Re: Two sanity checks

Dusan,

Regarding your first one. If I were you I'd push for partitioning of both the table and indexes as hard as I could. Especially when you consider what happens once it is OK for the old audit info to be deleted.

Commenting on your second one. You are correct they are wrong. I have had practical experience with this being told that the EMC disk took care of all this and no need for the DBA to worry etc. Then during a reorg with the help of a friendly UNIX admin moved two high transaction files so they no longer shared the same controller and wow what a performance increase. Eventually even the senior UNIX people admitted that the disk load was better distributed. Have also worked at a sight using HP SmartDisk system and that had to be the worst disk architecture I've ever had the displeasure of having Oracle on. What I tend to do these days is simply scale OFSA up to the controller level. This often leaves you compromising with regards to what goes where just like in the old days when disk was expensive.... you just have to prioritise what needs to be on it's own. Still can be really hard to convince people that your 700Mb of redo really does need it's own 18Gb mirrored disk set!

Kind Regards

Fraser McCallum
MVP Oracle Administration
www.brainbench.com

"Dusan Bolek" <pagesflames_at_usa.net> wrote in message news:1e8276d6.0202070251.73fa427e_at_posting.google.com...
> First One:
>
> There is a need for storing up to one terabyte of audit informations
> (not Oracle internal audit, just general one). Proposal is to use one
> single non-partitioned table with two indexes. One is on date and the
> second one is on system indentifier which is varchar2(256). Looks to
> me like a pretty crappy design and performance will be terrible.
>
> Second One:
>
> Some people says that OFSA (Oracle Flexible Storage Architecture) is
> obsolete, because new modern storages like SAN disk arrays with
> stripping, caching technologies etc. can spread I/O load with no
> support from special data placing.
> That sounds to me like (sorry for that word) a bullshit, because disk
> I/O is still very slow (comparing to memory, interfaces etc.) and
> having everything in one big mess can't bring the same performance as
> properly distributed files (OFSA) even with help of all stripes and
> caches in the universe.
>
> P.S. I love external consultants, especially from world-wide companies
> with all certifications and big bills.
> --
> _________________________________________
>
> Dusan Bolek, Ing.
> Oracle team leader
>
> Note: pagesflames_at_usa.net has been cancelled due to changes (maybe we
> can call it an overture to bankruptcy) on that server. I'm still using
> this email to prevent SPAM. Maybe one day I will change it and have a
> proper mail even for news, but right now I can be reached by this
> email.
Received on Thu Feb 07 2002 - 05:34:51 CST