Re: How to "Defragment" Tablespaces ?

> Good for you. You spread out IO among multiple drives and increased throughput,
> just like I suggested. So, what about it? Are you suggesting that there is some
> significance to the fact that they were all indexes? Time for a story...

Thanks for proving my point and I didn't need any old uncle Olaf stories to help me out...

This is the whole idea! If my data access requires both data from the table and the index, then by spreading the I/O among multiple drives increases throughput. If you put both the index and the table on the same drive, then this "can" hamper throughput. Erich just proved this point!!!! And you know what? If I use partitioning and split the table among 5 different drives and the index among another 5 different drives, then I'll get even better throughput!! How can you argue against that? Matter of fact, you can't since you just said "You spread out IO among multiple drives and increased throughput". The only way to *guarantee* that your tables and indexes are spread among multiple drives to increase throughput is to put them in separate tablespaces!

And your uncle increased throughput by decreasing demands for the same resources (i.e. mutliple highways). But I hope that you weren't saying that the blue trucks were tables and the yellow trucks were indexes. In that case, you are mixing metaphors because in the end, they are both trucks (color making no difference) where tables and indexes are completely different animals. They don't behave the same regardless of which color you paint them. It would have been better if your analogy had two different types of vehicles like a sports car and a moped. Putting them on the same highway will only cause trouble, but segregating them can improve life for everyone!

Just my 3.14159265 cents worth,
Brian

-- 
========================================
Brian Peasland
Raytheons Systems at
  USGS EROS Data Center
These opinions are my own and do not
necessarily reflect the opinions of my 
company!
========================================