HomeContact UsPast Events

Accepted Papers

  • GPU_MF_SGD : A NOVEL GPU-BASED STOCHASTIC GRADIENT DESCENT METHOD FOR MATRIX FACTORIZATION
    Mohammed Ali, Alexandria University, Egypt
    ABSTRACT

    Recommender systems are used in most of nowadays applications. Providing real-time suggestions with high accuracy is considered as one of the most crucial challenges that faces them. Matrix factorization (MF) is an effective technique for recommender systems as it improves the accuracy. Stochastic Gradient Descent (SGD) for MF is the most popular approach used to speed up MF. SGD is a sequential algorithm, which is not trivial to be parallelized, especially for large-scale problems. Recently, many researches have proposed parallel methods for parallelizing SGD. In this research, we propose GPU_MF_SGD, a novel GPU-based method for large-scale recommender systems. GPU_MF_SGD utilizes GPU resources by ensuring load balancing and linear scalability, and achieving coalesced access of global memory without preprocessing phase. Our method demonstrates 3.1X-5.4X speedup over the most state-of-the-art GPU method, CuMF_SGD.

  • ADAPTED BIN PACKING ALGORITHM FOR VIRTUAL MACHINES PLACEMENT INTO DATACENTERS
    Fréjus A. R. Gbaguidi1,3, Selma Boumerdassi1,2 and Eugène C. EZIN3, 1CNAM / CEDRIC, Paris, France, 2INRIA Hipercom, France and 3IMSP /UAC, Benin, France
    ABSTRACT

    The placement of virtual machines is a permanent routine that determines both performance and energy efficiency within Datacenters. Unfortunately, it is a task whose complexity is fully supported by the common sense of the system administrators who must try different scenarios in order to detect the one that best satisfies the constraints imposed by the environment. Bin packing techniques have been used to address similar issues in other areas such as transportation and mass distribution. We try to apply these methods to the problem of placing virtual machines on the physical servers within Datacenters. Our aim is to evaluate the efficiency of this technique at the optimum distribution of the VM while using the minimum number of physical machines and consequently reduce the amount of energy required for their power supply. The results obtained in comparison with the socalled brute force method makes it possible to conclude that the Bin packing techniques could help possible to rationalize the use of the physical resources allocated to the operation of the applications in the Datacenters.

                                                                                                                            Copyright © GridCom 2017                                                                                                 Designed By NnN