Jump label

Service navigation

Main navigation

You are here:

Sub navigation

Main content

Resource Management

 

Overview

Research Areas

Our group's interests lie in various areas, all related to advanced management strategies of services and resources. Although we are firmly commmitted to fundamental research in this area, we also have a tradition of being strongly involved in various hands-on projects and the development of applied solutions. Following, we give an overview of the main scopes of our work.

  • Grid Computing

    Regarding Grid Computing, we primarily work on Grid brokerage systems for the coordinated orchestration of computation, data and network requirements , developing strategies and implementations for complex environments with demanding applications. Moreover, we have great expertise in the management of High Performance Computing (HPC) resources.

  • Scheduling

    Here, our focus lies in the theoretical analysis and development of new approximation algorithms for practical scheduling problems. Additionally, we work on user group specific scheduling strategies and comprehensive workload models for the online scheduling of Massively Parallel Processing (MPP) systems on the basis of Computational Intelligence.

  • Service Infrastructures

    In this area, we are mainly interested in the development of advanced management strategies for Service Level Agreements (SLAs), including the automatic negotiation of QoS levels between service clients and suppliers. Furthermore, we work on the applicability of Service Oriented Architectures (SOAs) in different loosely federated, but highly distributed infrastructures.

  • Virtualization

    Here, we are evaluating the performance of virtual machines (VMs) on different hardware archictectures with a special focus on good allocation strategies for VMs for load-balanced systems.

Hardware Environment

Our institute is currently running a 128 node (Pentium IV at 2.8 GHz with 1024 MB RAM each) cluster on Linux and PBS Pro, which is mainly used for simulation purposes in the area of scheduler evaluation and workload modeling.

Furthermore, a new 200 node system (Xeon DP at 2.0 GHz with 4096 MB RAM each) is available since Q1/2007 for the D-Grid community. This system integrates various Grid middleware solutions (gLite, Globus Toolkit and UNICORE) into a cooperatively scheduled HPC environment.