shri - 1:55 am on Jan 2, 2011 (gmt 0)
>> shri, what does one's "experience" have to do with using an SSD drive?
Can you predict how your drive will behave? For example, the whole write wearing issue bugs the hell out of me, simply because I cannot find much written up about it on an enterprise level.
Say a cell can handle 100,000 writes. Might seem ok on a consumer grade device because cell are not hammered on a regular basis, but what happens if that cell contains the "views" column of a database table that catalogs the number of times a thread has been read. We have several threads that can get a 100K+ views a month.
Will the drive behave properly?
Do any of the drives really support TRIM on a RAID card under Linux / FreeBSD setups?
Are they being used in 1000's of similar servers and conditions? Is there local datarecovery expertise to rebuild SSD data? Rebuild SSD Raid arrays?
And then there is the price advantage. Is there really a price advantage on the performance of a good SAS 15K array v/s the performance of a good SSD based array?
Have you looked at other exsiting ways of improving your performance? Is the drive the only part of the hardware / software stack where the performance can be improved? (I doubt it is..)
There is still far to little documented about how SSDs will predictably behave under enterprise loads for me to bet my business on it.