H. Fröning, H. Litz, and U. Brüning (Germany)
High Speed Networking, Network Interface Architecture, Low Latency Message Passing, Fine Grain Communication
Cluster computing is still the most cost-effective solution to meet the increasing demand for computing power. Clusters are typically based on commodity computing hardware with specialized interconnection networks (IN). These cluster interconnects differ from commodity networks by higher bandwidth, lower latency, lower CPU utilization and improved scalability. But even with these sophisticated INs the latency of a message transfer between two nodes is still decades higher than a local memory access. Especially for fine grain communication the latency of a message transfer is crucial. An analysis of the latency shows that the main component originates from the I/O system. The goal of this paper is to present a new mechanism called Ultra Low Latency Message Transfer (ULTRA), which allows message passing with lowest latencies possible. Beside the usage of well-known techniques like User-Level Communication this work focuses on improving the Network Interface by an optimized and most efficient usage of the I/O system. The ULTRA mechanism and architecture presented here show a topmost optimized approach for low latencies, limited only by the used standard I/O system. With it a much closer coupling of the cluster nodes is possible and fine grain communication schemes are more suitable for cluster computing.
Important Links:
Go Back
