Utilizing Heterogeneous Networks in Distributed Parallel Computing Systems

Heterogeneity is becoming quite common in distributed parallel computing systems, both in processor architectures and in communication networks. For example, many distributed systems are now being constructed using a variety of different communication networks, such as Ethernet, ATM, Fibre Channel, and HiPPI, within a single system. In addition to this hardware heterogeneity, there is a heterogeneity in the types of messages produced by parallel application programs, such as short synchronization messages that require low latency, and large data transfers that require high-bandwidth, but can tolerate high start-up latency. Each of the different types of networks have different performance characteristics, while the different types of messages may have different communication requirements. In this work, we analyze two techniques for exploiting these heterogeneous characteristics and requirements to reduce the communication overhead of parallel application programs executed on distribute..