Re: DB_BLOCK_BUFFERS * DB_BLOCK_SIZE

SDC. Make sure you keep some memory for the Shared Pool and Java Pool, and that there is enough space for the Unix Buffer Cache (generally 10%) and for the Shadow Processes from your Oracle Clients.

After this, you are looking at database recovery time, checkpoint time, etc... The bigger the Buffer Pool the longer recovery and checkpointing will probably be.

You might not benefit from the 100,000 blocks since the database is not that big. If you only have a hot spot of 50Mb of data and the rest is read once and never again, why bother trying to keep this in memory! Remember with o8 that you can have multiple buffer pools, KEEP / REUSE and DEFAULT. When an object is assigned to the KEEP pool, you are saying please CACHE this table as much as possible. When an object is assigned to the REUSE pool, you are saying age this as soon as you want to, I don't want this data again.

What I am trying to say, is that a single large pool can be difficult and costly to manage, especially when things go wrong. (Instance CRASH)

Andrew

SDC wrote:

> Sun Solaris E4000 (Solaris 2.6) w/ 1.5gb RAM
> Oracle 8.0.5
>
> DB_BLOCK_SIZE: 8192 (8K)
> DB_BLOCK_BUFFERS: 100000
> Question:
>
> If this machine is dedicated to Oracle, why wouldn't I want to set the above
> parameters which would allocate around 800mb for SGA? The previous setting
> was 5,500 block buffers which works out to 44MB and the system was pitifully
> slow...
Received on Mon May 24 1999 - 19:14:38 CDT