Evaluation of Reallocation Heuristics for Moldable Tasks in Computational Dedicated and non Dedicated Grids
نویسندگان
چکیده
Grid services often consist of remote sequential or rigid parallel application executions. However, moldable parallel applications, linear algebra solvers for example, are of great interest but requires dynamic tuning which has mostly to be done interactively if performances are needed. Thus, their grid execution depends on a remote and transparent submission to a possibly different batch scheduler on each site, and means an automatic tuning of the job according to the local load. In this report we study the benefits of having a middleware able to automatically submit and reallocate requests from one site to another when it is also able to configure the services by tuning their number of processors and their walltime. In this context, we evaluate the benefits of such mechanisms on four multi-cluster Grid setups, where the platform is either composed of several heterogeneous or homogeneous, dedicated or non dedicated clusters. Different scenarios are explored using simulations of real cluster traces from different origins. Results show that a simple scheduling heuristic is good and often the best. Indeed, it is faster and thus can take more jobs into account while having a small execution time. Moreover, users can expect more jobs finishing sooner and a gain on the average job response time between 10% and 40% in most cases if this reallocation mechanism combined to auto-tuning capabilities is implemented in a Grid framework. The implementation and This text is also available as a research report of the Laboratoire de l’Informatique du Parallélisme http://www.ens-lyon.fr/LIP. in ria -0 05 09 39 3, v er si on 1 12 A ug 2 01 0 2 Y. Caniou, G. Charrier, F. Desprez the maintenance of this heuristic coupled to the migration mechanism in a Grid middleware is also simpler because less transfers are involved. Key-words: Reallocation, batch schedulers, meta-schedulers, computational grids, moldable tasks INRIA in ria -0 05 09 39 3, v er si on 1 12 A ug 2 01 0 Evaluation d’heuristiques de réallocation de tâches moldables dans les grilles de calculs dédiées et non dédiées Résumé : L’appel à des services présents sur les grilles de calcul correspondent généralement à l’exécution d’une application séquentielle ou rigide. Cependant, il est possible d’avoir des applications parallèles moldables, telles que des solveurs linéaires, qui sont d’un grand intérêt, mais qui demandent une adaptation dynamique pour obtenir de bonnes performances. Leur exécution nécessite donc d’avoir un accès distant et transparent à différents gestionnaires de ressources, demandant donc une adaptation automatique de l’application en fonction de la charge locale. Dans ce rapport, nous étudions les bénéfices découlant de l’utilisation d’un intergiciel de grille capable de soumettre et de réallouer des requêtes d’un site à l’autre tout en configurant automatiquement les services en choisissant le nombre de processeurs ainsi que la durée d’exécution estimée. Dans ce contexte, nous évaluons les gains apportés par de tels mécanismes sur quatre grilles de calcul différentes où la plate-forme est composée de plusieurs grappes, homogène ou hétérogènes, dédiées ou non. Nous explorons différents scénarios par la simulation de traces de tâches provenant de réelles exécutions. Les résultats montrent que l’utilisation d’une heuristique d’ordonnancement simple est efficace, souvent amplement suffisante, voire la meilleure. En effet, elle est plus rapide à l’exécution et permet de prendre plus de requêtes en compte. Les utilisateurs peuvent espérer une majorité de requêtes terminant plus tôt si elle est utilisée, ainsi qu’une réduction du temps d’attente du résultat d’entre 10% et 40% dans la plupart des cas lorsque le mécanisme de réallocation couplé à l’adaptation automatique sont présents dans l’intergiciel. De plus, l’implantation et la maintenance de cette heuristique couplée au mécanisme de migration de tâches dans un intergiciel de grille est aussi plus facile car moins de tranferts sont nécessaires. Mots-clés : Réallocation, ordonnanceurs batch, méta-ordonnanceurs, grilles de calcul, tâches moldables in ria -0 05 09 39 3, v er si on 1 12 A ug 2 01 0 4 Y. Caniou, G. Charrier, F. Desprez
منابع مشابه
Effective heuristics and meta-heuristics for the quadratic assignment problem with tuned parameters and analytical comparisons
Quadratic assignment problem (QAP) is a well-known problem in the facility location and layout. It belongs to the NP-complete class. There are many heuristic and meta-heuristic methods, which are presented for QAP in the literature. In this paper, we applied 2-opt, greedy 2-opt, 3-opt, greedy 3-opt, and VNZ as heuristic methods and tabu search (TS), simulated annealing, and pa...
متن کاملOn the Distribution of Dependent Tasks Over Non-dedicated Grids with High Bandwidth Links
The distribution of workflow-like dependent tasks over a set of heterogeneous resources is an NP-Complete problem. Thus, many solutions are heuristic-based, leading to simulations for the validation and comparison of dependent task scheduling algorithms. However, simulations concerning the algorithms’ performance when executed in non-dedicated resources with a wide range of bandwidth capacities...
متن کاملScheduling Problem of Virtual Cellular Manufacturing Systems (VCMS); Using Simulated Annealing and Genetic Algorithm based Heuristics
In this paper, we present a simulated annealing (SA) and a genetic algorithm (GA) based on heuristics for scheduling problem of jobs in virtual cellular manufacturing systems. A virtual manufacturing cell (VMC) is a group of resources that is dedicated to the manufacturing of a part family. Although this grouping is not reflected in the physical structure of the manufacturing system, but machin...
متن کاملEvaluation of dose distribution and dose gradient in brachytherapy cylindrical applicators using a dedicated Phantom for Iridium-192 and Cobalt-60 HDR sources
Introduction: A study was performed to evaluate radiation dose distribution and dose gradient around cylindrical applicators for high-dose-rate (HDR) brachytherapy systems with 192Ir, 60Co brachytherapy source applied for rectal and vaginal cancers treatments. Materials and Methods: Two additional computed tomography (CT) based plans were generated using a ...
متن کاملSelf-adaptive skeletal task farm for computational grids
In this work, we introduce a self-adaptive task farm for computational grids which is based on a single-round scheduling algorithm called dynamic deal. In principle, the dynamic deal approach employs skeletal forecasting information to automatically instrument the task farm scheduling and determine the amount of work assigned to each worker at execution time, allowing the farm to adapt effectiv...
متن کامل