Accepted Solutions
On a small MSA 2000 fibre SAN.
I figure even just 6 of these in a RAID 5 will be enough to push the controllers close to their limits. But if their just index files, they'll be small bursts of data. The only process that will probably stress the SAS/SATA bus is the Re-Indexing or Re-build of indexes on a weekly maint job.
/If I can get away with just the 6, I might just sneak the other 2 in my desktop PC.....shhhhhh don't tell anyone though. 
COS wrote:
On a small MSA 2000 fibre SAN.
I figure even just 6 of these in a RAID 5 will be enough to push the controllers close to their limits. But if their just index files, they'll be small bursts of data. The only process that will probably stress the SAS/SATA bus is the Re-Indexing or Re-build of indexes on a weekly maint job.
/If I can get away with just the 6, I might just sneak the other 2 in my desktop PC.....shhhhhh don't tell anyone though.
Coming from the NexentaStor platform (Solaris) where using flash for log + L2ARC cache pretty much gets you about 10x the bang for your buck compared to Dell/HP/EMC systems, I've only seen them in mirrors and striping them from there, I have no clue if this is because of MTBF or what. Anyone want to chime in on this?
I avoid Mirroring for wear leveling purposes.
In a mirror, if you write a 2GB file, you write 2GB onto SSD A and 2 GB onto SSD B. This means each SSD wrote 100% of the data each.
In a 3 SSD RAID 5, when you write a 2GB file, you write 33% of the file data on SSD A, 33% of the file data on SSD B and 33% of the file data on SSD C. Thus only using 33% on each disk.
That spreads the "wear out" across the SSD's giving you longer "wear leveling" life. But neither of the two can withstand 2 drive failures. 
But that's life and you can adjust.
Even just the 3 SSD's in a RAID 5 will be enough for for a greater than 10x boost in performance. At over 500Mb/s write each you can most likely get at least 800Mb/s actual write throughput.
My data here at work is write intensive because of SQL. Reads are just a big plus that comes along with it.
My desktop at home get's 819Mb/s :smileyshocked: write with 4 slower (270 write) SSD's in RAID.
COS wrote:
I avoid Mirroring for wear leveling purposes.
In a mirror, if you write a 2GB file, you write 2GB onto SSD A and 2 GB onto SSD B. This means each SSD wrote 100% of the data each.
In a 3 SSD RAID 5, when you write a 2GB file, you write 33% of the file data on SSD A, 33% of the file data on SSD B and 33% of the file data on SSD C. Thus only using 33% on each disk.
That spreads the "wear out" across the SSD's giving you longer "wear leveling" life. But neither of the two can withstand 2 drive failures.
Even just the 3 SSD's in a RAID 5 will be enough for for a greater than 10x boost in performance. At over 500Mb/s write each you can most likely get at least 800Mb/s actual write throughput.
My data here at work is write intensive because of SQL. Reads are just a big plus that comes along with it.
My desktop at home get's 819Mb/s :smileyshocked: write with 4 slower (270 write) SSD's in RAID.
Yeah, I understand, it makes massive sense numbers wise, it's why I'm confuse why I haven't seen it before (at least in various benchmarks) and just wanted to point it out.