NoSQL vs SQL

[Rich Braun]

As someone noted, big sites like Facebook started with hundreds or thousands
of users, growing over time to reach 5 or 6 orders of magnitude bigger.  If
you want to compete with one of those guys today, you have to start out with
the capacity to handle a lot more than a few thousand.  But meanwhile the
breadth and maturity of open-source tools available to tackle those challenges
has grown by about the same exponential amount.

With some savvy using APC cache, memcached, and MySQL query-caching, you can
easily hit 100,000 queries per second on a single back-end server.  With
replication and load-balancing, you can (by hiring a team of good systems
engineers) horizontally scale to even today's Facebook load (500 million+
users) with the same software that comes on a stock Linux distro.

Use the good stuff.  Don't settle for hobbled software.  Among the good stuff
are the technologies mentioned above: APC, memcache, MySQL 5.1, the Linux 2.6
kernel, WordPress, Drupal, Confluence, etc.  (There have been some debates in
the past here on the BLU list between MySQL and PostgreSQL--I know MySQL can
handle these high data read rates, and some very capable data-write
rates--will let others comment about the alternatives.)  What makes these
technologies good is they've been put to the test over time and been tightly
optimized to deliver exceedingly high transaction rates out of RAM, without
crashing.

-rich