I needed a solution for an Intranet project which requires fast processing of relatively large matrices (millions of cells); my language of choice is PHP. I discovered Octave, which is positively brilliant when it comes to the actual processing, but I wasn't happy with the time it took to start each Octave instance and to load the data files (0.3-0.5 seconds on my server), so I decided to write a network daemon which keeps a few Octave instances running at all times (and each Octave instance only loads the data once, so I'm slashing both the startup time and the loading time).
The resulting code is licensed under GNU AGPL, and lives at https://github.com/Gutza/octave-daemon (homepage and documentation at http://projects.moongate.ro/octave-daemon/). You can use the daemon regardless of the programming language you're using (simply connect to the server and issue plain ASCII commands). If you're writing PHP, you can use class Octave to either connect to the daemon or to control ad hoc Octave instances.
The project is currently in pre-alpha phase -- you can already start using it, but you shouldn't rely on it in production; the status will be updated on GitHub and Freshmeat, if you want to keep an eye out (http://frme.at/p/M0Q4p). I'm actively working on it these days, and would appreciate any feedback and suggestions. Contributors always welcome; also, it would be great if anyone was interested in helping with writing the documentation.
Regards,
Bogdan Stăncescu
--
