Tangentially related: I'm a bit concerned about using SSDs for server-side storage: Even if you RAID them, you still bought them at around the same time and wear leveling will work on all the drives in about the same way (when you consider RAID1). That means that the drives should die all together within a very short time span.<p>Hard Drives of course fail too, but they don't have a physical limit of how many times they can be written to, so at least in theory they shouldn't fail at the same time (especially when you buy them from different manufacturing batches).<p>Maybe I'm over-cautious, but IMHO the only safe way to use SSDs on a server is to replace half of them every safe period (which is unknown - 6 months?). This is very annoying maintenance you don't have to do with real harddrives where you can just swap in the hot-spare and then replace the broken one at some point close up.
Interesting stuff, thanks for writing this up :)<p>Personally, I've moved entirely away from VPSs to dedicated servers instead, mostly because I don't have the need to rapidly fire up lots of servers on demand. I think this may be the year where I start investing in SSDs for dedicated hardware, it seems the cost vs. performance gain is now heavily in their favour.
Very interesting, I found a thread on Quora with more info and pricing if anyone is interested... <a href="http://www.quora.com/Amazon-Web-Services/When-did-Amazon-Web-Services-begin-offering-SSDs-in-its-public-cloud" rel="nofollow">http://www.quora.com/Amazon-Web-Services/When-did-Amazon-Web...</a>