多核集群任务分配问题的0-1整数规划求解模型  

A model for solving task allocation problems in multi-core clusters using 0-1 integer programming

作　　者：杨际祥[1] 凌玲[1] 

机构地区：[1]重庆交通大学数学与统计学院,重庆400074

出　　处：《高技术通讯》2016年第4期344-348,共5页Chinese High Technology Letters

基　　金：国家自然科学基金(11401061);重庆市自然科学基金(KJ1400316);交通运输部建设科技基金(2014318223030)资助项目

摘　　要：研究了典型多核集群任务分配中的节点内通讯特性。基于0-1整数非线性规划模型和线性松弛技术,给出了一种0-1整数线性规划任务分配问题求解优化模型。由于节点内的通讯量与通讯延迟较大,以最小化计算代价和节点间通讯代价为研究目标的传统求解模型具有严重的局限性,而该求解模型考虑了节点内通讯代价,并采用了线性规划松弛技术,其目标是最小化计算代价、节点间通讯代价和节点内通讯代价。计算结果验证了提出的模型的有效性。The in-node communication characteristics of the task allocation of multi-core clusters was studied. An opti- mized model for solving task allocation problems in multi-core clusters using 0-1 integer linear programming was proposed based on the 0-1 integer linear programming model and the linear relaxation technique. Considering that the traffic and delay of in-node communications are bigger, which brings serious limitations to the traditional models aiming to minimize the computing cost and the cost of the node-node communications, the proposed model takes ac- count of the cost of in-node communications and adopts the technique of linear programming relaxation, with the aim to minimize the computing cost, the cost of node-node communications, and the in-node communication cost. The effectiveness of the model is verified by the computation result.

关 键 词：多核集群 任务分配问题(TAP) 0-1整数规划 线性规划松弛 

分 类 号：O221.4[理学—运筹学与控制论]