Besides maintence advances partitions may help to balance the load among your
disks, reduce contention and add problems. For example if your have 10 disks
your should create 10 partions, one for each disks and 10 local indexes for
each disk. Next time a user makes a query Oracle reads the index only from the
disk containing specified id. Without partions Oracle will have to read the same
index disk for all users. So, using partions you improve response time (as I see
you have 100% OLTP and this is exactly what you need)
Contact me for further explanations.
> We are currently looking to create a very large table >100GB 100,000,000
> rows, the only key we have that is always specified in every query is a
> customer id of which there is a range 1-1000. We are looking to partition
> the table based around this key. Most of the queries will be trawling the
> whole customers data, although some will also specify a date range, which we
> can index locally. Not all partitions will be in seperate tablespaces. This
> is a reporting database with a large number of online connected users
> running canned queries.
>
> Questions:
> Is there a better way of doing this?
> What is the overhead of having a large number of parititions?
> Would standard index/clusters work better?
>
> Many Thanks
>
> Mike
Received on Wed Jun 16 1999 - 11:08:53 CDT
