List:General Discussion« Previous MessageNext Message »
From:Fred Lindberg Date:May 5 1999 10:40pm
Subject:Re: Design help
View as plain text  
On Wed, 7 Apr 1999 13:03:47 +0200, list@stripped
wrote:

>Received: (qmail 25056 invoked from network); 6 May 1999 00:18:36 -0000
>Received: from cal040031.student.utwente.nl (130.89.229.165)
>  by www2.analytikerna.se with SMTP; 6 May 1999 00:18:36 -0000
>Received: from list by Cal040031.student.utwente.nl with local (Exim 2.12 #1)
>	id 10Uq7X-0005wt-00
>	for mysql@stripped; Wed, 7 Apr 1999 13:03:47 +0200

Either your mail server has the wrong date/time or it is _very_ slow
...

>Now my question is, how efficient is connecting/setting up a statement for
>all those paths?

Given the problems you have, I would argue that your database setup is
not efficient. If you need to distribute, you should do it in a way
that each subprocess has the data it needs to do the job. Why not store
the data as trees? Or use a hash calculated on the entire name to
decide which of e.g. 53 databases the data is stored in. Then you've
divided the space into 53 areas.

Yes, CHAR is more efficient than VARCHAR, but unless I misunderstand
your paying a lot for it.

Passing so much info between processes is also inefficient.

So, set up the database with distribution in mind. Either, so that a
given question can be directly assigned to a host, or so that a given
part of the question can be given to a particular host that is known to
have the resources (data) to solve the problem and report back to the
gathering process.


-Sincerely, Fred

(Frederik Lindberg, Infectious Diseases, WashU, St. Louis, MO, USA)


Thread
Design helplist6 May
Re: Design helpFred Lindberg6 May