High-Performance Cluster Computing Using SCI

The Scalable Coherent Interface (SCI) is a recent com- munication standard for cluster interconnects. We study the use of SCI in a high-performance parallel computing setting, using a cluster of UltraSparcs connected via Dol- phin SCI SBus-2 adapters. We chose SCI as network fabric since it offers very low latencies and high bandwidth. In this paper, we study how to map a variety of programming models efficiently onto the SCI hardware. We focus on message passing and global address space support, implementing Active Messages and Split-C. We present implementation trade-offs and present performance measurements. We found that the user-level load/store programming interface of SCI is very convenient to use, achieves low latencies, and is fully virtualized, simultaneously supporting multiple parallel programs and communication channels. On the other hand, neither of the programming models studied maps directly to SCI. Issues such as notification, atomic operations, and virtual address space limitations represent major implementation challenges, which we address with a combination of compiler and run-time support. Overall, we found the SCI network to form a good substrate for high-performance cluster computing.