Abstract
Large cluster-based machines require efficient high-performance interconnection networks. Routing is a key design issue of interconnection networks. Adaptive routing usually outperforms deterministic routing at the expense of introducing out-of-order packet delivery. Many of the commodity interconnects for clusters are based on fat-trees. The adaptive routing algorithm commonly used in fat-trees is composed of a fully adaptive upward subpath, followed by a deterministic downward subpath. As the latter is determined by the former, choosing the most adequate upward path for each packet is critical in fat-trees to achieve a good performance. In this paper, we present a mechanism for selecting the upward path in fat-trees, which enables optimum use of the available network resources to achieve a high network throughput. The proposed path selection is destination based, which allows reducing the head-of-line blocking effect. Indeed, the proposed mechanism can be used either as a selection function (the provided path is used as the preferred one), or as a deterministic routing algorithm (the path is the only possible one). The results show that the resulting selection function outperforms any other known one. Moreover, the proposed deterministic routing algorithm can achieve a similar, or even higher, level of performance than adaptive routing, while providing in-order packet delivery and a simpler switch implementation.
Similar content being viewed by others
Notes
Other options can be also considered. However, we have found out that it is not significant in the obtained results.
References
Abali B et al (2001) Adaptive routing on the new switch chip for IBM SP systems. J Parallel Distrib Comput 61(9):1148–1179
Bakker E, van Leeuwer J, Tan RB (1991) Linear interval routing. Algoritms Rev 2:45–61
Bogdanski B, Reinemo S-A, Sem-Jacobsen FO, Gran sFtree EG (2012) A fully connected and deadlock free switch-to-switch routing algorithm for fat-trees. ACM Trans Archit Code Optim 8(4):55-1–55-20
Bogdanski B, Dag B, Reinemo S-A, Flich J (2013) Making the network scalable: inter-subnet routing in InfiniBand. In: Proceedings of the Euro-Par 2013 international conference
Dally WJ, Towles B (2004) Principles and practices of interconnection networks. Morgan Kaufmann, Burlington
Duato J, Yalamanchili S, Ni L (2004) Interconnection networks: an engineering approach. Morgan Kaufmann, Burlington
Escudero-Sahuquillo J, Gunnar E, Garcia PJ, Flich J, Skeie T, Lysne O, Quiles FJ, Duato J (2014) Efficient and cost-effective hybrid congestion control for HPC interconnection networks. IEEE Trans Parallel Distrib Syst (to apear). doi:10.1109/TPDS.2014.2307851
Flich J, Malumbres MP, López P, Duato J (2000) Improving routing performance in Myrinet networks. In: Proceedings of the 14th international parallel and distributed processing symposium
García PJ, Flich J, Duato J, Johnson I, Quiles FJ, Naven F (2005) Dynamic evolution of congestion trees: analysis and impact on switch architecture. In: Proceedings of 1st HiPEAC conference, pp 266–285
Geoffray P, Hoefler T (2008) Adaptive routing strategies for modern high performance networks. In: IEEE HOTI
Gilabert F, Gómez ME, López P, Duato J (2006) On the influence of the selection function on the performance of fat-trees. In: European conference on parallel computing
Greenberg R, Leiserson C (1985) Randomized routing on fat-trees. In: Annual symposium on the foundations of computer science
Gómez ME, López P, Duato J (2005) A memory-effective routing strategy for regular interconnection networks. In: IEEE international parallel and distributed processing symposium
Gómez C, Gilabert F, Gómez ME, López P, Duato J (2007) Deterministic versus adaptive routing in fat-trees workshop on communication architecture on clusters. In: IEEE international parallel and distributed processing symposium
Hillis WD, Tucker L (1993) The CM-5 connection machine: a scalable supercomputer. Commun ACM 36(11):31–40
Hoefler T, Schneider T, Lumsdaine A (2009) Optimized routing for large-scale InfiniBand networks. In: Proceedings of the 2009 17th IEEE symposium on high performance interconnects
Infiniband Trade Association. http://www.infinibandta.org
Johnson G, Kerbbyson D, Lang M (2008) Optimization of InfiniBand scientific applications. In: 22nd international parallel and distributed processing
Kariniemi H (2006) On-line reconfigurable extended generalized fat tree network-on-chip for multiprocessor system-on-chip circuits. PhD. thesis, Tampere University of Technology
Karol MJ, Hluchyj MG, Morgan SP (1987) Input versus output queueing on a space-division packet switch. IEEE Trans Commun 35:1347–1356
Kim J, Park D, Theocharides T, Vijaykrishnan N, Das CR (2005) A low latency router supporting adaptivity for on-chip interconnects. In: 42nd annual conference on design automation
Kim J, Dally WJ, Dally J, Abts D (2006) Adaptive routing in high-radix clos network. In: SC 2006 conference, proceedings of the ACM/IEEE, Tampa, FL, 7 Nov 2006. doi:10.1109/SC.2006.10
Lin X, Chung Y, Huang T (2004) A multiple LID routing for fat-tree-based InfiniBand networks. In: IEEE international parallel and distributed processing symposium
Martínez JC, Flich J, Robles A, López P, Duato J (2004) Supporting adaptive routing in IBA switches. J Syst Archit 49:441–449
Martínez JC, Flich J, Robles A, López P, Duato J, Koibuchi M (2005) In-order packet delivery in interconnection networks using adaptive routing. In: IEEE international parallel and distributed processing symposium
Myricom. http://www.myri.com
Petrini F, Vanneschi M (1995) k-ary n-tress: high performance networks for massively parallel architecture. In: IEEE Micro, vol 15
Quadrics homepage. http://www.quadrics.com
Scott S, Abts D, Kim J, Dally WJ (2006) The BlackWidow high-radix clos network. In: International sympium on computer architecture
Ruemmler C, Wilkes J (1993) Unix disk access patterns. In: Winter Usenix conference
Tianhe. http://www.nscc-tj.gov.cn/en/
Top 500 Supercomputer site (2014). http://www.top500.org
Vishnu A, Koop M, Moody A, Mamidala A, Narravula S, Panda D (2007) Hot-spot avoidancce with multipathing over InfiniBand: an MPI perspective. In: International symposium on cluster computing and the grid
Zahavi E, Johnson G, Kerbyson DJ, Lang M (2010) Optimized InfiniBandTM fat-tree routing for shift all-to-all communication patterns. Concurr Comput Pract Experience 22:2
Acknowledgments
This work was supported by the Spanish Ministerio de Ciencia e Innovación (MICINN) and jointly financed with Plan E funds, under Grant TIN2009-14475-C04 as well as by Consolider-Ingenio 2010 under Grant CSD2006-00046.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Gómez, C., Gilabert, F., Gómez, M.E. et al. A HoL-blocking aware mechanism for selecting the upward path in fat-tree topologies. J Supercomput 71, 2339–2364 (2015). https://doi.org/10.1007/s11227-014-1303-x
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11227-014-1303-x