Evaluation of Meta-scheduler Architectures and Task Assignment Policies for High Throughput Computing

نویسندگان

  • Eddy Caron
  • Andreï Tsaregorodtsev
  • Vincent Garonne
چکیده

In this paper we present a model and simulator for many clusters of heterogeneous PCs belonging to a local network. These clusters are assumed to be connected to each other through a global network and each cluster is managed via a local scheduler which is shared by many users. We validate our simulator by comparing the experimental and analytical results of a M/M/4 queuing system. These studies indicate that the simulator is consistent. Next, we do the comparison with a real batch system and we obtain an average error of 10.5% for the response time and 12% for the makespan. We conclude that the simulator is realistic and well describes the behaviour of a large-scale system. Thus we can study the scheduling of our system called DIRAC in a high throughput context. We justify our decentralized, adaptive and opportunistic approach in comparison to a centralized approach in such a context. Key-words: Simulation, Model, Multi-clusters platform, Meta-scheduling, Grid Computing This text is also available as a research report of the Laboratoire de l’Informatique du Parallélisme http://www.ens-lyon.fr/LIP. Évaluation d’architectures de Meta-ordonnanceur et politique de distribution de tâches pour le calcul haute performance Résumé : Dans cet article, nous présentons une modélisation et un simulateur de grands systèmes de calcul distribué. Une telle plate-forme se compose de grappes de PCs hétérogènes appartenant à un réseau local inter-connectées entre elles par un réseau global. Ces grappes sont accessibles via un ordonnanceur local et sont partagées entre les utilisateurs. La confrontation du simulateur avec les résultats théoriques d’un système M/M/4 nous permet de conclure qu’il est analytiquement valide. Une deuxième confrontation avec un système batch réel, nous donne une différence moyenne de 10.5 % par rapport à la réalité pour les temps de réponse et de 12% pour le makespan. Notre simulateur est donc réaliste et décrit le comportement d’un système de batch réel. Fort de cet outil, nous avons analysé l’ordonnancement de notre système (appelé DIRAC) dans un contexte de calcul intensif. Nous avons justifié l’approche distribuée, adaptative et opportuniste utilisée dans notre système par rapport à une approche centralisée. Mots-clés : Simulation, Modélisation, Plate-forme multi-grappes, Calcul sur la Grille Évaluation of Meta-scheduler architectures ... 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

Hybrid Meta-heuristic Algorithm for Task Assignment Problem

Task assignment problem (TAP) involves assigning a number of tasks to a number of processors in distributed computing systems and its objective is to minimize the sum of the total execution and communication costs, subject to all of the resource constraints. TAP is a combinatorial optimization problem and NP-complete. This paper proposes a hybrid meta-heuristic algorithm for solving TAP in a ...

متن کامل

An Efficient Genetic Algorithm for Task Scheduling on Heterogeneous Computing Systems Based on TRIZ

An efficient assignment and scheduling of tasks is one of the key elements in effective utilization of heterogeneous multiprocessor systems. The task scheduling problem has been proven to be NP-hard is the reason why we used meta-heuristic methods for finding a suboptimal schedule. In this paper we proposed a new approach using TRIZ (specially 40 inventive principles). The basic idea of thi...

متن کامل

An Efficient Genetic Algorithm for Task Scheduling on Heterogeneous Computing Systems Based on TRIZ

An efficient assignment and scheduling of tasks is one of the key elements in effective utilization of heterogeneous multiprocessor systems. The task scheduling problem has been proven to be NP-hard is the reason why we used meta-heuristic methods for finding a suboptimal schedule. In this paper we proposed a new approach using TRIZ (specially 40 inventive principles). The basic idea of thi...

متن کامل

A Throughput Driven Task Scheduler for Batch Jobs in Shared MapReduce Environments

MapReduce is one of the most popular parallel data processing systems, and it has been widely used in many fields. As one of the most important techniques in MapReduce, task scheduling strategy is directly related to the system performance. However, in multi-user shared MapReduce environments, the existing task scheduling algorithms cannot provide high system throughput when processing batch jo...

متن کامل

Starvation Free Scheduler for Buffered Crossbar Switches (RESEARCH NOTE)

Need for high speed internet connectivity has lead to a substantial research in switching systems. Buffered crossbar switches have received a lot of attention from both research and industrial communities due of its flexibility and scalability. Designing a scheduling algorithm for buffered crossbar switches without starvation is a major challenge as of now. In this paper, we proposed a Delay ba...

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 2005