I didn’t agree with you first, but last paragraph makes sense for me

it is not that easy. Especially if functions or multiple columns are involved like in: SELECT * FROM tbl_name ORDER BY CONCAT(TO_DAYS(date_column), SUBSTRING(string_column, LOCATE(’substr’, string_column))) In order to merge all the results we would have to mimic mysql server logic (or we rewrite the query to include the order by conditions as hidden columns so we can use them for sort later on). Currently this is much better handled within your application. Remember that “full partition scan” is something you should try to avoid as much as you can anyway since it becomes quite a heavy operation. But you are right that simple ORDER BY could be integrated quite easy.

Greetings

Peter

currently LIMIT clauses are handled correctly (even though the natural order cannot be preserved due to the nature of having data spread along multiple tables). Queries that go to all tables (i.e. a full partition scan is performed) that contain ORDER BY are rejected. (Our application doesn’t use ORDER BY in full partition scans).

The only way to implement ORDER BY in these cases would be to perform the sort in the MySQL Proxy. I would rather solve this in the application layer, i.e. rewrite queries and application logic. Remember, HSCALE is not a “magical plug-in and forget” module that abstracts the MySQL Server in all its functions but a module to aid you writing application based partitioning.

Anyway, if you have any more ideas, please let me know!

Greetings

Peter

Just a quick question, how do you plan to distribute queries and combine results that contains ORDER BY and/or LIMIT statements ?

I’m curious because I’m working on something similar (written in Perl though).

Thanks

Didn’t even think of that. I hope this idea gets picked up and someone runs with it, in some shape or form! Thanks for getting the ball rolling!