>>>One thing to be wary of: to maintain performance and endurance, SSDs should periodically run the TRIM command. Many RAID controllers (especially 6 year old ones) don't know about SSDs so TRIM commands don't get passed to the drives. So, unless the RAID controller explicitly supports SSDs and TRIM, you'll be running any attached SSDs without the benefit of TRIM. This could lead to performance and endurance degradation faster than you might expect for standalone SSDs that get TRIMmed periodically.
>>>
>>>http://en.wikipedia.org/wiki/Trim_%28computing%29#RAID_issues
>>>http://en.wikipedia.org/wiki/Write_amplification#TRIM
>>
>>Hi Al,
>>
>>Thanks for the further info. I have found (for those who might be interested) the following:
>>
>>1) Samsung offer various software to monitor and improve their SSDs. See this link and look for "Magician Software" and "Performance Restoration Software";
>>
>>http://www.samsung.com/global/business/semiconductor/minisite/SSD/global/html/support/downloads.html
>>
>>2) My servers with SSD appear to be running TRIM which I tested using a command window and this command:
>>
>>http://www.eightforums.com/tutorials/39569-trim-support-ssd-check-enable-disable.html
>>
>>3) Samsung EVO SSD seem to have a very good long-life prospect. Tests done by AnandTech here:
>>
>>http://www.anandtech.com/show/7173/samsung-ssd-840-evo-review-120gb-250gb-500gb-750gb-1tb-models-tested/3
>
>Your links 1 and 2 are well and good but they don't address the issue of RAID.
>
>If you're using true hardware RAID controllers, the OS has no direct access to the drives at all. Yes, if the OS supports TRIM you can configure it to send TRIM commands, but if the drives are behind a RAID controller it depends entirely on how the controller responds to them.
>
>Six years ago SSDs didn't exist so there's no way a RAID controller of that era could possibly know what to do with a TRIM request. My guess is they'd be ignored. If you're lucky you might get Windows Event Log or RAID management software log entries saying "Unrecognized command - ignored" or words to that effect.
>
>A TRIM operation effectively takes a drive offline for an unknown and potentially significant amount of time. For sure a RAID controller can't apply it to all member drives at once. If the drives have no redundancy and are purely striped e.g. RAID0, the whole array would have to be taken offline, or incoming requests buffered or cached somehow until the TRIM was complete. With redundant RAID the controller could (at some risk) take one drive at a time offline, then resync when TRIM is complete. All this can get complicated in a hurry, it doesn't surprise me that HW RAID controller makers are struggling with this.
>
>I believe Intel has put some effort into supporting TRIM in its low-end ICH-based pseudo-RAID (what the *n?x community calls FakeRAID). I'm not sure of the status of that support, or even what its target market would be - it's not enterprise grade.
>
>Link 3 - yes, as I understand it Samsung enterprise drives have a good rep. If you're putting SSDs behind hardware RAID it's a good idea to use enterprise drives, since despite what your OS may be reporting, the drives almost certainly aren't getting TRIMmed.
>
>Somewhat related - more fun with Apple: http://hothardware.com/News/AntiCompetitive-Apple-Disables-Trim-Support-On-3rd-Party-SSDs-In-OS-X/


Understand your post, thanks for the info. It seems that the TRIM command performs the erase command on blocks so that when it comes to write to those blocks it will be done as fast as possible (since the SSD will not need to perform an erase first). This is a performance enhancement which I understand. However, I am not really worried since the nature of how we use all our servers tends to be 95% adding content to the disk which is never deleted. We just add content without hardly ever removing content. So this is not going to be a real issue for us. Switching two servers to SSD has already resulted in orders of magnitude faster operations than the HDDs that we were using.

Of much more importance to us is life-span of the disk and reliability. These seems to be quite good with Samsung EVO drives as per the tests mentioned above which seemed to use continuous overwrites to "wear" the drive down. But like I said above, we don't do much overwriting since we don't delete content. Seems to me we can expect our drives to last for years (barring manufacturing faults). And I am quite happy to replace drives every 5 years - cost of business.

Thanks again for the info.