Ushahidi Mailing Lists

I used to do Linux capacity planning for a living and my guess is that
we're looking for the simplest thing that will work. In any event, the
first order of business is usually defining what we mean by a "slow
system". Is it not responding fast enough under load? Or is it not
producing enough throughput? As a general rule of thumb, more
throughput means slower response.

Once you've defined the criteria, the next step is profiling. Are you
processor or I/O bound? You determine this using "iostat" or "sar",
both of which are in the "sysstat" package. If you're processor-bound,
you can focus your investigations with "oprofile". If you're I/O
bound, "iostat" will tell you a lot, and if you want more detail,
there's "blktrace".

If someone has an Ushahidi system that's slow, I'd recommend
installing "sysstat" and collecting some data before throwing rocks at
the database, PHP, or the architecture.
-- 
Twitter: http://twitter.com/znmeb Computational Journalism Studio
http://j.mp/CompJournStudio

Data is the new coal - abundant, dirty and difficult to mine.