Application-Oriented Adaptive MPI Bcast for Grids
نویسندگان
چکیده
Due to the importance of collective communications in scientific parallel applications, many strategies have been devised for optimizing collective communications for different kinds of parallel environments. Recently, there has been an increasing interest to evolve efficient broadcast algorithms for computational Grids. In this paper, we present application-oriented adaptive techniques that take into account recent resource characteristics as well as the application’s usage of broadcasts for deriving efficient broadcast trees. In particular, we consider two broadcast parameters used in the application, namely, the broadcast message sizes and the time interval between the broadcasts. The results indicate that our adaptive strategies can provide 20% average improvement in performance over the popular MPICH-G2’s MPI Bcast implementation for loaded network conditions.
منابع مشابه
Optimized Broadcast for Deep Learning Workloads on Dense-GPU InfiniBand Clusters: MPI or NCCL?
Dense Multi-GPU systems have recently gained a lot of attention in the HPC arena. Traditionally, MPI runtimes have been primarily designed for clusters with a large number of nodes. However, with the advent of MPI+CUDA applications and CUDA-Aware MPI runtimes like MVAPICH2 and OpenMPI, it has become important to address efficient communication schemes for such dense Multi-GPU nodes. This couple...
متن کاملMPI Support on the Grid
Grids as infrastructures offer access to computing, storage and other resources in a transparent way. The user does not have to be aware where and how the job is being executed. Grid clusters in particular are an interesting target for running computation-intensive calculations. Running MPI-parallel applications on such clusters is a logical approach that is of interest to both computer scienti...
متن کاملTowards an Internet Operating System: Middleware for Adaptive Distributed Computing
Large-scale, dynamic, and heterogeneous networks of computational resources (a.k.a. grids) promise to provide high performance and scalability to computationally intensive applications. These environments also introduce the need for complex resource management strategies. We propose decentralized middleware-triggered dynamic reconfiguration strategies to enable application adaptation to the con...
متن کاملThe Internet Operating System: Middleware for Adaptive Distributed Computing
Large-scale, dynamic, and heterogeneous networks of computational resources (a.k.a. grids) promise to provide high performance and scalability to computationally intensive applications. To fulfill this promise, grid environments require complex resource management. We propose decentralized middlewaretriggered dynamic reconfiguration strategies to enable application adaptation to the constantly ...
متن کاملApplication of Malmquist Index in Two-Stage DEA for Measurement of Productivity Growth
The purpose of this paper is to develop an output oriented methodology for calculating productivity growth by using Malmquist productivity index (MPI) and two different data envelopment analysis (DEA) views (optimistic and pessimistic) simultaneously, and apply it to five Iranian Commercial Banks over the four time period (2009-2013). Consequently, we have proposed a new approach called the dou...
متن کامل