i have a raid 5 configuration on HP gen9 server with eight hard disks (hp-10k-1.2TB disks). that server intended to serve as a Microsoft SQL sever and contains a database with 143GB size.

i have run a time consuming query on it and it accomplished in 10 minutes.(maximum read speed=200 MB/s)

then i stored that database on a single SSD drive(Samsung EVO 1TB) and same query accomplished in 3 minutes.(maximum read speed=380 MB/s)

the question is: why performance difference is so big? (considering the amount of 10k disks in use)


While the difference in throughput might not be so different between an 8-Disk RAID5 and a single SSD, the difference in latency will be orders of magnitude.

This implies, that every write - and there are many: Journal, Tablespace, ... - will take much longer and every random read will do as well (but not by so much as writes).

You will see, that the difference for an index-read-only query is much less (as long as the DB doesn't process any other load in the meantime), but the moment you do writes or index-less joins you will experience magnitudes of higher latency.

It is very likely, that in your second test the server was CPU-bound, or you might have experienced an even greater difference. And don't get me started on RAID5 with small writes ...

What you experience is exactly the reason, why databases on spinning rust are going the way of the Dodo.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.