Re: Bad performance?

[Harrison Fisk]

Jacob Christensen wrote:
> Yes is was a bad select i posted. The select
> allways look like this:
>
> SELECT * FROM cologin WHERE user='XXXX' and
> k_id='XXXX';
>
> Explain from the select above is okay and confirm
> index is okay.
>
> I even tried making a index on all the fields with
> the same result.

You should ideally have an index across both (user, k_id). That will allow it to filter and return only the records in question. After adding the combined index, are you still only getting 5-6 q/s as you indicated in the first post?

>
> This select take 3-4 sec with no load on the
> cluster:
>
> SELECT * FROM cologin LIMIT 350000, 30;
>

Again this query is doing a full table scan. Check EXPLAIN.

If you want to do this in a more optimal way (in a method using indexes), you should look into doing something like:

SELECT * FROM cologin WHERE id > 350000 LIMIT 30;

Keep in mind these queries aren't identical, but might be possible get your application to work with it anyways.


One thing you are missing (from both the first and the second query) is that is the queries *need* to use indexes fully in order to be fast with cluster. Otherwise you end up pulling huge amounts of data across the network since the nodes are seperated.

Harrison Fisk, Trainer and Consultant
MySQL AB, www.mysql.com



Edited 1 time(s). Last edit at 12/07/2004 06:22PM by Harrison Fisk.