Pipelined broadcast on Ethernet switched clusters

Patarasuk, Pitch; Faraj, Ahmad; Yuan, Xin

doi:10.1109/ipdps.2006.1639364

Cited by 15 publications

(9 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A general analytical performance model for tree-based broadcast algorithms with message segmentation has been proposed by Patarasuk et al [17]. Unlike traditional models, this model introduces a new parameter, Maximum nodal degree of the tree.…”

Section: A Analytical Performance Models Of Mpi Collective Algorithmsmentioning

confidence: 99%

Efficient and Accurate Selection of Optimal Collective Communication Algorithms Using Analytical Performance Modeling

Nuriyev

Lastovetsky

2021

IEEE Access

View full text Add to dashboard Cite

The performance of collective operations has been a critical issue since the advent of Message Passing Interface (MPI). Many algorithms have been proposed for each MPI collective operation but none of them proved optimal in all situations. Different algorithms demonstrate superior performance depending on the platform, the message size, the number of processes, etc. MPI implementations perform the selection of the collective algorithm empirically, executing a simple runtime decision function. While efficient, this approach does not guarantee the optimal selection. As a more accurate but equally efficient alternative, the use of analytical performance models of collective algorithms for the selection process was proposed and studied. Unfortunately, the previous attempts in this direction have not been successful. We revisit the analytical model-based approach and propose two innovations that significantly improve the selective accuracy of analytical models: (1) We derive analytical models from the code implementing the algorithms rather than from their high-level mathematical definitions. This results in more detailed models.(2) We estimate model parameters separately for each collective algorithm and include the execution of this algorithm in the corresponding communication experiment. We experimentally demonstrate the accuracy and efficiency of our approach using Open MPI broadcast and gather algorithms and two different Grid'5000 clusters and one supercomputer.

show abstract

Section: A Analytical Performance Models Of Mpi Collective Algorithmsmentioning

confidence: 99%

Efficient and Accurate Selection of Optimal Collective Communication Algorithms Using Analytical Performance Modeling

Nuriyev

Lastovetsky

2021

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Yet another model in pipelining consecutive broadcasts to maximize system throughput was first proposed in [22]. By splitting a large message into segments and broadcasting the segments in a pipeline fashion, an optimal schedule of unicastbased broadcasting can be derived.…”

Section: Related Workmentioning

confidence: 99%

Scheduling for atomic broadcast operation in heterogeneous networks with one port model

Hsu

Tsai

2009

J Supercomput

View full text Add to dashboard Cite

With the emergence of the network technologies, heterogeneous computing has become a wide accept paradigm for distributed and network computing. In this paper, we present different algorithms aiming to efficiently perform atomic oneto-all broadcast in a heterogeneous network with a one port model. The proposed algorithms are divided into graph-based and tree-based ones. In graph-based algorithms, we present Nearest Neighbor First and Maximum Degree Neighbor First schemes. A prescheduling strategy with constructing a message forwarding table for avoiding redundant transmissions is applied as runtime support. In the tree-based approaches, there are five heuristic algorithms: Nearest Neighbor First, Maximum Degree Neighbor First, Maximum Height Subtree First, Maximum Subtree First, and Maximum Weighted Subtree First, proposed based on different network characteristics. To evaluate the performance of the proposed techniques, we have developed a simulator that contains a parametric graph generator for generating network graphs with various characteristics. We have implemented all of the proposed scheduling algorithms on the simulator. The performance results show that the Maximum Weighted Subtree First performs best in high degree heterogeneous environments. On the contrary, with homogeneous-like environments, the graph-based Nearest Neighbor First will be the best choice. In summary, contribution of this study relies on informing significant suggestions for adapting proper broadcasting mechanism in different heterogeneous platforms.

show abstract

“…A combined compiler and library approach was proposed in [11]. There is a large body of work on the development of algorithms for MPI collective operations [4,5,18,25]. This work, however, does not focus on developing collective communication algorithms.…”

Section: Related Workmentioning

confidence: 99%

Star-Mpi

Faraj

Yuan

Lowenthal

2006

Proceedings of the 20th Annual International Conference on Supercomputing

View full text Add to dashboard Cite

Message Passing Interface (MPI) collective communication routines are widely used in parallel applications. In order for a collective communication routine to achieve high performance for different applications on different platforms, it must be adaptable to both the system architecture and the application workload. Current MPI implementations do not support such software adaptability and are not able to achieve high performance on many platforms. In this paper, we present STAR-MPI (Self Tuned Adaptive Routines for MPI collective operations), a set of MPI collective communication routines that are capable of adapting to system architecture and application workload. For each operation, STAR-MPI maintains a set of communication algorithms that can potentially be efficient at different situations. As an application executes, a STAR-MPI routine applies the Automatic Empirical Optimization of Software (AEOS) technique at run time to dynamically select the best performing algorithm for the application on the platform. We describe the techniques used in STAR-MPI, analyze STAR-MPI overheads, and evaluate the performance of STAR-MPI with applications and benchmarks. The results of our study indicate that STAR-MPI is robust and efficient. It is able to find efficient algorithms with reasonable overheads, and it out-performs traditional MPI implementations to a large degree in many cases. these applications to achieve high performance, it is crucial that an MPI library realizes the communications efficiently.Studies have indicated that MPI collective operations are used in most MPI applications and account for a significant portion of the total time in some applications [21]. Developing efficient MPI collective communication routines, however, is challenging. For collective communication routines to achieve high performance for

show abstract

Pipelined broadcast on Ethernet switched clusters

Cited by 15 publications

References 17 publications

Efficient and Accurate Selection of Optimal Collective Communication Algorithms Using Analytical Performance Modeling

Efficient and Accurate Selection of Optimal Collective Communication Algorithms Using Analytical Performance Modeling

Scheduling for atomic broadcast operation in heterogeneous networks with one port model

Star-Mpi

Contact Info

Product

Resources

About