Jason R. White

An architecture for fast processing of large unstructured data sets

This paper presents a general system architecture tailored to perform searching, filtering, compression, encryption, and other operations on unstructured data streaming from a disk system. The system achieves high performance on such applications by providing for parallelism, hardware-application specialization and reconfiguration, and hardware placement near the disk systems. A limited prototype of a single compute node has been implemented and is described. The prototype is tailored to applications involving complex searching and its performance is compared to a pure software implementation having the same search capabilities. Performance is considered in terms of data set size, query string hit rate and query complexity. Performance results as a function of these parameters are presented and the results indicate that, for data set sizes above 1.4 MB, the prototype compute node is between one and two orders of magnitude faster than a pure software implementation. At high data set sizes, on an individual node, speedups of about 200 and a sustained throughput of 300 MB/sec have been achieved.

Massively parallel data mining using reconfigurable hardware: approximate string matching

Summary form only given. Data mining is an application that is commonly executed on massively parallel systems, often using clusters with hundreds of processors. With a disk-based data store, however, the data must first be delivered to the processors before effective mining can take place. Here, we describe the prototype of an experimental system that moves processing closer to where the data resides, on the disk, and exploits massive parallelism via reconfigurable hardware to perform the computation. The performance of the prototype is also reported.

Modeling the power consumption of audio signal processing computations using customized numerical representations

This paper explores the impact that numerical representation has on the power consumption of audio signal processing applications. The motivation is digital hearing aids, for which minimizing the power consumption is a critical design goal. We investigate two aspects of this problem. First, we evaluate the validity of using signal transition counts to model actual power consumption within this problem domain, and second, we compare the relative power consumption of multiply-accumulate operations for several customized numerical representations.