March 5, 2014 at 9:17 am
Hi,
I want to improve performacne of my SQL servers.
recently I've used a test server where I placed the tempdb on a raid1 SSD drive.
I got impresive perforamnce boost.
Now I want to apply this on my productions servers.
On one server I have a single slot free. That means that I can only use one disk with no mirroring.
Assuming I'll encounter a disk failure - do you see any other risks besides the down time required to replace the disk?
Thanks.
March 5, 2014 at 9:31 am
Nope, but understand, when you say Raid 0, that's not a single disk.
Raid 0 is a multidisk stripe that if any fail you lose the stripe. Single disk = same risk, just less throughput.
But no, not for tempdb. TempDB is rebuilt on server reboots so if it goes down, swap the disk and bounce the box. You'll only be down for the time of the physical swap and the box reboot.
Never stop learning, even if it hurts. Ego bruises are practically mandatory as you learn unless you've never risked enough to make a mistake.
For better assistance in answering your questions[/url] | Forum Netiquette
For index/tuning help, follow these directions.[/url] |Tally Tables[/url]
Twitter: @AnyWayDBA
March 6, 2014 at 5:06 am
Thanks, Kraig!
March 6, 2014 at 5:11 am
The risk is that when the SSD fails, your SQL Server instance shuts down and you won't be able to restart it until you replace the SSD or move TempDB to another drive.
Is management comfortable with that level of risk? Is the downtime allowance (RTO) for this server such that you will always be able to obtain and install a new SSD within the time permitted?
Gail Shaw
Microsoft Certified Master: SQL Server, MVP, M.Sc (Comp Sci)
SQL In The Wild: Discussions on DB performance with occasional diversions into recoverability
March 6, 2014 at 6:39 am
Yes, the consequences of downtime are understood. I just wanted to make sure I wouldn't find my self in a position where the disk failure will result in loss of data or will prevent me from reopening the SQL server after disk is replaced.
March 6, 2014 at 7:28 am
FYI, my experience with a SSD which is installed in my laptop has not been good. I've had 2 failures within less than 1 year. I will agree that the technology is great and hope that the industry offers an industrial grade SSD to improve overall reliability.
In theory along with on paper, it is a great idea.
My previous position we had a very large server that had allocated a large chunk of system memory as a "memory drive" to which tempdb was loaded. That solution worked fine.
Kurt
Kurt W. Zimmerman
SR DBA
Lefrak Organization
New York, NY
http://www.linkedin.com/in/kurtwzimmerman
March 6, 2014 at 11:28 am
I just read an article that stated they do NOT recommend TEMPDB be on SSD. Reason being is you can only write so many times to them before their performance starts to slow down and then eventually fail. SSDs are better for low writes and low read stuff. I am still not 100% comfortable with this technology. I can't remember where I found that article otherwise I would add the link to it here... sorry.
March 7, 2014 at 3:05 am
Evil Kraig F (3/5/2014)
Raid 0 is a multidisk stripe that if any fail you lose the stripe. Single disk = same risk, just less throughput.
Actually, a single disk would be a lower risk than the RAID-0--if any of the drives in a RAID-0 fail it takes down the entire array, and multiple disks means a higher chance of one of them failing compared to the single drive situation.
March 7, 2014 at 7:36 am
I have come across SSD raid arrays where there are basically two SSDs mounted on the same PCB along with a RAID controller. These are RAID 0 arrays but it may be possible to reconfigure as RAID 1. The ones I have seen reviewed are consumer devices but there may be similar server products available. This would bypass your single slot limitation.
March 7, 2014 at 8:29 pm
paul.knibbs (3/7/2014)
Evil Kraig F (3/5/2014)
Raid 0 is a multidisk stripe that if any fail you lose the stripe. Single disk = same risk, just less throughput.
Actually, a single disk would be a lower risk than the RAID-0--if any of the drives in a RAID-0 fail it takes down the entire array, and multiple disks means a higher chance of one of them failing compared to the single drive situation.
Well, point. Allow me to restate: Raid 0 and single disk have the same risk exposure on the failure of any drive in either setup. 🙂
Never stop learning, even if it hurts. Ego bruises are practically mandatory as you learn unless you've never risked enough to make a mistake.
For better assistance in answering your questions[/url] | Forum Netiquette
For index/tuning help, follow these directions.[/url] |Tally Tables[/url]
Twitter: @AnyWayDBA
Viewing 10 posts - 1 through 9 (of 9 total)
You must be logged in to reply to this topic. Login to reply