This is what I'm semi-ranting about. RAID 5 doesn't give you
enough bang for the cost (processing, time, or money) to justify
it in most cases. But a lot of people out there have this idea
that RAID 5 is going to provide you with maximum uptime, nevermind
the rest of the issues. It's not so. RAID 0+1 and RAID 5 are
nearly equivalent in terms of data availability. And the >very<
small gain you get from RAID 5 is offset by increased costs.
To bring this back to database-land, I'll reiterate this: if
more than about 10% of your db queries are writes, you'll suffer
significant performance degradation if you use RAID 5.
But RAID 5 does have maximum *cool* factor... :)
And, yes, you can certainly hot-swap in a RAID 0+1 environment.
I've done it, using SEVM with disks in multipacks.
From: maxfield%one.ctelcom.net@Internet on 03/19/99 09:29 AM
To: Derick H Siddoway/TC/TRS/American Express@AMEX
cc: mysql%lists.mysql.com@Internet
Subject: Re: fastCPU vs moreRAM
{snip}
It really is up to the final customer. I prefer raid 5 over mirroring
for maximum up time in the commercial products I've seen.
