Distributed Computing

With the ubiquitous presence of the personal computer, the vast majority of the world’s computing power and disk space is no longer concentrated with a relative small number of supercomputer but with the hundreds of millions of personal computers currently in use. The underutilized resources of these PCs had become a valuable commodity considering the rising computational requirements of many research and development projects. One way to tap into these resources is with the use of distributive computing.

Areté has developed its distributive computing capacity with the help of the Berkeley Open Infrastructure for Network Computing (BOINC). BOINC is designed to take advantage of a computer’s processing power and results in an automated cyclical process between a project server and participating PC that is repeated indefinitely with little to no user involvement. This process, as shown in the figure below, consists of (1) the PC receiving instructions and user preferences as well as downloading any required applications and input files followed by (2) computing of work and, when finished, (3) the uploading of output files and reporting of results.

Distributive computing has already been teamed with a knowledge discovery and dissemination (KDD) algorithms for the detection of weak signals in large, noisy text databases in order to output sets of documents that suggest relevant, non-obvious "stories" for a human analyst to review. Unfortunately, relatively small increases in the size and depth of analysis of the databases can result in substantial increases in computational requirements. BOINC was employed to help meet these computational burdens and resulted in a tremendous reduction in overall computation time and the solving of problems which had never been attempted.

The cyclical process of the BOINC distributive computing framework.