High Performance Computing
Boston University's College of Engineering relies upon high-performance computing throughout its academic and research areas. Our facilities include an experimental CPU/GPU cluster, a GPU cluster, a CPU cluster, cycle-harvesting of six computer classrooms, three private clusters in three different research labs, and easy access to University-level and large HPC-center facilities.
All of these resources are tied together by the Eng-Grid, which provides single-sign-on access to all queues. All College of Engineering users, from undergraduate through faculty, are automatically granted access to the grid's public queues. This allows us to easily incorporate HPC methodologies into our curricula, which includes High Performance Computing with Multicore and GPUs, Parallel Computer Architecture, Applied Algorithms for Engineers, Numerical Methods and Modeling in Biomedical Engineering, and the GPU@BU Workshop.
Faculty and students in the College of Engineering are utilizing high performance computing in numerous research projects, including:
- Design and runtime management of manycore systems, Dr. Ayse Coskun
- Lattice methods for statistical mechanics, Dr. Richard Brower
- Configurable computing with FPGAs, Dr. Martin Herbordt
- Algorithms for large-scale micromagnetic modeling and molecular dynamics simulation, Dr. Roscoe Giles
- Cochlear and binaural hearing simulation, Dr. David Mountain and Dr. Steven Colburn
- Photon transport modeling in tissue for optical pharmacokinetics, Dr. Irving Bigio
Boston University is participating in the SC11 Student Cluster Competition. Half of the students on this team, as well as the team advisor, are from the College of Engineering.
Grid computing offers a model for solving massive computational problems by making use of the unused CPU cycles of large numbers of disparate, often desktop, computers treated as a virtual cluster embedded in a distributed telecommunications infrastructure. Grid computing’s focus on the ability to support computation across administrative domains sets it apart from traditional computer clusters or traditional distributed computing