New architecture to reduce I/O bottlenecks in high-performance systems

Reviews large-scale application input/output requirements that are driving the need for high-performance distributed hierarchical storage. The focus is on evolving, high-performance, network-based storage-system architectures. The authors discuss some ideas about paradigm shifts that provide perspective on what is happening in system architectures; they outline, using examples, application storage and I/O requirements; they review relevant historical trends in storage-system architecture and their limitations; and they outline the IEEE Mass Storage Reference Model, focusing on the experience and issues in the use of network-connected devices and third-part protocol architectures to reduce the overall I/O bottleneck in distributed system environments.<>