[TriLUG] SSD-based NFS servers for production - ready for primetime?

Ron Kelley rkelleyrtp at gmail.com
Sun Jan 15 13:08:22 EST 2012


Greetings all,

I need to add some NFS horsepower to our production network and find myself at a crossroads.  Our current NFS servers each have 8x Seagate 1TB 7200-RPM SATA drives hooked to Areca ARC1220 controllers.  They have been very stable/reliable over the past few years but the drives are starting to die off (3-4yrs @ 24/7 operation).  In addition, we have added more NFS clients to the mix (combination of ESX servers and CentOS clients) which has pushed the NFS servers to the point of I/O starvation.

I have been scouring the 'net looking for information about spinning drives vs SSDs in an enterprise NFS deployment.  After a ton of research, I am torn between the 10K-RPM/15KRPM 2.5" Seagate drives and the Intel 510 SSDs.  The SSDs would certainly fix the performance issue, but I am concerned they won't last as long especially given the large number of writes our NFS servers perform.  I know the 510s are not "enterprise grade" SSDs, but Anandtech and Tom's hardware has some very glowing reviews for them.  Here is a good link to a Tom's article from July 2011 about SSD reliability:  http://www.tomshardware.com/reviews/ssd-reliability-failure-rate,2923.html 

As for the configuration, I will put them in a 6-drive RAID-5 array using Linux software RAID.  If a drive fails, we can easily swap it out with another spare on the shelf.  And, as for space, we need about 1.5T of usable NFS space to store VMDKs and other high-sensitive items.  

Does anyone have an opinion or real-life experience as to the reliability of SSD drives in an enterprise setting? 



Thanks,

-----------------------------
Ron Kelley
rkelleyrtp at gmail.com




More information about the TriLUG mailing list