Re: Multiple columns, huge number of inserts and selects - what to do?

From: <sybrandb_at_yahoo.com>
Date: 27 May 2003 03:54:49 -0700
Message-ID: <a1d154f4.0305270254.3cebc1b8_at_posting.google.com>


pgaur_at_cdotd.ernet.in (Praveen Gaur) wrote in message news:<781decc.0305262158.919ec0e_at_posting.google.com>...
> Hi,
> I'm in the very initial phases of designing a large database using
> OPS 9.1 on COMPAQ Alpha Servers. The requirements for the DB are as
> follows :
>
> 1. There will be a huge amount of data insertions required to be done,
> on the avg of 2.5 Gigabytes per hour.
> 2. All these inserts are of similar type and we need to do a lot of
> selects also from this inserted data later on after a short while (may
> be not that real time that as soon as the row is inserted, its made
> available for select immediatly, but whenever the select query is run,
> it should be fast).
>
> In the above process, I've found that at least 75 attributes are there
> which I need to insert into my DB. Almost all these 75 attributes will
> be used in seperate queries of selects later on.
> Now the question is - How should I design my database. I wish a very
> good performance of insertion (space isn't a problem, i've got 10 Tera
> Bytes to store), wish to make use of OPS to Maximum. Parallaly I wish
> to give a good performance for selects too, which may be made after
> some time of a row insertion. Should I create indexes? What all
> paramters should I take care while designing such a kind of DB?
> Thanx in advance
> Praveen Gaur

You should design a *datamodel* first, and the instance in a second phase.
You should NOT (I repeat NOT) take into account any hardware during design of the *logical* model.
As your question is much too vague (what is 'fast', 'very good'), and much too broad ('design a database'), I suggest you'd hire a consultant.

Regards

Sybrand Bakker
Senior Oracle DBA Received on Tue May 27 2003 - 12:54:49 CEST

Original text of this message