On Thu, 2007-02-22 at 18:48 -0500, Brandon Aiken wrote:
> Digg and Slashdot use MySQL databases, so clearly they *can* be made to
> support a high-load, high-performance, limited-write style web
> application.
>
> You might remember a few months back when SlashDot had to turn off
> threaded replies because the schema for the parent-child field was still
> an UNSIGNED INT4 instead of an UNSIGNED INT8, and they reached the
> maximum value of the field (16.7 million). Obviously, I have no
> knowledge of the server configuration, hardware configuration, or
> schema, but in-the-wild examples of high performance MySQL installations
> are trivial to find (as are PostgreSQL installations such as the .org
> DNS TLD root).
Actually this has been mentioned before, Slashdot is a good example of
how poorly MySQL scales. Almost every page you view on slashdot is
actually a static page harvested every x minutes by another process
because dynamically generating those pages is very expensive.
If slashdot was able to run on top of MySQL without all that hand
holding, then it would be a showcase for it.
> I'd like to see a tuned MySQL vs a similarly tuned PostgreSQL system
> (that is, fsync in the same state and with the same level of ACID
> compliance) subject to a battery of test schema types (OLTP, OLAP,
> etc.).
Me too. But I gave up on using MySQL for serious uses some time ago
when I realized that updates for serious bug fixes took years to come
out of the pipeline, and some will simply never be fixed.