DIET-ethic: Fair Scheduling of Optional Computations in GridRPC Middleware

نویسندگان

  • Frédéric CAMILLO
  • Eddy CARON
  • Ronan GUIVARCH
  • Aurélie HURAULT
  • Cristian KLEIN
  • Christian PÉREZ
چکیده

Most HPC platforms require users to submit a pre-determined number of computation requests (also called jobs). Unfortunately, this is cumbersome when some of the computations are optional, i.e., they are not critical, but their completion would improve results. For example, given a deadline, the number of requests to submit for a Monte Carlo experiment is difficult to choose. The more requests are completed, the better the results are, however, submitting too many might overload the platform. Conversely, submitting too few requests may leave resources unused and misses an opportunity to improve the results. This paper introduces and solves the problem of scheduling optional computations. An architecture which auto-tunes the number of requests is proposed, then implemented in the DIET GridRPC middleware. Real-life experiments show that several metrics are improved, such as user satisfaction, fairness and the number of completed requests. Moreover, the solution is shown to be scalable. Key-words: HPC; GridRPC; malleable applications; Grid’5000 ha l-0 06 96 10 4, v er si on 2 18 M ay 2 01 2 Diet-ethic: Ordonnancement équitable des calculs optionnels dans un intergiciel GridRPC Résumé : La plupart des plate-formes HPC demandent à l’utilisateur de soumettre un nombre pré-déterminé de requêtes de calcul (aussi appelées « job »). Malheureusement, cela n’est pas pertinent quand une partie des calculs est optionnelle, c’est-à-dire, que l’exécution des requêtes n’est pas critique pour l’utilisateur, mais que leur complétion pourrait améliorer les résultats. Par exemple, étant donnée une date limite, le nombre de requêtes à soumettre pour une expérience Monte Carlo est difficile à choisir. Plus il y a des requêtes qui sont exécutées, meilleures sont les résultats. Cependant, en soumettant trop de requêtes, on risque de surcharger la plate-forme. À l’opposé, en ne soumettant pas assez de requêtes, les ressources sont sous-exploitées alors qu’elles auraient pu être utilisées pour améliorer les résultats. Cet article introduit et résout le problème d’ordonnancer des requêtes optionnelles. Une architecture qui choisit automatiquement le nombre de requêtes est proposée puis implémentée dans l’intergiciel GridRPC DIET. Les expériences faites sur de vraies plate-formes – telles que Grid’5000 – montrent que plusieurs métriques peuvent être améliorées, telles que la satisfaction des utilisateurs, l’équité et le nombre des requêtes exécutées. Enfin, la solution proposée passe à l’échelle. Mots-clés : HPC; GridRPC; applications malleables; Grid’5000 ha l-0 06 96 10 4, v er si on 2 18 M ay 2 01 2 Diet-ethic: Fair Scheduling of Optional Computations in GridRPC Middleware 3

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Evaluation of the OGF GridRPC Data Management library, and study of its integration into an International Sparse Linear Algebra Expert System

Gaël Le MAHEC Université de Picardie Jules Verne E-mail: [email protected] The Data Management API for the GridRPC describes an optional API that extends the GridRPC standard. It provides a minimal subset of functions to handle a large set of data operations, among which movement, replication, migration and stickyness. We already showed that its use leads to 1) reduced time to complet...

متن کامل

Diet: A Scalable Toolbox to Build Network Enabled Servers on the Grid

Among existing grid middleware approaches, one simple, powerful, and flexible approach consists of using servers available in different administrative domains through the classical client-server or Remote Procedure Call (RPC) paradigm. Network Enabled Servers implement this model also called GridRPC. Clients submit computation requests to a scheduler whose goal is to find a server available on ...

متن کامل

Parallelization of the Treecode Algorithm for N-Body Simulation Using MPI, Hybrid, and GridRPC Programming Paradigms

This dissertation describes the parallelization of the treecode algorithm for N-Body problem and performance comparison among three different parallel programming paradigms, MPI, hybrid MPI-OpenMP, and GridRPC. In N-Body simulation, the specific routine for calculating the forces on the bodies which accounts for upwards of 90% of the cycles in typical computations is eminently suitable for obta...

متن کامل

Tunable scheduling in a GridRPC framework

Among existing grid middleware approaches, one simple, powerful, and flexible approach consists of using servers available in different administrative domains through the classic client-server or Remote Procedure Call (RPC) paradigm. Network Enabled Servers (NES) implement this model also called GridRPC. Clients submit computation requests to a scheduler whose goal is to find a server available...

متن کامل

Diet: New Developments and Recent Results

Among existing grid middleware approaches, one simple, powerful, and flexible approach consists of using servers available in different administrative domains through the classic client-server or Remote Procedure Call (RPC) paradigm. Network Enabled Servers (NES) implement this model also called GridRPC. Clients submit computation requests to a scheduler whose goal is to find a server available...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 2012