Re: Basic question on RAID array / Tablespace configuration.

"TR" <tman_at_tman.dnsalias.com> wrote in message news:<6t%Q8.26123$XF6.3372526731_at_newssvr10.news.prodigy.com>...
> Have an approx 50 GB database (that is data+indexes) that will be used for
> aggregation queries and other OLAP type of stuff. E.g. write performance
> next to irrelevant, massive sequential reads from index and tables, I guess
> not a whole lot of probe-type random reads. Beware some of the queries do
> heavy writes to TEMP space.
>
> Hardware available is 8 80GB drives. Loss of data in the event of a drive
> loss is of very little concern.
>
> Thoughts?:
>
> -> Stripe (RAID0) all 8 drives, then logically partition into Data, Index,
> Temp, etc.
> -> Stripe (RAID0) 3 drives for Data, 3drives for Index, 2 for Temp. E.g.
> ensure that index and data are on separate physical devices.
> -> Any better configurations?
>

This a good way to get started. Sequential access is what OLAP is about so it definitely doesn't hurt to be thinking about it and tweaking your hardware in that direction. Whether it can be done depends on a number of things. The larger the database the more you can benefit from doing so. 50GB is quite small by OLAP standards. Other posters already pointed out another thread that touched upon separation of tables and indexes so I don't want to get into it again.

More important factors are the database design and the OLAP tool you choose. A properly designed dimensional schema is more flexible than a relational schema, and is almost always easier for you to run analytical queries. On the OLAP tool, Microsoft Analysis Services would be a good bet for medium sized databases. A good decision made on these issues will go a long way.

> Thanks,
> TR.
Received on Thu Jun 27 2002 - 00:27:01 CDT