2022/03/24 by Hengzhao Ma, Xiangyu Gao, Ma, Hengzhao +5
Computer Science · #Complexity and Algorithms in Graphs #Parallel Computing and Optimization Techniques #Stochastic Gradient Optimization Techniques #cs.DC
paper · pdf · doi:10.48550/arxiv.2203.12811
This is a preview version. A lot of revisions are needed
arxiv created 2022/03/24 · arxiv updated 2022/03/25
In the research area of parallel computation, the communication cost has been extensively studied, while the IO cost has been neglected. For big data computation, the assumption that the data fits in main memory no longer holds, and external memory must be used. Therefore, it is necessary to bring the IO cost into the parallel computation model. In this paper, we propose the first parallel computation model which takes IO cost as well as non-uniform communication cost into consideration. Based on the new model, we raise several new problems which aim to minimize the IO and communication cost on the new model. We prove the hardness of these new problems, then design and analyze the approximate algorithms for solving them.