论文部分内容阅读
当前,集群是流行的并行计算平台,既具有所有并行计算都具备的一般特性,又具有自己独有的特性———节点计算资源的异构性、节点计算资源的非独占性、对节点间通信的敏感性。针对基于集群环境的并行计算,对非独占、异构、同步LogGP模型NHBL(Nondedicated Heterogeneous Barrier Log GP model)进行改进,提出了一种更实用的计算模型,即考虑网络拥塞的非独占、异构、同步LogGP模型NHCBL(Nondedicated Heterogeneous Barrier Log GP model considering Conges-tion)。NHCBL模型既比较准确地描述了并行计算的一般特性,又比较准确地描述了基于集群的并行计算独有的节点计算资源的异构性、节点计算资源的非独占性、对节点间通信的敏感性,其描述比NHBL模型更全面也更准确。
At present, the cluster is a popular parallel computing platform. It not only has the general characteristics of all parallel computing, but also has its own unique characteristics --- the heterogeneity of computing resources, the non-exclusive computing resources of nodes, Sensitivity to communications. Aiming at the problem of parallel computing based on cluster environment, a more practical computing model is proposed to improve the Non-Exclusive, Heterogeneous, and Synchronous Log-NHGP (Non-Heterogeneous Heterogeneous Barrier Log GP model), which is to consider the non-exclusive and heterogeneous , Non Logical Heterogeneous Barrier Log (GP) GP modeling considering Congestion. NHCBL model not only describes the general characteristics of parallel computing more accurately, but also more accurately describes the heterogeneity of computing resources of nodes based on cluster-based parallel computing, the non-exclusiveness of computing resources of nodes, and the sensitivity to the communication among nodes The description is more comprehensive and accurate than the NHBL model.