Section: Research Program
Axis 3: Parallelism
This third axis investigates a supplementary way to increase performances and scalability of genomic treatments. There are many levels of parallelism that can be used and/or combined to reduce the execution time of very time-consuming bioinformatics processes. A first level is the parallel nature of today processors that now house several cores. A second level is the grid structure that is present in all bioinformatics centers or in the cloud. This two levels are generally combined: a node of a grid is often a multicore system. Another possibility is to add hardware accelerators to a processor. A GPU board is a good example.
GenScale does not do explicit research on parallelism. It exploits the capacity of computing resources to support parallelism. The problem is addressed in two different directions. The first is an engineering approach that uses existing parallel tools to implement algorithms such as multithreading or MapReduce techniques [4]. The second is a parallel algorithmic approach: during the development step, the algorithms are constrained by parallel criteria [2]. This is particularly true for parallel algorithms targeting hardware accelerators.