75 Hits in 5.0 sec

Hardware supported multicast in fat-tree-based InfiniBand networks

Jiazheng Zhou, Xuan-Yi Lin, Yeh-Ching Chung
2007 Journal of Supercomputing  
With the hardware supported multicast of the InfiniBand Architecture (IBA), in this paper, we propose a cyclic multicast scheme for fat-treebased (m-port n-tree) InfiniBand networks.  ...  We implement the proposed multicast scheme along with the OpenSM multicast scheme and the unicast scheme on an m-port n-tree InfiniBand network simulator.  ...  We propose a cyclic multicast scheme for the m-port n-tree (a fat-tree) InfiniBand networks [12] based on the hardware supported multicast feature of the IBA and the characteristics of m-port n-tree  ... 
doi:10.1007/s11227-006-0019-y fatcat:ojfjhxpz7jbc5ini3dnxo4magq

Scalable Hierarchical Aggregation and Reduction Protocol (SHARP)TM Streaming-Aggregation Hardware Design and Evaluation [chapter]

Richard L. Graham, Lion Levi, Devendar Burredy, Gil Bloch, Gilad Shainer, David Cho, George Elias, Daniel Klein, Joshua Ladd, Ophir Maor, Ami Marelli, Valentin Petrov (+3 others)
2020 Lecture Notes in Computer Science  
Some network-hardware-based solutions have been implemented, with those relevant to the focus of this paper reviewed in Sect. 2.  ...  This paper describes the new hardware-based streamingaggregation capability added to Mellanox's Scalable Hierarchical Aggregation and Reduction Protocol in its HDR InfiniBand switches.  ...  Summary This paper describes the Mellanox SHARP Streaming-Aggregation capability introduced in Mellanox's HDR InfiniBand network hardware.  ... 
doi:10.1007/978-3-030-50743-5_3 fatcat:zcadrld7hzcm3izseyc65o5ld4

Towards A Data Centric System Architecture: SHARP

2017 Supercomputing Frontiers and Innovations  
The use of UD-Multicast to distribute aggregation results is introduced, reducing the letency of an eight-byte MPI Allreduce() across 128 nodes by 16%.  ...  Use of reduction trees that avoid the inter-socket bus further improves the eight-byte MPI Allreduce() latency across 128 nodes, with 28 processes per node, by 18%.  ...  An aggregation tree can be used to distribute the data in these cases. The new hardware capability described in this paper is that the target may also be a userdefined InfiniBand multicast address.  ... 
doi:10.14529/jsfi170401 fatcat:ul33psqlf5bltnzbzcdbbcmlni

HPC Ethernet

Ben Matthews
2021 Zenodo  
This presentation presents details on a multi-home EVPN network configuration for a 100Gb ethernet backbone that was deployed at NCAR.  ...  • Can't we just do a fat-tree like we'd do on IB?  ...  based on the source and destination addresses (LIDs) in a predetermined forwarding table • ECMP is kind of similar, except we don't do it based on a static table of possible hashes and we might hash more  ... 
doi:10.5281/zenodo.5552573 fatcat:4qsbfjvtcbbhzcz52awg45fiem

HPP Switch: A Novel High Performance Switch for HPC

Dawei Wang, Zheng Cao, Xinchun Liu, Ninghui Sun
2008 2008 16th IEEE Symposium on High Performance Interconnects  
In this paper, HPP Switch, as the core component of interconnection network of a HPC prototype, is introduced to meet these requirements.  ...  The applications of HPC not only demand on the low latency and high bandwidth of the switch, but also need the effective support of collective communication, such as broadcast, multicast, and barrier etc  ...  The implementation method can be classified into two classes: software-based and hardware-based. The soft-ware-based multicast usually implements in the Spanning Tree algorithm, with high overhead.  ... 
doi:10.1109/hoti.2008.17 dblp:conf/hoti/WangCLS08 fatcat:maawhiw2tzd67oxeqlltmw5jde

High-Performance Routing with Multipathing and Path Diversity in Ethernet and HPC Networks [article]

Maciej Besta, Jens Domke, Marcel Schneider, Marek Konieczny, Salvatore Di Girolamo, Timo Schneider, Ankit Singla, Torsten Hoefler
2020 arXiv   pre-print
Thus, we cover architectures and protocols based on Ethernet, InfiniBand, and other HPC networks such as Myrinet.  ...  In this work, to facilitate high-performance routing in modern networks, we analyze existing routing protocols and architectures, focusing on how well they exploit the diversity of minimal and non-minimal  ...  Finally, QsNet offers hardware support for broadcasts, and for multicasts to physically contiguous QsNet endpoints.  ... 
arXiv:2007.03776v3 fatcat:zwex5rqaobdttnpduurzww6gze

Scheduling-Aware Routing for Supercomputers

Jens Domke, Torsten Hoefler
2016 SC16: International Conference for High Performance Computing, Networking, Storage and Analysis  
Our routing method is implemented in the standard InfiniBand tool set and can immediately be used to optimize systems.  ...  We first demonstrate this by defining the dark fiber metric as a measure of unused resource in networks.  ...  InfiniBand network ports must support up to 15 VLs for data traffic-8 data VLs is common on current hardware-and one VL for management traffic.  ... 
doi:10.1109/sc.2016.12 dblp:conf/sc/DomkeH16 fatcat:fq3smia6yjh3lelvzz2riocdk4

Performance evaluation of deterministic routings, multicasts, and topologies on RHiNET-2 cluster

M. Koibuchi, K. Watanabe, T. Otsuka, H. Amano
2005 IEEE Transactions on Parallel and Distributed Systems  
In this paper, we implement and evaluate deadlock-free routings and unicast-based multicasts under various topologies and channel buffer sizes on a PC cluster called RHiNET-2 with 64 hosts.  ...  System Area Networks (SANs), which usually accept arbitrary topologies, have been used to connect nodes in PC/WS clusters or high-performance storage systems.  ...  Tree-based multicast is efficient especially at a broadcast operation in regular topologies like fat tree.  ... 
doi:10.1109/tpds.2005.97 fatcat:e3yyeh7fzbh5rcvq2hqeydzqci

The Case for Network Coding for Collective Communication on HPC Interconnection Networks

2015 IEICE transactions on information and systems  
Quantitative analysis show that the aggregate path hop counts by our hierarchical network coding decrease as much as 94% when compared to conventional unicast-based multicasts.  ...  Our proposed network coding scheme has a hierarchical multicasting structure with intra-group and inter-group unicasts.  ...  Acknowledgements This work was partially supported by JST CREST and KAKENHI # 25280018 and 25280043.  ... 
doi:10.1587/transinf.2014edp7255 fatcat:awba7wxj45bpzktild54ah3u5i

Efficient Two-Opt Collective-Communication Operations on Low-Latency Random Network Topologies

Ke CUI, Michihiro KOIBUCHI
2020 IEICE transactions on information and systems  
However, a few network topologies, i.e., fat tree, torus, Dragonfly [1]-[3], have been used to interconnect compute nodes in parallel computers.  ...  In this study, we firstly apply a two-opt algorithm for building efficient multicast on random network topologies.  ...  Acknowledgments This work was partially supported by JSPS KAKENHI 19H01106.  ... 
doi:10.1587/transinf.2020pap0004 fatcat:3kn4iv76treilbnwjlhm7dnbmq

An Abstract Interface for System Software on Large-Scale Clusters

J. Fernandez
2006 Computer journal  
In this paper we propose and demonstrate an abstract network interface in the cluster interconnect to facilitate the implementation of a simple yet powerful global operating system.  ...  These challenges may seem daunting with commodity hardware and operating systems, since they were not designed to support a global, single management view of a large-scale system.  ...  ACKNOWLEDGEMENTS This work is supported by the U.S. Department of Energy through Los Alamos National Laboratory contract W-7405-ENG-36.  ... 
doi:10.1093/comjnl/bxl020 fatcat:a6jlukw5wrdtzptxojmiguquvu

A Cost-Effective, High Bandwidth Server I/O network Architecture for Cluster Systems

Hsing-bung Chen, Gary Grider, Parks Fields
2007 2007 IEEE International Parallel and Distributed Processing Symposium  
improvement through reducing large number of network components in server I/O network, and (5) global storage/file systems support in heterogeneous multi-cluster and Grids environments.  ...  We use the PaScal server I/O network to support data-intensive scientific applications running on very large-scale Linux clusters.  ...  ACKNOWLEDGEMENTS We are thankful to co-workers from LANL's HPC-3, CTN-5, and HPC-5 groups for their support on the design and implementation of PaScal Server I/O Infrastructure, Benjamin McCelland (LANL's  ... 
doi:10.1109/ipdps.2007.370221 dblp:conf/ipps/ChenGF07 fatcat:4sktmbchvjacbl4iznptvrvpxe

Scalable and Reliable Data Broadcast with Kascade

Stephane Martin, Tomasz Buchert, Pierric Willemet, Olivier Richard, Emmanuel Jeanvoine, Lucas Nussbaum
2014 2014 IEEE International Parallel & Distributed Processing Symposium Workshops  
saturate a 1 Gbit/s network, even at large scale; (3) handles failures of nodes during the transfer gracefully thanks to a fault-tolerant design.  ...  That distribution of data often has a key role in the overall performance of the operation. In this paper, we present Kascade, a solution for the broadcast of data to a large set of compute nodes.  ...  In [16] and [17] , implementations of the MPI_Bcast operation on top of InfiniBand hardware multicast support are proposed. IP multicast can also be used to achieve reliable broadcast.  ... 
doi:10.1109/ipdpsw.2014.191 dblp:conf/ipps/MartinBWRJN14 fatcat:oo6m3vuquffghiet6nyi3mvxqi

A Novel Query Caching Scheme for Dynamic InfiniBand Subnets

Evangelos Tasoulas, Ernst Gunnar Gran, Bjorn Dag Johnsen, Tor Skeie
2015 2015 15th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing  
Xinoula Song for believing in me, supporting me in so many different ways and praising me unconditionally; To my friend and soon-to-be Dr.  ...  The advent of the Internet of Things, sensor and social networks, to mention just a few examples, all contribute towards the solid establishment of the Big Data era.  ...  Different variations of Fat-Trees are presented in the literature, including k -ary-n-trees [101] , Extended Generalized Fat-Trees (XGFTs) [98] , Parallel Ports Generalized Fat-Trees (PGFTs) and Real  ... 
doi:10.1109/ccgrid.2015.10 dblp:conf/ccgrid/TasoulasGJS15 fatcat:hsxswl63gfbtbdipuhvrhx6y7u

Enhancing InfiniBand with OpenFlow-Style SDN Capability

Jason Lee, Zhou Tong, Karthik Achalkar, Xin Yuan, Michael Lang
2016 SC16: International Conference for High Performance Computing, Networking, Storage and Analysis  
Consider an HPC cluster whose interconnect has a fat-tree topology with our proposed SDNenhanced InfiniBand.  ...  Interconnect Topology and Routing The interconnects are assumed to be fat-tree topologies.  ... 
doi:10.1109/sc.2016.35 dblp:conf/sc/LeeTAY016 fatcat:3vqnhtbjdbfileymb62a3yzifa
« Previous Showing results 1 — 15 out of 75 results