Re: Is Raid 5 really that bad for Oracle?

"omlet" <notrolls_at_notrolls.omlet.org.notrolls> wrote in message news:f22beb8f20202b298380f42c31b81a7f_at_localhost.talkaboutdatabases.com...
>
> I would suggest renaming RAID to RAED; as stuffing as many fiber spindles
> with the fastest 15Krpm disks ain't cheap!
> Ok may be a better name is DLDMNIT.
>
> PS. Just in case you wonder what is DLDMNIT; keep guessing? If you figure
> it out; you can redeem a prize!
>
> Hint: It has something to do with the Iranian Jewish girl that busted
> Daniel Morgans balls for not knowing "vi"! Call Netapp for a clue?!
>
> BAARF is a shameless gang who promotes self destructive not-so expertly
> hogwash.
>
> You would think that Oracle knows what is best for Oracle! What are those
> idiots take the world for?!

Er, that's the same company that invented PCTINCREASE and made it default to 50, and who then went on record claiming that having your segments comprised of a single extent would be good for performance. Their track record on these things is not as perfect as you are suggesting it ought to be.

But regardless... I don't recall Oracle ever recommending RAID-5.

Indeed, in the appendix E of their Performance Tuning course notes they still publish a table showing performance/safety scores of various RAID flavours (drawing heavily on the work of Cary Millsap in the process, who is one of BAARF's leading lights of course). That table shows RAID-5 in the middle of the pack for performance, but with quite good safety. But it also shows RAID1+0 with stellar performance and equally good safety. If Oracle recommends anything in their training documentation, therefore, it is RAID1+0.
> Yahoo! uses RAID4/5 technology; CISCO uses RAID4/5 technology, to just to
> name few?!

And Microsoft engages in monopolistic practices. So do we follow that example, too? Of course not: these issues are debatable on their merits, and the question of which corporation uses which IT technologies and whether that should be taken as some kind of recommendation is not a good point to start from. The average business using Oracle is not likely to find Yahoo's or CISCO's reasoning and cost/benefit analysis applicable to their own particular set of circumstances, after all.

However, you seem unable to debate these sorts of issues on their technical merits, but seek merely to reduce everything to personalities and cheap-shot ad hominem attacks... so I guess I've just wasted my time.

HJR Received on Mon Aug 02 2004 - 02:35:51 CDT