I have been speaking a bit with the authors of the core algorithm that I implemented in phpdfs and they directed me to some follow up research and papers that contain information on some scaling improvements.
If you are interested, you can read the papers here:
http://users.soe.ucsc.edu/~
http://www.ssrc.ucsc.edu/
these papers are great. I am just now getting into collecting empirical data about phpdfs to prove that the implementation does, in fact, do what it is supposed to do and the first paper above has a section on "RUSH in Practice" (section 3) which can serve as a reference when I do my own tests. sweet.
Since I am able to easily swap out the locating algorithm in phpdfs I think what I will do before I get into working on the improved implementation of RUSH is test what I currently have and develop a solid test suite that can be used going forward. I know that I should have more fully fleshed out tests before implementation, but I was just too damn excited about doing this. sue me. :P
peace, look for test results soon...
Thanks for this. It looks interesting. I'm printing off the docs now so I can read them on the train home.
ReplyDelete