Resource Matching in Non-dedicated Multicluster Environments ⋆ J. Ll. Lérida 1 , F. Solsona 1 , F. Giné 1 , J.R. García 2 and P. Hernández 2 1 Departamento de Informática e Ingeniería Industrial, Universitat de Lleida, Spain. {jlerida,francesc,sisco}@diei.udl.cat 2 Departamento de Arquitectura y Sistemas Operativos, Universitat Autònoma de Barcelona, Spain. {jrgarcia}@aomail.uab.es, {porfidio.hernandez}@uab.cat Abstract. We are interested in making use of Multiclusters to execute parallel applications. The present work is developed within the M-CISNE project. M- CISNE is a non-dedicated and heterogeneous Multicluster environment which includes MetaLoRaS, a two-level MetaScheduler that manages the appropriate job allocation to available resources. In this paper, we present a new resource-matching model for MetaLoRaS, which is aimed at mitigating the degraded turnaround time of co-allocated jobs, caused by the contention on shared inter-cluster links. The model is linear program- ming based and considers the availability of computational resources and the contention of shared inter and intra-cluster links. Its goal is to minimize the av- erage turnaround time of the parallel applications without disturbing the local applications excessively and maximize the prediction accuracy. We also present a parallel job model that takes both computation and communi- cation characterizations into account. By doing this, greater accuracy is obtained than in other models only focused on one of these characteristics. Our preliminary performance results indicate that the linear programming model for on-line resource matching is efficient in speed and accuracy and can be suc- cessfully applied to co-allocate jobs across different clusters. 1 Introduction A Multicluster system has a network topology made up of interconnected clusters, lim- ited to a campus- or organization-wide network. There are collections of several clusters formed by commodity workstations in many laboratories, Universities, and research centers. The main goal of the present work is to make use of wasted computational resources of non-dedicated and heterogeneous Multiclusters to execute parallel appli- cations efficiently without disturbing the local applications excessively. In order to manage the collective computational power of a Multicluster efficiently, special scheduling mechanisms are required to select and map jobs to available resour- ces. We refer to these schedulers as MetaSchedulers. In general, we consider a Me- taScheduler to be the software that decides where, when, and how to schedule jobs in a Multicluster. In previous works [13,12], we presented MetaLoRaS, an efficient Me- taScheduler made up of a queuing system with two-level hierarchical architecture for ⋆ This work was supported by the MEyC-Spain under contract TIN2007-64974