Linux applied to real time and distributed systems

Commodity hardware running the Linux operating system will serve various roles in the next few years at Fermilab. We discuss several applications of Linux to distributed systems with real time or other performance constraints. First, we describe the level-3 trigger of the CDF experiment. In Run II, starting in the year 2000, this trigger will consist of hundreds of Intel-based PC´s and will be responsible for filtering events arriving via ATM at up to 250 MBytes/sec. The events will be distributed via fast ethernet to subfarms providing at least 45000 MIPS of computational power for the online reconstruction codes. Second, we discuss investigations of commodity hardware applied to lattice-gauge calculations, requiring hundreds of systems coupled via very low latency, high bandwidth networks. Third, we describe a high performance shared file system running on a cluster of Linux systems attached via fibre channel to a common disk pool