Re: A DBA philosopical question

>What people that defend RAID 5 can not do, however, is repeal the laws
>of physics. Parity calculations and 200%+ more writes can not be free.

The trick with dealing with the laws of physics is to avoid the ones that penalize you. Hydrogen power is much more efficient than gasoline, but what do you put in the Jag? Engineering a solution to any problem requires dealing with actual costs and economic realities as well as physical laws and existing infrastructure. Implementing a solution adds dealing with social realities as well.

I hate defending RAID 5. If it were up to me it would be BAARF all the way. But no one pays for it to be up to me. The important thing is... it works for some sets of problems. Where it doesn't work is a much more powerful argument against it. If you are spec'ing systems to be near capacity at light-off, something is wrong. So the RAID-5 performance issue can be put off to the future, after other tuning, given sufficient initial capacity. When it becomes a problem, the cause can be demonstrated empirically. Or you can yank one disk and make a demonstration, if you must.

Hans, for many products the cache is there however you array the disks, so the delta is how many/size disks you must buy. We techies know performance is often dependent on number of spindles, but we hat-in-hand-consultants have to specify how big the db and all related usage is going to be and someone has to fit it in a fixed budget. When the choice is between me or hardware... I feel fortunate to be chosen.

jg

--
@home.com is bogus.
"It's about time for RAID-5 to be more generally accepted, so must be
about time for me to be a dinosaur against it." - member 127
http://www.miracleas.com/BAARF/BAARF_members_sql.php