The architecture of supercomputers: Titan, a case study
The architecture of supercomputers: Titan, a case study
Theory, Volume 1, Queueing Systems
Theory, Volume 1, Queueing Systems
Models of machines and computation for mapping in multicomputers
ACM Computing Surveys (CSUR)
The performance of cache-coherent ring-based multiprocessors
ISCA '93 Proceedings of the 20th annual international symposium on computer architecture
A case for uniform memory access multiprocessors
ACM SIGARCH Computer Architecture News
Techniques for reducing overheads of shared-memory multiprocessing
ICS '95 Proceedings of the 9th international conference on Supercomputing
An extension to the SCI flow control protocol for increased network efficiency
IEEE/ACM Transactions on Networking (TON)
The GLOW cache coherence protocol extensions for widely shared data
ICS '96 Proceedings of the 10th international conference on Supercomputing
Efficient synchronization: let them eat QOLB
Proceedings of the 24th annual international symposium on Computer architecture
Proceedings of the 24th annual international symposium on Computer architecture
Performance Evaluation of Hierarchical Ring-Based Shared Memory Multiprocessors
IEEE Transactions on Computers
Performance Evaluation of the Slotted Ring Multiprocessor
IEEE Transactions on Computers
The Impact of Pipelined Channels on k-ary n-Cube Networks
IEEE Transactions on Parallel and Distributed Systems
Kiloprocessor Extensions to SCI
IPPS '96 Proceedings of the 10th International Parallel Processing Symposium
Scalability of SCI Workstation Clusters: A Preliminary Study
IPPS '97 Proceedings of the 11th International Symposium on Parallel Processing
Integrating reliable memory in databases
The VLDB Journal — The International Journal on Very Large Data Bases
A supercomputer system interconnect and scalable IOS
MSS '95 Proceedings of the 14th IEEE Symposium on Mass Storage Systems
| Hi-index | 0.01 |
The Scalable Coherent Interface (SCI) is an emerging IEEE standard that provides computer-bus-like services to a set of nodes via fast, unidirectional links. This paper presents the first detailed performance study of the SCI ring, using both analytical models and simulation. Performance is analyzed for uniform and nonuniform traffic, and the effect of the ring's flow control protocol is studied.The queueing model is based on an M/G/1 queue, augmented to include the effect of packet trains on the mean and variance of the source transmission time. The model is validated against simulation results, and shown to be quantitatively accurate for uniform workloads, and at least qualitatively accurate for nonuniform workloads. The flow control mechanism is shown to effectively prevent node starvation and reduce the ability of nodes to unfairly consume ring bandwidth, but at the cost of decreased overall ring utilization. The SCI ring is also compared to a standard bus, modeled with a simple M/G/1 queue, and shown to provide substantially higher throughputs and lower latency than a bus with realistic clock speeds.