We present and evaluate an architecture for highthroughput pattern matching of regular expressions. Our approach matches multiple patterns concurrently, responds rapidly to changes in the pattern set, and is well suited for synthesis in an ASIC or FPGA. Our approach is based on a new and easily pipelined state-machine representation that uses encoding and compression techniques to improve density. We have written a compiler that translates a set of regular expressions and optimizes their deployment in the structures used by our architecture. We analyze our approach in terms of its throughput, density, and efficiency. We present experimental results from an implementation in a commodity FPGA, showing better throughput and density than the best known approaches.