SEMINARS

Month Week Day
August 2017
Monday Tuesday Wednesday Thursday Friday Saturday Sunday
  • June 29, 2017 @ Bâtiment IMAG (404) -- A stochastic approach for optimizing green energy consumption in distributed clouds by Fanny Dufossé (Inria)

    A stochastic approach for optimizing green energy consumption in distributed clouds

    The energy drawn by Cloud data centers is reaching worrying levels, thus inciting providers to install on-site green energy producers, such as photovoltaic panels. Considering distributed Clouds, workload managers need to geographically allocate virtual machines according to the green production in order not to waste energy. In this paper, we propose SAGITTA: a Stochastic Approach for Green consumption In disTributed daTA centers. We show that compared to the optimal solution, SAGITTA consumes 4% more brown energy, and wastes only 3.14% of the available green energy, while a traditional round-robin solution consumes 14.4% more energy overall than optimum, and wastes 28.83% of the available green energy.

  • July 6, 2017 @ Bâtiment IMAG (406) -- Séminaire Josu Doncel : Under-Approximation Computation Through Optimal Control

    Title: Under-Approximation Computation Through Optimal Control

    Abstract: Under-approximation provides a subset of the reachable set of an uncertain dynamical system which can then be used to formally falsify properties of quantitative models. Using Pontryagin’s principle, our approach computes an under-approximation for a linear combination of state variables of nonlinear ordinary differential equations and time-varying uncertainties. By a numerical comparison against state-of-the-art tools Flow^∗ and CORA, we show that our methodology provides tight under-approximations in benchmarks, and that it can scale to models that are out of reach with these over-approximation techniques.

  • July 6, 2017 @ Bâtiment IMAG (406) -- I/O performance for HPC: finding the right access pattern and avoiding interference by Francieli Zanon-Boito

    Title:
    I/O performance for HPC: finding the right access pattern and avoiding interference

    Abstract:
    Scientific applications are executed in a high performance computing (HPC) environment, where a parallel file system (PFS) provides access to a shared storage infrastructure. The key characteristic of these systems is the use of multiple storage servers, from where data can be obtained by the clients in parallel. The performance observed by applications when accessing a PFS is directly affected by the way they perform this access, i.e. their access pattern.
    In this seminar, I'll discuss my work with the Ondes3D seismic simulation, which was focused into changing the application's access pattern to improve I/O performance without changing the output format. Moreover, I'll discuss my previous and current work on I/O scheduling at different levels of the I/O stack, pointing current challenges for future work.

  • September 7, 2017 @ -- keynote LIG
  • September 14, 2017 @ Bâtiment IMAG (406) -- Computing with coins (by Jean-Marc Vincent)

    En cette rentrée et pour nous remettre les idées bien en place, Jean-Marc nous expliquera tout ce que nous voulons savoir mais n'avons jamais osé demander sur les aiguilles de Buffon, les urnes de Polya, et d'autres questions fondamentales.

  • September 21, 2017 @ Bâtiment IMAG (406) -- Kleinberg 's Grid Unchained (by Fabien Mathieu, Nokia)

    One of the key features of small-worlds is the ability to route messages with few hops only using local knowledge of the topology. In 2000, Kleinberg proposed a model based on an augmented grid that asymptotically exhibits such property.

    In this paper, we propose to revisit the original model from a simulation-based perspective. Our approach is fueled by a new algorithm that can draw an augmenting link in Õ(1).

    The resulting speed gain enables detailed numerical evaluations. We show for example that in practice, the augmented scheme proposed by Kleinberg is more robust than predicted by the asymptotic behavior, even for very large finite grids. We also propose tighter bounds on the performance of Kleinberg's routing algorithm. At last, we show that, fed with realistic parameters, the model gives results in line with real-life experiments.

Comments are closed