Larry.org

Distributed, or Peer to Peer, Applications still have lots of room for improvement because currently none of them really demonstrate enough business value to unseat the Client Server model. There's some people that would very much like to change that of course. Here's an overview of why we need a better solution and a link to one company that's trying to change it now.


There have been two proof of concept applications that are 1) looking for signs of extraterrestrial life and 2) gathering data to figure out how to fold proteins. Both of which are worthy and interesting in their own right and certainly in dire need of spare CPU cycles. Thing is, both these efforts rely on custom developed data processing clients. In the two examples above I have to install SETI at Home or Folding at Home on my computer to start receiving, processing and returning data.


I've heard these data processing clients called P2P applications but for large data processing problems like this there is a centralized server or server farm controlling the data flow so to my mind P2P doesn't seem right.


Thing is that many projects that could benefit from picking up CPU cycles from a network have more complex data processing requirements than either of these two shining examples. These two examples don't demonstrate that we can apply a distributed application, they demonstrate we can distribute the data processing associated with a client server application. To put it more simply, if the data needs to be processed in many different ways how would a distributed data processing application know how to chose which way to process the data. The answer is to build a distributed application that can not just process finely chopped up data sets but can also apply rules and conditions to how and why it will process that data.


Neither SETI at Home nor Folding at Home provide any sort of demonstration of applying businesses rules. Green Tea Technologies, Inc. Has some interesting ideas and apparently a working solution that does. Could be real interesting to see who picks it up. Certainly the BioInformatics crowd could stand to benefit from a solution like this that is more application oriented and less data processing oriented.


Thanks for the link Mariane!