Quality of service

QoS for Service Based Workflow on Grid
L Guo, A S McGough, A Akram, D Colling, J Martyniak, M Krznaric London e-Science Centre Imperial College London London, UK {asm, aakram, liguo}@inf.ed.ac.uk {d.colling, janusz.martyniak}@imperial.ac.uk

Abstract
As the main computing paradigm for resource-intensive scientific applications, Grid enables resource sharing and dynamic allocation ofcomputational resources, promotes access to distributed data, operational flexibility and collaboration, and allows service providers to distribute both conceptually and physically to meeting different requirements. Large-scale grids are normally composed of huge numbers of components from different sites. This increases the requirements of workflows and Quality of Service (QoS) upon these workflows as many ofthese components have real-time requirements. In this paper, we describe a QoS-aware workflow management system(WfMS) from GridCC project[7] and show how our WfMS ensures workflows meet the pre-defined QoS requirements and optimise them accordingly. keywords: Grid, Quality of Service (QoS), Web Services, Service Oriented Architecture (SOA)

1. Introduction
In order to perform constructive sciencea scientist will in general need to perform multiple tasks in order to achieve their goal. These may include such things as configuring an instrument, collecting and storing relevant data from the instrument, processing of this information and potentially further iterations of these tasks. This can be seen as a set of tasks which interact with each other in order to achieve the final result and canbe considered as a workflows. Workflow management systems (WfMSs) have been used to support various types of e-science workflows on Grid for a while and as the Grid has evolved into a ServiceOrientated Architecture (SOA)[2] with Web Services[3] emerging as the de-facto communication mechanism. Worflow languages such as BPEL[4], WS-Choreography[5] are powerful languages for developing workflows based onWeb Services. However, the development and execution of workflows within the Grid is a complex process due to the mechanisms used to describe them and the Web Services they are based upon. The selection of the best resources to use within the Grid is complicated due to its

dynamic nature with resources appearing and disappearing without notice and the load on these resources changingdramatically over time. These are issues that the scientist will, in general, not wish to be concerned about. Quality of Service (QoS) constraints for interactions between the different Grid components thus becomes crucial to enable resources oriented workflows. In such workflow processes, service providers and consumers define a binding agreement or contract between the two parties, specifying quality ofservice (QoS) properties such as response times of particular resources such as instruments elements, available storage elements (disk space, memory) for certain tasks, etc. Management of such QoS directly impacts success of parties’ participating in the coordination processes. A good management of quality leads to the creation of quality products and services, which in turn fulfills users’expectations and achieves their satisfaction. Therefore, when services are created or managed using workflow processes, the underlying WfMS should be able to accept, monitor, and control the QoS provided by users. In this paper we present those parts of the GRIDCC[7]

architecture responsible for providing QoS support in WfMS to show how QoS component can be possibly used in Grid environment. In Section 2we present related work. Section 3 gives the brief analysis of Grid Qos requirements. The architecture our QoS based WfMS is explained in Section 4. Section 4 also contains a more detailed breakdown of our system components: the workflow editor-section 4.1, performance repository-section 4.2, planner-section 4.3 and observer-section 4.4. We conclude our work in Section 5.

3. Workflow with...