Low-power cluster using OMAP3530

Performance is always a delicate matter for algorithms to function properly in real-time or any other critical scenario. Depending on the application complexity, the processor power from a single core may not be enough to accomplish all the desired tasks; in order to surpass this boundary without multi-core solutions, it is only natural to consider the use of more than one machine, which is widely known as clustering. This paper intends to discuss a cluster built at the “Universidade de Sao Paulo - USP” using Beagle Boards, a Texas Instruments OMAP3530 processor based board, by describing the steps involved in its assembly, as well as presenting results of a parallel algorithm running on it.