15 Hits in 5.7 sec

Netloc: A Tool for Topology-Aware Process Mapping [chapter]

Cyril Bordage, Clément Foyer, Brice Goglin
2018 Lecture Notes in Computer Science  
We show that our Netloc tool for gathering network topology in a generic way can be combined with the state-of-the-art Scotch partitioner for computing topology-aware MPI process placement.  ...  Topology-aware process mapping consists in matching the application communication pattern with the network topology to improve the communication cost by placing related tasks close on the hardware.  ...  We discuss in the next sections the use of Netloc for topology-aware process placement.  ... 
doi:10.1007/978-3-319-75178-8_13 fatcat:p4ufuvdpbrdwxfwwi4gcxi3unu

Netloc: Towards a Comprehensive View of the HPC System Topology

Brice Goglin, Joshua Hursey, Jeffrey M. Squyres
2014 2014 43rd International Conference on Parallel Processing Workshops  
Additionally, netloc provides the ability to merge the network topology with the server-internal topologies resulting in a comprehensive map of the HPC system topology.  ...  Using a modular infrastructure, netloc provides support for a variety of network types and discovery techniques.  ...  Special thanks to Douglas MacFarland and Nicholas Buroker for their work on early prototypes of this project.  ... 
doi:10.1109/icppw.2014.38 dblp:conf/icppw/GoglinHS14 fatcat:w53ulau63bhazkw2cpnxxn4y3a

Topology-Aware Mapping Techniques for Heterogeneous HPC Systems: A Systematic Survey

Saad B. Alotaibi, Fathy alboraei
2018 International Journal of Advanced Computer Science and Applications  
In this survey paper, we have studied various topology-aware mapping techniques and algorithms.  ...  Given that, the efficient topology-aware process mapping has become vital to efficiently optimize the data locality management in order to improve the system performance and energy consumption.  ...  [16] proposed a Netloc tool for collecting the physical topology that is integrated with a Scotch practitioner for computing the topology-aware MPI process placement.  ... 
doi:10.14569/ijacsa.2018.091045 fatcat:taeescyyjjej7pbqutm4kulagy

Managing the topology of heterogeneous cluster nodes with hardware locality (hwloc)

Brice Goglin
2014 2014 International Conference on High Performance Computing & Simulation (HPCS)  
Thus there is a strong need for a portable tool gathering and exposing this information.  ...  We also describe how hwloc now helps process managers and batch schedulers to deal with the topology of multiple cluster nodes, together with compression for better scalability up to thousands of nodes  ...  This is useful for developing topology-aware algorithms and testing on a variety of different platform topologies.  ... 
doi:10.1109/hpcsim.2014.6903671 dblp:conf/hpcs/Goglin14 fatcat:o52gb56fnncbtoqyocwiaq7gou

Hardware topology management in MPI applications through hierarchical communicators

Brice Goglin, Emmanuel Jeannot, Farouk Mansouri, Guillaume Mercier
2018 Parallel Computing  
It provides the user with tools to address hardware topology and locality issues while improving application performance.  ...  Since its inception in the mid 90s it has ensured portability and performance for parallel applications on a wide spectrum of machines and architectures.  ...  The authors would like to thank the MPI Forum for its feedback, especially Daniel Holmes, as well as Cyril Bordage from implementing the netloc extension for retrieving network coordinates of nodes.  ... 
doi:10.1016/j.parco.2018.05.006 fatcat:u6zhfpnzyzgyrlcq633u7p642i

Hardware Locality-Aware Partitioning and Dynamic Load-Balancing of Unstructured Meshes for Large-Scale Scientific Applications

Pavanakumar Mohanamuraly, Gabriel Staffelbach
2020 Proceedings of the Platform for Advanced Scientific Computing Conference  
We present an open-source topology-aware hierarchical unstructured mesh partitioning and load-balancing tool TreePart.  ...  The tool was successfully integrated into our in-house code and we present results from a large-eddy simulation of a combustion problem.  ...  This work was granted access to the HPC resources of IDRIS under an allocation by GENCI for the Grand Challenges Jean Zay (2019).  ... 
doi:10.1145/3394277.3401851 dblp:conf/pasc/MohanamuralyS20 fatcat:thhnjsb7ufbgbmvmf6abiyzlri

A hierarchical model to manage hardware topology in MPI applications

Emmanuel Jeannot, Farouk Mansouri, Guillaume Mercier
2017 Proceedings of the 24th European MPI Users' Group Meeting on - EuroMPI '17  
Such a tool actually exists: netloc [4] . netloc is a hwloc extension that specically addresses network hierarchies and topologies.  ...  Once again, the lowest level is chosen and in this 6 A "by node" mapping policy in conjunction with a "by core" binding policy for processes.  ... 
doi:10.1145/3127024.3127030 dblp:conf/pvm/JeannotMM17 fatcat:qep5w4i6yvalfgueno7dfmvcci

Topology-aware resource management for HPC applications

Yiannis Georgiou, Emmanuel Jeannot, Guillaume Mercier, Adèle Villiermet
2017 Proceedings of the 18th International Conference on Distributed Computing and Networking - ICDCN '17  
This paper introduces a new topology-aware resource selection algorithm to determine the best choice among the available nodes of the platform based upon their position within the network and taking into  ...  To validate our approach, we integrated this algorithm as a plugin for Slurm, a popular and widespread HPC resource and job management system (RJMS).  ...  Acknowledgments Experiments presented in this paper were carried out using the Grid'5000 testbed, supported by a scientic interest group hosted by Inria and including CNRS, RENATER and several Universities  ... 
doi:10.1145/3007748.3007768 fatcat:sflvwmx525gubn6svjanowe6fe

Topology-aware job mapping

Yiannis Georgiou, Emmanuel Jeannot, Guillaume Mercier, Adèle Villiermet
2017 The international journal of high performance computing applications  
We show that transparently taking into account a job communication pattern and the topology allows for relevant performance gains.  ...  To validate our approach, we integrate this algorithm as a plugin for Slurm, a well-known and widespread RJMS.  ...  Acknowledgments Experiments presented in this paper were carried out using the Grid'5000 testbed, supported by a scientific interest group hosted by Inria and including CNRS, RENATER and several Universities  ... 
doi:10.1177/1094342017727061 fatcat:35fnfho4wje4nmlgi3wv7wbilu

Trends in Data Locality Abstractions for HPC Systems

Didem Unat, Anshu Dubey, Torsten Hoefler, John Shalf, Mark Abraham, Mauro Bianco, Bradford L. Chamberlain, Romain Cledat, H. Carter Edwards, Hal Finkel, Karl Fuerlinger, Frank Hannig (+9 others)
2017 IEEE Transactions on Parallel and Distributed Systems  
Support for expression of data locality has been explored in the past, but those efforts have had only modest success in being adopted in HPC applications for various reasons.  ...  However, with the increasing complexity of the memory hierarchy and higher parallelism in emerging HPC systems, locality management has acquired a new urgency.  ...  Netloc [26] , [49] is a network model extension of hwloc to account for locality requirements of the network, including the fabric topology.  ... 
doi:10.1109/tpds.2017.2703149 fatcat:vjalwrujhrex7cibod3qerf3z4

Specification Of Hpc Hardware And Program Components To Enable Further Optimized Mappings

Carlchristian Helmut Johannes Eckert, Wolfgang E Nagel, Jerónimo Castrillón
2016 Zenodo  
Through these features, Dodo forms the base for tools that can specialize in the creation of optimized domain decompositions and mappings.  ...  Domain decompositions are expressed through graph mappings. These mappings combine the different models to form a comprehensive view of a simulation run.  ...  As with hwloc, it is possible to use netloc as a tool to gather relevant data for Dodo.  ... 
doi:10.5281/zenodo.163329 fatcat:mkb4ewc4nvdfrcclwlqzrbg7ry

Process Affinity, Metrics and Impact on Performance: An Empirical Study

Cyril Bordage, Emmanuel Jeannot
2018 2018 18th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (CCGRID)  
Process placement, also called topology mapping, is a well-known strategy to improve parallel program execution by reducing the communication cost between processes.  ...  It requires two inputs: the topology of the target machine and a measure of the affinity between processes.  ...  We thank the JLESC structure for providing us with some hours on the BlueWaters machine. We also want to thank Clément Foyer for providing us with the Open MPI Inria  ... 
doi:10.1109/ccgrid.2018.00079 dblp:conf/ccgrid/BordageJ18 fatcat:rlf5jhd6grgxzizrf3sqvvjbue

Evaluating locality-aware extensions for task migration in distributed memory

Marvin Porsil, Jannis Klinkenberg, Matthias S. Müller, Karl Fuerlinger
Considering the topological distances between computing nodes or processes running on these nodes in a cluster network when choosing a migration victim for load balancing proved to result in only small  ...  Chameleon, a library for reactive load balancing for hybrid MPI+OpenMP task-parallel applications, provides a way to balance the load in distributed memory systems, across process boundaries.  ...  In this section, the implementation and evaluation of a tool for the Chameleon-Tools interface for network topology-aware task migration is described.  ... 
doi:10.18154/rwth-2021-09161 fatcat:sov5xbys2jcfnioyy7asci2ybm

Active Learning for Source Localization

Victor Cărbune
I'm very grateful to Sînziana for always being by my side throughout the master's degree.  ...  I would like to thank my parents, Lucia and Vilut , , and my sister, Maria, for their endless positive energy and support.  ...  , in Section A.1.  ... 
doi:10.3929/ethz-a-010144433 fatcat:xe26hcy3c5esndjrfan2xvx6su

Large Scale Benchmarking of Broadband Access Networks: Issues, Methodologies, and Solutions

Walter De Donato
As a result of the preceding analysis we identified the ideal characteristics for an architecture able to evaluate the performance of broadband access networks on a large scale.  ...  we conduct research activities on both the analysis and characterization of traffic generated by new-generation applications and the identification of relevant metrics, methodologies, techniques and tools  ...  appropriate stochastic processes for both IDT (Inter Departure Time) and PS (Packet Size) random variables (exponential, uniform, cauchy, normal, pareto, ...), which also acts as a tool for measuring  ... 
doi:10.6092/unina/fedoa/8954 fatcat:2j5d4dgjije5za7wmezphfn3hy