The BlackWidow High-Radix Clos Network

Scott,; Abts,; Kim, John; Dally,

doi:10.1109/isca.2006.40

Cited by 123 publications

(57 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The second level of the Fat Tree uses 16-port routers with down channels only. The routers' operating frequency is set at 1GHz and the zero-load latency through each router is 26ns [24]. The network uses nearest-common-ancestor routing.…”

Section: Methodsmentioning

confidence: 99%

Network congestion avoidance through Speculative Reservation

Jiang

Becker

Michelogiannakis

et al. 2012

IEEE International Symposium on High-Performance Comp Architecture

View full text Add to dashboard Cite

Congestion caused by hot-spot traffic can significantly degrade the performance of a computer network. In this study, we present the Speculative Reservation Protocol (SRP), a new network congestion control mechanism that relieves the effect of hot-spot traffic in high bandwidth, low latency, lossless computer networks. Compared to existing congestion control approaches like Explicit Congestion Notification (ECN), which react to network congestion through packet marking and rate throttling, SRP takes a proactive approach of congestion avoidance. Using a light-weight endpoint reservation scheme and speculative packet transmission, SRP avoids hot-spot congestion while incurring minimal overhead. Our simulation results show that SRP responds more rapidly to the onset of severe hot-spots than ECN and has a higher network throughput on bursty network traffic. SRP also performs comparably to networks without congestion control on benign traffic patterns by reducing the latency and throughput overhead commonly associated with reservation protocols.

show abstract

Section: Methodsmentioning

confidence: 99%

Network congestion avoidance through Speculative Reservation

Jiang

Becker

Michelogiannakis

et al. 2012

IEEE International Symposium on High-Performance Comp Architecture

View full text Add to dashboard Cite

show abstract

“…The control dependency between stages on the pipeline structure can be relaxed by look-ahead routing, and speculation, thus achieving a further reduction of the switch delay [18]. Recently, high-radix low-latency switches with dozens of ports have become available, as seen in the YARC routers for folded-Clos or Fat tree networks [20]. Switch delay with a latency of 31.25 ns (800 MHz, 25 cycles) are reported to have been implemented, and a custom chip provides 40.1 ns for forwarding a short message on an Anton-2 supercomputer [2].…”

Section: Switch Delaymentioning

confidence: 99%

Optical network technologies for HPC: computer-architects point of view

Koibuchi

Fujiwara

Ishii

et al. 2016

IEICE Electron. Express

View full text Add to dashboard Cite

Optical network technologies, such as circuit switching, wavelength division multiplex and silicon photonics, have been considered for high-performance computing (HPC) systems to achieve low communication latency, high link bandwidth and low power consumption. However, conventional HPC systems still use packet networks with electric switches. Only active optical cables for inter-cabinet long links are borrowed from optical network technologies. This paper firstly reviews the gap between the conventional HPC networks and feasible optical network technologies. We explain our pessimism that this gap will continue to exist by the beginning of the post-Moore era, i.e. 2025-2030. It secondly illustrates our research vision that HPC networks will be able to adopt optical circuit switching, possibly using free-space optics in the post-Moore era.

show abstract

“…This triggered renewed interest in interconnection structures, in particular in those supported by high radix switches [20]. These switches, instead of providing only 8 to 24 bi-directional ports, pushed the port count to 64 [21] by 2006, and to more than 100 by today [22]. With such high port counts, more highly connected and compact topologies become feasible.…”

Section: Related Workmentioning

confidence: 99%

Design Methodology for Optimizing Optical Interconnection Networks in High Performance Systems

Rumley

Glick

Hammond

et al. 2015

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. Modern high performance computers connect hundreds of thousands of endpoints and employ thousands of switches. This allows for a great deal of freedom in the design of the network topology. At the same time, due to the sheer numbers and complexity involved, it becomes more challenging to easily distinguish between promising and improper designs. With ever increasing line rates and advances in optical interconnects, there is a need for renewed design methodologies that comprehensively capture the requirements and expose trade-offs expeditiously in this complex design space. We introduce a systematic approach, based on Generalized Moore Graphs, allowing one to quickly gauge the ideal level of connectivity required for a given number of end-points and traffic hypothesis, and to collect insight on the role of the switch radix in the topology cost. Based on this approach, we present a methodology for the identification of Pareto-optimal topologies. We apply our method to a practical case with 25,000 nodes and present the results.

show abstract

The BlackWidow High-Radix Clos Network

Abstract: This paper describes the radix-64 folded-Clos network of the

Cited by 123 publications

References 18 publications

Network congestion avoidance through Speculative Reservation

Network congestion avoidance through Speculative Reservation

Optical network technologies for HPC: computer-architects point of view

Design Methodology for Optimizing Optical Interconnection Networks in High Performance Systems

Contact Info

Product

Resources

About