HP will soon introduce enterprise-class solid state drives (SSD) with a 6Gbit/s SAS interface for its servers. Currently it offers 3Gbit/s SSDs from Samsung for its G6 and certain G5 servers. These have a sustained read speed of 230MB/sec and a sustained write speed of 180MB/sec. TechEye reckons the coming SSDs have sustained …
I thought the biggest SSDs available were 256GB, and they're very expensive.
It's so very exciting to have SSD more widespread. I'm looking forward to the price (slowly) coming down so I can afford one.
they have that expensive part of it covered...
The existing 60GB hot plug 3g/s SATA goes for around $1400. $3k+ for the 120GB version.
Why are they so expensive? And why do they only have a 1 year warranty? --There are no moving parts and a lot less heat generated, so you would think a SSD would fail a lot less often than a spinning hard drive.
Totally unimpressed by this piece of news.
Can enterprise SSDs really be given multi-year warranties?
I'm not sure about the cost of the SSDs ($1400 for a 60 GB SSD sure seems high to me), but I'm not surprised by the limited 1 year warranty. I assume that the inherent limitations on the number of writes a NAND cell can handle prevents the vendor from guaranteeing the drives for a longer 3 to 5 year span, which are usually seen with traditional HDDs. If a customer uses an SSD in a write intensive environment, the drive life span can be dramatically shortend through no fault of the SSD vendor - this is just the nature of the technology at the moment.
Yes, no moving parts and less heat generated, but that does not contribute to their failure rate. With SSDs, failure rate is determined by write cycles. HP figured that chunking data under a 100% workload for 1 year (up to the 3-5yr range for the non "boot drive" SSDs) would be the reliable write-endurance for these drives. Of course, if you don't continuously stream data to the drive, you could see longer life expectancies from these drives. HP just won't replace them under warranty if they crap out on you. I'd love to stick a few of these bad boys in my DB server under RAID6. I'd just feel sorry for my RAID card.
OT: Anyone know the dynamics of TRIM with the SSDs in a RAID? One would think that since the drives are in a RAID, each delete request wouldn't be able to request a TRIM on each disk affected...but I'm not sure.
TRIM on RAID seems iffy at the moment
It seems that TRIM support, at least for for Intel RAID volumes, is not available at the moment.
"Intel® Rapid Storage Technology 9.6 supports TRIM in AHCI mode and in RAID mode for drives that are not part of a RAID volume."
It seems very unlikely to me that trim will work with RAID arrays unless the RAID array itself supported trim. Easy enough with RAID1, but much more difficult with RAID5, although I imagine it could be done. The same would be true for any logical volume manager - it would somehow have to pass through trim requests to the physical drives.
In the case of a DB server then trim might not give you much anyway. DB files tend to remain in situ and are not created and deleted frequently.
Trim: RAIDZ is the answer?
What RAID6? RAIDZ needs no specific controller, which is a huge safety feature imho, takes care of bit decay, and support trim as far as I know. What is there no to like?