Re: primary and foregin keys
Date: 29 Dec 2004 15:59:07 -0800
Message-ID: <1104364747.512095.29190_at_f14g2000cwb.googlegroups.com>
>> it is a DB2 trick and
needs to be used with caution.
...<<
yep. One guy kept typing UNION instead of UNION ALL and complaining about poor performance...
>> Teradata got parallelism by using hashing -- one hash bucket gets
a processor and hash buckets are a partitioning enforced by
mathematics. That works because the engine is doing the partitioning
at the PHYSICAL level without a human being to screw it up
...<<
so does DB2, they both implement "shared nothing" approach
>> The goal of a partitioning like this is parallelism
...<<
not only. There are some tricks with different clustering indexes, such
as fiction books ordered by author's last name, but non fiction ones by
subject, then author's last name. That's how some public libraries
store books on their shelves, and for a good reason, this clustering is
best for most searches such as
- some mystery by Agatha Christie
- something on hiking in Arizona
so on
And yes, there are views and instead of triggers on top of that, so the
developers don't need to know that.
What do you think?
Received on Thu Dec 30 2004 - 00:59:07 CET