hyc's Journal: Fedora-what?

On the OpenLDAP Project we've taken a lot of flack about having outdated, clunky, badly performing software for a long time, thanks in large part to the obsolete releases that RedHat and other distros bundled in their offerings. People have said that RedHat acquired the old Netscape assets because OpenLDAP wasn't adequate for their needs. Of course, those people have no idea what they're talking about; OpenLDAP slapd was massively refactored in OpenLDAP 2.1 over 6 years ago, yielding a couple orders of magnitude improvement. But the RedHat faithful never saw it because RedHat continued to ship the clunky old OpenLDAP 2.0.27 release for so many years.

For the past few years OpenLDAP has been the fastest, most scalable directory server on the planet, bar none. Nobody else comes close. The benchmarks we've been conducting since 2005 all bear this out. Over the past few days I ran another set of tests against current versions of OpenLDAP (2.3.34), FedoraDS (1.0.4), OpenDS (0.1-34) and ApacheDS (1.0.1) just to clear the air.

Most of the data is posted right now on http://www.connexitor.com/blog/

At it's best FedoraDS is 3.5 times slower than OpenLDAP. Not even in the same league. For example, authentication rates on a database with 250,000 entries:

• OpenLDAP: 15907 auths/sec
• FedoraDS: 4526 auths/sec
• OpenDS..: 4178 auths/sec
• ApacheDS: 632 auths/sec

So, to the decision makers at RedHat, who didn't take my advice to work with the OpenLDAP Project and decided to spend millions of dollars on obsolete technology instead - nice job. You've helped prove that a well written Java program (OpenDS) can perform as well as a poorly written C program. Did you get a free bowl of soup with that hat?