This section of the directory pages is a look at the history of Stanford's directory project, in order to document what we have done, and why we have made the choices to do things in this manner. Portions of this documentation will likely be useful to other who want to set up their own OpenLDAP servers, but please remember that portions of our process entirely reflect Stanford's technology needs, and may not be applicable to other locations.
History
In May 1996, Stanford rolled out a Michigan ULDAP based directory service. It was used for rerouting email to @stanford.edu addresses, where previously, everything was @leland.stanford.edu for these systems. This service was very unstable, however, and so Stanford continued to research other options for its directory services.
In 1999, Stanford rolled out a Netscape based directory service. It was deployed on Sun Ultra 10's, and was used primarily for 3 things:
- A way to enforce privacy rules on Stanford related information via the use of Stanford.Who
- A method of allowing our Web Authentication methods to query a directory server for further information if needed, that could also be used for access restriction
- A location to store mail delivery locations that our mail routers could query
The directory project initially had 5 servers -- One master, four replica's running on Sun Ultra 10's. Over time, the needs of the mailrouters drove this up to 10 servers (Running on Sun E220R's and Sun Netra 1405's) -- One master, 9 replica's, 6 of those dedicated entirely to the mailrouters.
One thing we noticed over time, was that the stability of our Netscape directory service was decreasing. We were having steady corruption of our databases, which would then require the rebuild of the affected server. If the server happened to be the master, it required the rebuild of the entire set of directory servers. The reload process could take up to 22 hours per server, which resulted in up to a week of time being dedicated just to bringing the service back up.
Because of these (and a few other) issues, it was decided that we would start researching a new directory technology to deploy at Stanford.
