Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: What is 'Shared Disk (Oracle)' and 'Shared Nothing(DB2)' ?

Re: What is 'Shared Disk (Oracle)' and 'Shared Nothing(DB2)' ?

From: Mark D Powell <mark.powell_at_eds.com>
Date: 11 Feb 2002 06:23:01 -0800
Message-ID: <178d2795.0202110623.1d886c8a@posting.google.com>


xtanto_at_hotmail.com (Krist) wrote in message news:<cb48a3b.0202110336.3ac2db9_at_posting.google.com>...
> Hi,
>
> I read that DB2 is based on 'Shared Nothing' approach while Oracle is
> using 'Shared Disk'.
>
> Could somebody please explain : What do they mean ?
>
> Thanks,
> Krist

Krist, it sounds like you are discussing the Parallel schemes of Oracle vs DB2, which I did not know had a parallel version though I guess someone could also have used these terms when describing or contrasting Oracle OPS/RAC to a distributed system.

Anyway with Oracle Parallel Server, OPS, and its decendant Real Application Clusters, RAC, multiple Oracle instances share access to one physical database located on sharable disk, that is, clustered disk.

With a share nothing system also known as massively parallel processing, MPP, disk units are allocated to CPU's and any request to read from that disk have to be routed through that CPU. In an MPP environment each CPU has dedicated memory. The CPU do not share memory or disk hence, share nothing.

The MPP setup can be contrasted to the more common SMP configuration on most multi-cpu computers where each CPU has access to all of memory and all the disk, or share everything.

And prior to version 9 Oracle did have an MPP version for MPP machines; Oracle may still be supporting this version going forward since it is for a specific type of hardware platform.

If the terms were used to constract a distributed approach to Oracle OPS/RAC then with a distributed approach you create several separate databases usually on different machines. Each of these databases probably constains the objects for specific applications. When data is needed from several applications in one query (or job task) then distributed queries are used to process the data. Oracle also supports this arrangement.

HTH

Received on Mon Feb 11 2002 - 08:23:01 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US