Malleable iterative MPI applications

Maghraoui, Kaoutar El; Desell, Travis; Szymanski, Boleslaw K.; Varela, Carlos A.

doi:10.1002/cpe.1362

Cited by 23 publications

(20 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The work in [8] explores how malleability can be used in checkpoint/restart applications. There, a checkpoint-and-reconfigure mechanism is leveraged to restart applications with a different number of processes from data stored in checkpoint files.…”

Section: Related Workmentioning

confidence: 99%

Efficient Scalable Computing through Flexible Applications and Adaptive Workloads

Iserte

Mayo

Quintana‐Ortí

et al. 2017

2017 46th International Conference on Parallel Processing Workshops (ICPPW)

View full text Add to dashboard Cite

Abstract-In this paper we introduce a methodology for dynamic job reconfiguration driven by the programming model runtime in collaboration with the global resource manager. We improve the system throughput by exploiting malleability techniques (in terms of number of MPI ranks) through the reallocation of resources assigned to a job during its execution. In our proposal, the OmpSs runtime reconfigures the number of MPI ranks during the execution of an application in cooperation with the Slurm workload manager. In addition, we take advantage of OmpSs offload semantics to allow application developers deal with data redistribution. By combining these elements a job is able to expand itself in order to exploit idle nodes or be shrunk if other queued jobs could be initiated. This novel approach adapts the system workload in order to increase the throughput as well as make a smarter use of the underlying resources. Our experiments demonstrate that this approach can reduce the total execution time of a practical workload by more than 40% while reducing the amount of resources by 30%.

show abstract

Section: Related Workmentioning

confidence: 99%

Efficient Scalable Computing through Flexible Applications and Adaptive Workloads

Iserte

Mayo

Quintana‐Ortí

et al. 2017

2017 46th International Conference on Parallel Processing Workshops (ICPPW)

View full text Add to dashboard Cite

show abstract

“…However, most of them are very restrictive with respect to the kind of supported applications. Indeed, only iterative or master-slave applications are considered (e.g., [61,86,14]), because in these cases the modification of the number of processes is much easier than in a general application. Furthermore, in these approaches reconfiguration can only take place in very specific points within the applications.…”

Section: 22mentioning

confidence: 99%

On Processing Extreme Data

Petcu¹,

Iuhasz²,

Pop³

et al. 2016

SCPE

View full text Add to dashboard Cite

Abstract. Extreme Data is an incarnation of Big Data concept distinguished by the massive amounts of data that must be queried, communicated and analyzed in near real-time by using a very large number of memory or storage elements and exascale computing systems. Immediate examples are the scientific data produced at a rate of hundreds of gigabits-per-second that must be stored, filtered and analyzed, the millions of images per day that must be analyzed in parallel, the one billion of social data posts queried in real-time on an in-memory components database. Traditional disks or commercial storage nowadays cannot handle the extreme scale of such application data.Following the need of improvement of current concepts and technologies, we focus in this paper on the needs of data intensive applications running on systems composed of up to millions of computing elements (exascale systems). We propose in this paper a methodology to advance the state-of-the-art. The starting point is the definition of new programming paradigms, APIs, runtime tools and methodologies for expressing data-intensive tasks on exascale systems. This will pave the way for the exploitation of massive parallelism over a simplified model of the system architecture, thus promoting high performance and efficiency, offering powerful operations and mechanisms for processing extreme data sources at high speed and/or real time.

show abstract

“…[14], [20]. The survey by Belbekkouche [2] provides a nice overview of allocation and embedding algorithms.…”

Section: Related Workmentioning

confidence: 99%

Optimizing Long-Lived CloudNets with Migrations

Schaffrath

Schmid

Feldmann

2012

2012 IEEE Fifth International Conference on Utility and Cloud Computing

View full text Add to dashboard Cite

Abstract-This paper attends to the problem of embedding flexibly specified virtual networks connecting cloud resources (e.g, storage or computation) on a given substrate (e.g., a data center, an ISP backbone, a router site, or a virtual provider network). We study a scenario where a substrate provider (or a potential intermediate broker or reseller) wants to optimize the embedding of these so-called CloudNets by migrating them to more suitable locations. For instance, such re-embeddings can be useful if the CloudNets were requested at short notice and initially placed heuristically. Subsequent optimizations can, e.g., reduce the peak resource loads in the network by spreading CloudNets across the infrastructure or save energy by moving CloudNets together and switching off unused components.We present the generic mathematical programming algorithm used in our CloudNet prototype to compute optimal embeddings. For example, this algorithm supports different objective functions (such as load minimization or energy conservation), arbitrary resource combinations and the mapping of multiple virtual nodes of a CloudNet to a single substrate node, cost-aware migrations, and it can deal with all link types that arise in practice (e.g., fullduplex or even wireless or wired broadcast links with multiple endpoints). Of course, such rigorous CloudNet optimizations are time consuming, and we report on the time complexities obtained from our experiments with our network virtualization prototype architecture. It turns out that optimizing CloudNets over moderate sized infrastructures is feasible, even for scenarios with high flexibility and without tuning the solver software to speed up computations further.

show abstract

Malleable iterative MPI applications

Cited by 23 publications

References 16 publications

Efficient Scalable Computing through Flexible Applications and Adaptive Workloads

Efficient Scalable Computing through Flexible Applications and Adaptive Workloads

On Processing Extreme Data

Optimizing Long-Lived CloudNets with Migrations

Contact Info

Product

Resources

About