Conference Paper

DynamicWorkflow Management Using Performance Data

Cardiff University, UK
DOI: 10.1109/CCGRID.2006.37 Conference: Cluster Computing and the Grid, 2006. CCGRID 06. Sixth IEEE International Symposium on, Volume: 1
Source: IEEE Xplore

ABSTRACT

An approach to dynamic workflow management and optimisation using performance data is presented. We discuss strategies for choosing an optimal service (based on user specified criteria) from several semantically equivalent Web Services. Such an approach involves finding "similar" services, by first pruning the set of discovered services based on service metadata, and subsequently selecting an optimal service based on data recorded during prior executions of a service and/or current machine loads. We describe the current implementation of the system, and demonstrate this by a BLAST (used in BioInformatics for Protein-alignment) example by using the Ganglia monitoring tool to get performance data.

0 Followers
 · 
5 Reads
  • Source
    • "It is because the Grid offers a huge amount of heterogeneous computing resources, making the traditional code optimization cycle difficult. Instead, the performance optimization is shifted to Grid middleware services which aim at optimizing the execution of Grid workflows by means of the optimization of resource selections and execution control, and self-adaptive techniques [7] [22] [11] [10] [18]. To support this paradigm shift, we need performance analysis services that are capable of detecting performance problems of Grid workflows during runtime. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Grid workflows are executed on diverse resources whose interactions are highly complicated and hardly predicted. Often the user and the workflow middleware services want to be informed about the performance behavior of workflows, as early as possible, so that they can steer the execution of workflows to compensate the performance loss or execution failures. This paper describes a distributed performance analysis service that supports tracing execution, analyzing performance overheads, and searching for performance problems of Web services-based workflows in the Grid. We present how the user and the Grid workflow middleware can utilize the distributed performance analysis service in order to optimize the execution of workflows.
    Full-text · Article ·
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The Grid computing community is now converging on a service-oriented architecture in which applications are composed from geographically dispersed, interacting Web services, and expressed in a workflow description language, typically based on XML. Workflow techniques generally enable a collection of services to be combined dynamically. However, although there is broad consensus on the overall architecture of the Grid there are many unresolved issues that are still active research areas and for which implementations are not publicly available. Grid computing presents many challenges to both middleware developers and application scientists. Amongst these is the discovery of resources to perform a particular task or application, and where multiple resources perform the same function, selecting the optimal one with respect to a set of user-specified criteria. The importance of such optimisation issues is highlighted in service-rich environments in which an application may be able to select from a number of semantically equivalent services that may be characterised by differing performance, cost, and quality of solution. By definition, such optimisation needs to be undertaken in a dynamic environment, within which resource properties can change. Optimisation therefore must utilise a number of heuristic techniques to enable service selection. The Workflow Optimisation Services for e-Science Applications (WOSE) project has investigated
    Full-text · Article · Jul 2008
  • [Show abstract] [Hide abstract]
    ABSTRACT: Grid workows are executed on diverse resources whose interactions are highly complicated and hardly predicted. Often the user and the workow middleware services want to be informed about the performance behavior of workows, as early as possible, so that they can steer the execution of workows to compensate the performance loss or execution failures. This paper describes a distributed performance analysis service that supports tracing execution, analyzing performance overheads, and searching for performance problems of Web services-based workows in the Grid. We present how the user and the Grid workow middleware can utilize the distributed performance analysis service in order to optimize the execution of workows.
    No preview · Article ·
Show more