Application-Specific Microprocessors

ing line speeds, executing the many different networking protocols is becoming the main bottleneck in high-speed communications. Gigabit Ethernet is already available, and products exist for 10-Gbps transfer rates. New band-width-eager software applications and faster processors in desktop and server systems place enormous demands on the current networking infrastructure. As a rule, network bandwidth use doubles every four months. Moreover, guaranteed quality and priority customization are on the way for many data, voice, and video applications. To meet these stringent processing demands, design-ers can either use an ASIC to create a custom solution for their application or use a commercially available net-work processor.1,2 The ASIC approach can achieve very high processing speeds but is inflexible, because changes in chipset behavior are either impossible or have limited support. This inflexibility is a significant problem, because applications and protocols continually evolve and extend to meet user-desired functionality. In addi-tion, deploying an ASIC-based design requires more time than deploying a network-processor-based one. Commercial network processors such as Intel’s IX