Re: webmail user mailbox schema design?

From: Tegiri Nenashi <TegiriNenashi_at_gmail.com>
Date: Fri, 1 May 2009 13:55:57 -0700 (PDT)
Message-ID: <dd2fa48b-d749-4318-93d8-515a26f252af_at_c18g2000prh.googlegroups.com>



On May 1, 12:01 pm, Michael Austin
> You, my friend have obviously only ever worked on very small databases.

Define "small"...
"Terradata" -- a spectacular name 10 years ago sounds pretty ordinary today.

> In a DW enviroment, you would be insane to not use partitions.

Ah, the one that stubbornly resist indexing. Why indexing by the record time is such a bad idea? For example, why table Sales partitioned by the TransactionTime column can't be index organized?

> When you need to query with this much data, being able to restrict your
> search to a single or even a couple of partitions - you WILL become a
> believer in the use of partitions.

Again, from query access path perspective partitioning looks like rudimentary indexing. Because everything that partition pruning does index range scan does better. Yet, implementation of this rather unsophisticated idea of chopping one table into many is not stellar: I have seen optimizer failing to do partition pruning about as often as it failing to find an indexed access path. Received on Fri May 01 2009 - 15:55:57 CDT

Original text of this message