Accomplishments

The LTC has had a very good second year vastly different than the first year.
Specifically : BU LTC Website

IBM Power5 development environment – the p720 running Redhat Linux (RH purchased for $60) is being configured to allow web users to port their code from other platforms to the p5. This is for the IBM p5 development team, lead by Robert MacFarlan, for providing 10 p5 720s. Our graduate student is developing this environment on Plone CMS. Visit PPC to see the progress.

Linux Technology Center Website (http://bultc.binghamton.edu) - Two iterations have been completed. The final site graphical design was provided by Ashok Subramanian, Assoc. Director of Research Advancement. The system is running on Redhat. Having the site complete was a pre-requisite for the proposed BU LTC conference and the LTC renaming contest.

Pavithra Mahamani – has been our graduate student developer since 1/2007. She has installed, configured and set up the p5 720, Redhat and the LTC website. We meet two to three times a week for updates and support. Pavithra has moved on over the 2007 summer and the LTC’s new graduate student is Sandesh SK. Sandesh has been working on the LTC website and Plone site, to get them updated and available.

Bio-Engineering IBM p720 – the LTC worked to provide Craig Laramee and his graduate student, Heika Sichtig with an IBM p720 with Redhat installed. The 8-way p720 provides the performance necessary in this highly computational centric environment ( The previous system was a small Intel box). According to Heika, “A multi-processor Linux system is highly desired compute environment for running time costly experiments efficiently”. The LTC installed, configured and debugged the p720 and Redhat for the Bio-Engineering team. They utilize the Linux server to run spiking neural network simulation with a computing expensive genetic algorithm to explore various parameter sets. The genetic algorithm usually employs an x-length chromosome to search through complex multi parameter space with the Darwinian model in mind. For example, a spiking neural network may be comprised of a simple 2 neuron - 5 synapse topology that requires certain weight and delay parameters for each synapse within certain ranges. In the most simple scenario, 10 parameters, 2 for each synapse, need to be tuned. This can be accomplished by encoding these parameters into a chromosome of length 100 bits. Here, each parameter requires 10 bits, which allows a certain range of values to be explored for the desired parameter. Moreover, the search space expands easily to 2 power 100, which requires high performance computing to run several experiments in a reasonable amount of time.