Re: Cluster vs Unclustered

Charles McDonald <cmcdon12_at_ford.com> writes:

> Hey Konstantin
>
> Nastarovia ??

Spasibo ;-) (Vodka, Balalayka,

        Matreshka, Gagarin, Perestroika, KGB, Gorbachev, Eltsin, Big Mac)

>
> Anyway, You stated that maintanence of Clusters are a nightmare.
> Especialy in a large system. Can you please elaborate as these tables
> we will be creating will have a lot of activity in the form of mass
> additions and updates. Why are index rebuilds difficult?
Probably index rebuilding is not difficult . Building new index can be difficult for tables that have entries for only a few cluster keys (it will make a full cluster scan to find just a few records)

>
> What type of impact does a cluster have on the mass addition of data?
>

If add records for a given key into say ten tables in the case of cluster you modify only one block in the datafile and one block in the index. This will turn out into ten data blocks and ten blocks for index. Don't forget to add rollback and redo logs here.

Why it is actually a nightmare. Imagine that you got 'Oracle block corrupted' in one of your tables. In this case you only have to export data from it and then import it back. Then you build necessary indexes.

If it is a cluster you will have to export all data from all tables, then recreate the cluster and build all indexes for all tables. It will take much more time (x10 or x20 times).

You probably can afford 20 minutes downtime, but 5 hours is not acceptable.

--
Sincerely Yours, Konstantin Kivi, Russia, konst_at_sirena.rinet.ru Received on Fri Jun 18 1999 - 08:20:50 CDT