UNIVERSITY OF CALIFORNIA, IRVINE
This administrative supplement award provides funding for additional equipment to facilitate and speed up research in haplotype analysis over general pedigrees via distributed processing on a grid of computers. Preliminary work conducted under the parent grant has yielded highly promising results, but at this point further progress is limited by the availability of computational resources.
Additional equipment will be used to extend an existing, small-scale prototype system for distributed haplotype computation. Having more computational power will allow a thorough theoretical and empirical investigation into the challenge of creating an effective parallel scheme for linkage and haplotype analysis. The central issue is scaling: how algorithms can exploit those resources as effectively as possible, to yield as high speedup as possible. We identified several parameters that control our parallel scheme and govern the tradeoff between parallel time and resources. Our initial results are promising yet their scope is limited by our small number of machines.
The computational resources we hope to get through this supplement will greatly enhance our ability to open up an extensive investigation and reach conclusive results at a faster rate. We hope to conclusively determine the interaction and the impact on performance of those parameters which we identified, eventually facilitating the anticipated move to massively parallel computation using thousands of computers over the Internet.
Finally, a public Internet portal will be established allowing interested professionals submit their pedigree problem instances for solving and have the most likely haplotype configurations presented in a transparent and user-friendly way. The additional equipment will constitute the computational basis for such a platform and guarantee satisfactory performance.
A small fraction of the budget will be allocated to machines having extensive external memory. These will facilitate experiments with memory intensive algorithms for genetic linkage analysis which currently develop.