Asynchronous and multithreaded communications on irregular applications using vectorized divide and conquer approach

Thébault, Loïc; Petit, Éric

doi:10.1016/j.jpdc.2017.12.004

Cited by 3 publications

(3 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Cluster systems are classi ed into three types of server groups based on how they are connected to the device responsible for storing con guration information (called a node) [7]. These three types include single (or standard) quorum clusters, majority node set clusters, and single node clusters, each of which is discussed in more detail below.…”

Section: Types Of Clustersmentioning

confidence: 99%

Analysis and modeling of Linux server clustering methods

Zeynalli

2023

Preprint

View full text Add to dashboard Cite

Currently, cluster servers and workstation systems are primarily utilized in high-performance applications and scientific computations. A cluster involves the use of numerous computers, typically PCs or UNIX workstations, along with a significant number of storage devices and unnecessary interconnections to create what appears to users as a single high-capacity system. Cluster computing can be employed for load balancing and achieving high availability. It serves as a relatively cost-effective form of a parallel processing machine for scientific and other applications that require parallel operations. Cluster server systems combine a group of servers to provide a common processing service for clients on the network. The functionality of the methods used in the cluster is crucial for enhancing the continuity and efficiency of operations on servers. This dissertation investigates the specific features, similarities, and differences of these methods. The purpose of this article is to analyze Linux server cluster methods, examine their characteristics, and conduct a deeper analysis of cluster topologies to gain further insights. To achieve this goal, a literature review has been conducted to analyze the discoveries made by other researchers in this field. By utilizing the information gathered from the literature review and following the steps of several operational researches, this study demonstrates that the agility proposed by the Linux Operating System is applicable not only to older devices but also to server clustering and the configuration requirements for various clustering techniques in a real production environment.

show abstract

Section: Types Of Clustersmentioning

confidence: 99%

Analysis and modeling of Linux server clustering methods

Zeynalli

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…This fork-join approach is usually suboptimal since the synchronization point after the parallel region limits both the intra-and inter-node parallelism. Thébault et al [30] use the fork-join strategy to parallelize irregular applications with tasks. Möller et al [31] also mix GASPI and OpenMP tasks with a similar strategy in an electromagnetic simulation application.…”

Section: Related Workmentioning

confidence: 99%

Combining One-Sided Communications with Task-Based Programming Models

Sala

Macià

Beltrán

2021

2021 IEEE International Conference on Cluster Computing (CLUSTER)

View full text Add to dashboard Cite

Hybrid programming combining task-based and message-passing models is an increasingly popular technique to exploit multi-core clusters. The Task-Aware MPI (TAMPI) library integrates both models enabling the safe overlap of computation and communication tasks using two-sided MPI communications. Two-sided primitives combine data transfers with implicit synchronizations, but one-sided models usually offer more efficient data transfers decoupling synchronizations. MPI offers four distinct one-sided synchronization modes, while GASPI is a PGAS API providing one-sided operations with remote notifications for fine inter-process synchronizations.In this paper, we study the challenges of integrating MPI and GASPI one-sided operations with the OpenMP and OmpSs-2 tasking models. We propose and implement several extensions to the GASPI and OmpSs-2 programming models, which are leveraged by a new library called Task-Aware GASPI (TAGASPI). The TAGASPI library allows the efficient and safe use of one-sided operations with remote notifications inside tasks. Both TAGASPI and TAMPI transparently manage communications issued by tasks and allow these to overlap with computation tasks naturally, following a data-flow model. These libraries are complementary and can be mixed in the same application.Our experience porting several mini-apps to this hybrid model shows that TAGASPI helps leverage one-sided communications with similar complexity to pure and hybrid two-sided MPI approaches. We show that our hybrid one-sided approach outperforms the pure MPI strategies, but it also surpasses the TAMPI's performance when stressing communication phases, e.g., increasing the communication parallelism and reducing the communication tasks' sizes.

show abstract

“…The multithreading technique is one processing technique that aims to create a virtual multiprocessor environment to execute multiple tasks on single processor [3] [4]. The recent hardware revolution plays an important a role in improving system performance through multi-core technology.…”

Section: Introductionmentioning

confidence: 99%

Parallel Platform for Supporting Stream Ciphers Over Multi-core Processors

Almanasra¹

2019

IJACSA

View full text Add to dashboard Cite

Designing secure and fast cryptographic primitives is one of the critical issues in the current era. Several domains, including Internet of Things (IoT), military and banking, require fast and secure data encryption over public channels. Most of the existing stream ciphers are designed to work sequentially and therefore not utilizing available computing power. Also, other stream ciphers are designed based on complex mathematical problems which makes these ciphers slower due to the complex computations. For this purpose, a novel parallel platform for enhancing the performance of stream ciphers is presented. The platform is designed to work efficiently over multi-core processors using multithreading techniques. The architecture of the platform relies on independent components that can operate over multiple cores available on the corresponding communication ends. Two groups of stream ciphers were considered as case studies in our experiments. The first category includes stream ciphers of a sequential design, while the second category includes parallelizable stream ciphers. Performance tests and analysis shows that the parallel platform was able to maximize the encryption throughput of the selected stream ciphers dramatically. The enhancements on the encryption throughput is relative to the constructional design of the stream ciphers. Parallelized stream ciphers (Salsa20, DSP-128, and ECSC-128) was able to achieve higher throughput compared to other sequentially designed stream ciphers.

show abstract

Asynchronous and multithreaded communications on irregular applications using vectorized divide and conquer approach

Cited by 3 publications

References 41 publications

Analysis and modeling of Linux server clustering methods

Analysis and modeling of Linux server clustering methods

Combining One-Sided Communications with Task-Based Programming Models

Parallel Platform for Supporting Stream Ciphers Over Multi-core Processors

Contact Info

Product

Resources

About