fpga acceleration of information management services.docVIP

FPGA Acceleration of Information Management Services Richard W. Linderman, Mark H. Linderman Air Force Research Laboratory Information Directorate {Richard.linderman, Mark.linderman}@ Chun-Shin Lin Electrical and Computer Engr. Univ. of Missouri-Columbia linc@ ABSTRACT Field Programmable Gate Arrays (FPGAs) are widely known for their ability to accelerate “number crunching” applications, such as filtering for signal and image processing. However, this paper reports on the ability of FPGAs to greatly accelerate non-numerical applications, in particular fundamental operations supporting publish—subscribe information management environments. The specific core service accelerated by FPGAs is the brokering of XML metadata of publications against the XPATH logical predicates expressing the types of publications that the subscribers wish to receive. The acceleration is not achieved solely by the FPGA, but by its close coordination with a programmable (XEON?) microprocessor within a Heterogeneous HPC architecture (HHPC). Two subtasks addressed by the FPGA are the parsing of the ASCII XML publication metadata into an exploitable binary form, followed by the partial evaluation of up to thousands of subscription predicates, with results reported back to the programmable processor. Using the FPGA, parsing times are reduced to 20 microseconds, a 100-fold improvement over the microprocessor. After parsing, all predicates are evaluated simultaneously using combinational logic while only consuming a small percentage of a 6M-gate FPGA for 2000 predicates. Recently, the brokering implementation has been enhanced to support 2-bit logic encodings of clauses to more completely evaluate the XPATH predicates within the FPGA and reduce the burden on the programmable processor at the expense of some increased FPGA resource consumption. Another recent advance has been to allow incremental design on the FPGA to allow a mix of stable and volatile predicates to be accommodated in separately