** big time. Using some sort of RAID 1+0 / RAID 0+1 sounds like a winner to me in this case - it has the lowest number of logical reads & writes that are implemented at the RAID level compared to the other RAID types. And keep a spare on the shelf.

Otherwise get a non-FLASH based SSD (10x the price or so)

Brad

P.S.:
I took a look at the OCZ drive specs at http://www.ocztechnology.com/products/solid-state-drives/pci-express/revodrive/ocz-revodrive-pci-express-ssd-.html

It is sorta weird to me that a solid-state component manufacturer cites MTBF (Mean Time Between Failure) at 2,000,000 hours (288+ years) and the limited warranty is 3 years (26,280 hours or 1.3% of MTBF).
Post #1032065
Posted Wednesday, December 8, 2010 1:21 PM
Ten Centuries

Ten CenturiesTen CenturiesTen CenturiesTen CenturiesTen CenturiesTen CenturiesTen CenturiesTen Centuries

Group: General Forum Members
Last Login: Thursday, July 17, 2014 12:34 PM
Points: 1,414, Visits: 4,539
has anyone used SSD's on SQL 2005 on Windows 2003? I just sent for a quote for 4 Proliant DL 380 G7's with 2 of them getting SSD's for tempdb. For now we're staying on Windows 2003 R2 and SQL 2005 and the only thing i'm worried about is the lack of TRIM support

https://plus.google.com/100125998302068852885/posts?hl=en
http://twitter.com/alent1234
x-box live gamertag: i am null
[url=http://live.xbox.com/en-US/MyXbox/Profile[/url]
Post #1032138
Posted Friday, December 17, 2010 2:57 AM
Forum Newbie

Forum NewbieForum NewbieForum NewbieForum NewbieForum NewbieForum NewbieForum NewbieForum Newbie

Group: General Forum Members
Last Login: Saturday, May 10, 2014 10:27 AM
Points: 7, Visits: 194
Hi again,

I have done some more testing and one thing that took me by surprice at first , was the fact that compressed tables are slower on SSD drives....

Running a ETL job between two REVO Drives (Running the same ETL with everything on one drive is just as fast... )is actually faster if the tables involved are not compressed. I believe that this is because there is no disk latency what so ever and compressing an uncompressing actually adds an extra overhead. CPU's are not fully utilized so that is prabably not the reason either. I have no waits on the server while running the ETL.
When I run the same ETL on normal disks, compressed disks are a bit faster then uncompressed, but there is no dramatic perfromance gain. Space usage on the other hand is wastly reduced.

Any thoughts on this ?

Staffan
Post #1036412
Posted Friday, December 17, 2010 8:01 AM
Forum Newbie

Forum NewbieForum NewbieForum NewbieForum NewbieForum NewbieForum NewbieForum NewbieForum Newbie

Group: General Forum Members
Last Login: Friday, December 23, 2011 2:34 PM
Points: 3, Visits: 27
Hi Stefan,

What are the CPU'S and memory when you are running ETL. Are you using SSIS or SQL Scripts?

Best,

mac
Post #1036547
Posted Monday, December 20, 2010 12:41 AM
Forum Newbie

Forum NewbieForum NewbieForum NewbieForum NewbieForum NewbieForum NewbieForum NewbieForum Newbie

Group: General Forum Members
Last Login: Saturday, May 10, 2014 10:27 AM
Points: 7, Visits: 194
CPU's are only running at 65-80 %, it's dual CPU dual core machine, part of it is due to calculations taking place in the ETL process. The only waits I can obeserve is different latch waits and I believe that is ok.

ETL is run as an SSIS package.

/Staffan
Post #1037047
Posted Monday, December 20, 2010 12:42 AM
Forum Newbie

Forum NewbieForum NewbieForum NewbieForum NewbieForum NewbieForum NewbieForum NewbieForum Newbie

Group: General Forum Members
Last Login: Saturday, May 10, 2014 10:27 AM
Points: 7, Visits: 194
Memory is all 28 Gb( 4 Gb left for OS) allocated to SQL server.
/S
Post #1037048
Posted Wednesday, December 29, 2010 7:06 AM
Forum Newbie

Forum NewbieForum NewbieForum NewbieForum NewbieForum NewbieForum NewbieForum NewbieForum Newbie

Group: General Forum Members
Last Login: Friday, September 20, 2013 6:35 AM
Points: 1, Visits: 7
SS 2005, Win 2003 Server, running under ESXi 4.0, 3ware controller, Crucial SSDs.

We started out with one 3ware 9650 controller that apparently could not keep up and so it corrupted our databases. We worked with 3ware support and we are now also running a 9750-4i to run these SSDs in RAID, leaving the 9650 for the spindles. I have yet to move any files back on to the array for testing. Make backups AND pay attention to what is going on in the environment. You best be running DBCC's and staying on top of it. Work with the vendors and manufacturers as much as possible - even if you spec everything out with plenty of fudge, you just never know.

My opinion would be that the most prominent performance boost would be from tempdb and log files being on the SSD. Since changes are written back to the db file lazilly does it make sense that it is not as critical to have them on SSD? But I guess it all depends on whether your heavy transactional or heavy reads.
Post #1040260
« Prev Topic | Next Topic »

Add to briefcase ««12

Permissions Expand / Collapse