Re: large database loading

From: Devan F. Dewey <dfd_at_epsilon.com>
Date: Tue, 2 Aug 1994 13:21:27 GMT
Message-ID: <dfd-020894092127_at_sebastian2.quantum.epsilon.com>


In article <94210.085735KJOHNSON_at_ESRIN.BITNET>, <KJOHNSON_at_ESRIN.BITNET> wrote:

> In article <7912_at_dsacg3.dsac.dla.mil>, nfm1556_at_dsacg3.dsac.dla.mil (Steven R
> Jackson) says:
> >
> >We are starting a new project to load an Oracle
> >database on a Sunsparc from flat files. The
> >design is being done in case tools which will generate the
> >ddl.
> >Has anyone any ideas on the quickest way to load very large
> >databases? Will loader handle this in a reasonable time?
> >
> >Could the 7.1 parallel option be of help?
> >
> The Parallel Query Option is described in the Summer edition of
> ORACLE MAGAZINE.
>
> The documentation for ORACLE7.1 (including Parallel Query Option) is
> available on ORACLE PRODUCT DOCUMENTATION LIBRARY CD (V 1.0.4 May 1994)
> svadd.obd - ORACLE7 Server Documentation
>
> The Parallel Query Option CAN improve SQL*Loader performance
> as well as index building performance.

Oracle 7.1 also supports a direct path load option.

SQL*Loader in conventional mode essentially uses insert statements of load data into your tables. This is relatively slow.

The driect path laod option pumps data directly into the database an entire block at a time. having a large block size helps this along. The drawback to driect path load is that recoverability from failure is low to non-existent.

Direct path loading is great for building large, new tables fast.

-- 
Devan F. Dewey
Business Analyst
Epsilon Data Management
dfd_at_epsilon.com
Received on Tue Aug 02 1994 - 15:21:27 CEST

Original text of this message