Re: Comparison of PGSQL and DB2

On Tue, Mar 09, 2004 at 10:22:24AM -0800, Josh Berkus wrote:
> > User Defined Task Prioritization
> > (Ability to define rules on how the database should prioritize workloads
> either by query cost or user profiles. Ability to monitor resource usage and
> adjust the priority of queries or cancel runaway queries that exceed
> predefined limits.)
>
> This would be nice.   We're not sure how to implement it; don't
> expect it soon for PostgreSQL.

Note that we do have a part of this, if "runaway queries" == "queries
which run too long".  I suspect that what would really be needed,
however, is something running on the OS to detect timeslice or
memory hogs.

> > Parallel Backup / Restore (Ability to perform backup/restore of
> > large databases faster by leveraging
> all the processors in a multi-processor machine.)
>
> This is directly related to the above feature.

We can simulate it though, right?  Use the binary file format for
pg_dump, and hand out pieces of recovery to different restorers
depending on your number of processors, one table at a time per
restorer.  (Afilias -- Chris Browne, really -- wrote something along
these lines for the import of the .org database.  I talked about it
a little at OSCON last year.)  This is very far from the
user-friendly tool that IBM offers, of course.

A

--
Andrew Sullivan  | ajs@crankycanuck.ca