Latency Numbers Every Programmer Should Know: https://gist.github.com/jboner/2841832 discussion: https://hackernews.hn/item?id=4047623

DRAM (~100 ns) is very nearly 1000x better latency than SSD (~70 us) which is only 100x faster than disk (~10 ms).

Actually, modern SSDs have lower latency than 70 microseconds:

26 microseconds for OCZ Vertex 4 (reads)

http://thessdreview.com/our-reviews/ocz-vertex-4-128gb-ssd-r...

No more than 20 microseconds for Corsair Neutron GTX and Vertex 4 (writes):

http://www.storagereview.com/samsung_ssd_840_pro_review

While that's a significant improvement over the 70ms assumption, it's still "around three orders of magnitude slower than dram and two orders of magnitude faster than spinning disks", and doesn't really change any of the conclusions you'd have drawn using the 70ms number.

OCZ compromises a lot to get high marks on AnandTech benchmarks and great Farcry 3 FPS reports on newegg.

You should be looking at companies like STEC or higher end Intel SSDs for server applications.

This whitepaper goes into some detail about how/why to use SSDs as cache in a zfs fileserver: http://info.nexenta.com/rs/nexenta/images/white_paper_nexent...