A low cost, low power, high scalability and dependability processor-cluster platform

Nowadays, more and more applications are computing-intensive, such as search engines, online game servers, or even fMRI, used in medical examination. Supercomputers are setup to meet the computing demand as well as huge data storage. However, some applications favor computing power much more than data storage. To emphasize on providing computing power with low cost, low power, high scalability and dependability, this work presents a platform which involves a group of computing clusters. A cluster consists of up to 12 processors. A computing job is distributed to a group of processors for speed-up. Moreover, an FPGA may join for further acceleration. A set of front-end servers is responsible for job dispatching and maintains reliability. The system details are described. Current status and ongoing work are also reported.