Re: RAC and undo tablespace

On 2006-03-08, Joel Garry <joel-garry_at_home.com> wrote:
> wagen wrote:
>
>>Why are separate UNDO tablespaces required for each instance in a RAC
>>cluster?
>
> Not knowing anything about it, I'll throw out some guesses anyways,
> least to most likely:
>
> Instance crash & recovery: So you don't affect the other instances
> multiversioning.

        So? Just associate undo segments with threads. That's all you're really doing with the separate undo tablespaces. Now there's the issue that even if you could share undos you probably would still want to segregate them since undos tend to be an IO bottleneck.

>
> Performance & Scalability: some OLTP db's do more i/o to undo than any
> other TS.
>
> Different tuning for instances: OLTP on one and DW on other could have
> quite different requirements for undo.
>
> Inter-instance communication: reduce having to get data from other
> instance in general for read consistency. RAC shares database files,

        RAC needs to worry about that anyway doesn't it? One of the instances (for whatever reason) could be trying to do some sort of recovery and accessing another threads undo (wherever it resides).

> but each instance has to handle the buffering of that data. There's
> generally enough talk going on just to resolve who owns what anyways,
> you probably don't want to add undo to that too.
>
> jg
> --
> @home.com is bogus.
> http://orcasthoughts.blogspot.com/2005/07/new-oracle-packaged-mini-triggers-with.html
>

-- 
     If you think that an 80G disk can hold HUNDRENDS of           |||
hours of DV video then you obviously haven't used iMovie either.  / | \

 Posted Via Usenet.com Premium Usenet Newsgroup Services
----------------------------------------------------------
    ** SPEED ** RETENTION ** COMPLETION ** ANONYMITY **
----------------------------------------------------------        
                http://www.usenet.com