Re: Splitting query results among machines?

[Fred Lindberg]

On Thu, 11 Mar 1999 13:21:15 -0800 (PST), Brian Bray wrote:

>My question is how do I run a query and then reliably split the results of
>the query among multiple machines.  Say for example I do a SELECT * FROM
>Bookmarks.  I want machine 1 to handle the first 1/3 of the
>resulting rows, machine 2 the 2nd 1/3 and so on.
>
>Any good Idea's?

For ezmlm, a 53 way hash is calculated and stored with each address.
The load can then be split. You can to your data add a column with
either a hash or a random number (depending on if the number has to be
calculatable from the data), then into your select add e.g. WHERE hash
BETWEEN 0 AND 17, etc. As long as you cover the entire range without
overlap, you can split any way you want.


-Sincerely, Fred

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