Wildfalcon

Books

1. Component Models and Systems for Grid Applications
   Volume 1 of CoreGRID series, Chapter ICENI: An Integrated Grid Middleware to Support e-Science
   Workflow specification and enactment is a critical operation in e-science. In this paper we describe how an abstract workflow, specified by an end-user in the form of an Execution Plan, is instantiated within the ICENI environment through the enactment pipeline. The pipeline starts with the workflow specification, includes the mapping of work onto resources through a workflow enabled scheduler which is able to make use of performance results captured from previous executions within the ICENI environment, and ends with the orchestration of the concrete execution plan on the specified collection of distributed resources. We show that ICENI is capable of deploying the specified components of the workflow over the available resources.

Papers

1. Making the Grid Predictable through Reservations and Performance Modelling
   The Computer Journal: 48(3)
   Workflow specification and enactment is a critical operation in e-science. In this paper we describe how an abstract workflow, specified by an end-user in the form of an Execution Plan, is instantiated within the ICENI environment through the enactment pipeline. The pipeline starts with the workflow specification, includes the mapping of work onto resources through a workflow enabled scheduler which is able to make use of performance results captured from previous executions within the ICENI environment, and ends with the orchestration of the concrete execution plan on the specified collection of distributed resources. We show that ICENI is capable of deploying the specified components of the workflow over the available resources.
   Available on The Computer Journal website
2. Workflow Enactment in ICENI
   Workflow specification and enactment is a critical operation in e-science. In this paper we describe how an abstract workflow, specified by an end-user in the form of an Execution Plan, is instantiated within the ICENI environment through the enactment pipeline. The pipeline starts with the workflow specification, includes the mapping of work onto resources through a workflow enabled scheduler which is able to make use of performance results captured from previous executions within the ICENI environment, and ends with the orchestration of the concrete execution plan on the specified collection of distributed resources. We show that ICENI is capable of deploying the specified components of the workflow over the available resources.
   View as PDF
3. Performance Architecture within ICENI
   This paper describes the architecture built into the Imperial College e-Science Infrastructure (ICENI) for handling performance meta-data. The architecture provides a means to gathering performance information, processing this information to populate the performance store, and to use this performance information to aid in the selection of resources and component implementations. The Performance Framework is developed in a .pluggable. manner allowing alternate implementations of the three main features to be used. Performance Stores may be either data stores or based on analytical models.
   View as PDF
4. Transfer Report
   We present an account of scheduling component based applications to run on a computational Grid. We discuss the structure of a Grid, where scheduling fits within the architecture and examine the aims and purpose of a scheduler. A classification for Grid schedulers is given, and applied to a number of existing scheduling algorithms. We examine the scheduling service within the ICENI Grid Middleware and present scheduling algorithms based on random scheduling, exhaustive search scheduling, simulated annealing and game theory. Experimental analysis of the effectiveness of these algorithms is given, which concludes that a simulated annealing algorithm is generally preferable, and that in some situations increasing the size of a Grid decreases its effectiveness. We examine the role of scheduling in a wider Grid context, involving multiple organisations and machines, each exhibiting stochastic behaviour. We finish with a discussion of future work which we would like to examine in order to improve the quality of Grid scheduling.
   View as PDF
5. Scheduling Architecture and Algorithms within the ICENI Grid Middleware
   All Hands Meeting - Sept 03
   The ICENI Grid Middleware system enables the use of component based applications in a distributed heterogeneous environment and is implemented as a layered service oriented architecture. The scheduling of component applications across this environment takes place through the ICENI Scheduling Framework which allows pluggable. scheduling algorithms and platform specific launchers. The architecture of the Scheduling Framework is described, which decouples the scheduling algorithm from both the application and the architecture. Four scheduling algorithms (game theory, simulated annealing, random and best of n random) are presented and compared under the criteria of time and cost optimisation. We show that the simulated annealing scheduler produces more optimal results in the case of both time and cost optimisation.
   View as PDF
6. Meaning and Behaviour in Grid Oriented Components:
   The ICENI middleware utilises information captured within a component based application in order to facilitate Grid-based scheduling. We describe a system of application related meta-data that features a separation of concerns between meaning, behaviour and implementation, which allows for both communication and implementation selection at run-time, while providing the user with a flow-based programming model. It is shown that this separation enables a flexible approach to scheduling, and eases the integration of components with disparate control flow patterns or data types, by means of converters and tees for collective communication. By explicitly recording application information and supporting multiple scheduling approaches, communication protocols and component applications, while retaining OGSA compatibility, the ICENI component model is ideally suited to Grid computing.
   View as PDF
7. 9 Month PhD Report:
   This report summarise the work and progress that has been made in nine months of research in the area of Grid Resource Allocation and Grid Scheduling. It contains an overview of the area, and contributes a mathematical formalisation for exploring the suitability of a schedule. It is proposed to use this formalisation in scheduling algorithms, due to be developed as ongoing work. the work also looks briefly at research in connected areas, such as scheduler architecture and scheduler simulation, as these areas will be of importance to future work on this project.
   View as PDF

Presentations

1. Transfer Report
   We present an account of scheduling component based applications to run on a computational Grid. We discuss the structure of a Grid, where scheduling fits within the architecture and examine the aims and purpose of a scheduler. A classification for Grid schedulers is given, and applied to a number of existing scheduling algorithms. We examine the scheduling service within the ICENI Grid Middleware and present scheduling algorithms based on random scheduling, exhaustive search scheduling, simulated annealing and game theory. Experimental analysis of the effectiveness of these algorithms is given, which concludes that a simulated annealing algorithm is generally preferable, and that in some situations increasing the size of a Grid decreases its effectiveness. We examine the role of scheduling in a wider Grid context, involving multiple organisations and machines, each exhibiting stochastic behaviour. We finish with a discussion of future work which we would like to examine in order to improve the quality of Grid scheduling.
   View as PPT
2. Scheduling Architecture and Algorithms within the ICENI Grid Middleware
   All Hands Meeting - Sept 03
   The ICENI Grid Middleware system enables the use of component based applications in a distributed heterogeneous environment and is implemented as a layered service oriented architecture. The scheduling of component applications across this environment takes place through the ICENI Scheduling Framework which allows pluggable. scheduling algorithms and platform specific launchers. The architecture of the Scheduling Framework is described, which decouples the scheduling algorithm from both the application and the architecture. Four scheduling algorithms (game theory, simulated annealing, random and best of n random) are presented and compared under the criteria of time and cost optimisation. We show that the simulated annealing scheduler produces more optimal results in the case of both time and cost optimisation.
   View as PPT
3. ISN Scheduling Workshop - 12th May 2003:
   Componentised applications can be used in a grid environment. This provides the benefit of allowing reusable generic components, as well as allowing the composite application to be split over multiple compute resources, thus increasing the performance of the application. However this approach, where multiple components can each run on any one of multiple resources, drastically increases the search space when looking for the most optimal solution. Two heuristic algorithms, those of simulated annealing and game theory have been implemented and compared with two benchmark algorithms; random scheduling and exhaustive search, showing that results with the same or comparable quality as an exhaustive search, can be obtained in a significantly shorter time. Empirical data showing the scalability of both execution time and the quality of the result is presented for both algorithms.
   View as PPT
4. 9 Month PhD Report:
   This presentations summarises the work and progress that has been made in nine months of research in the area of Grid Resource Allocation and Grid Scheduling. It contributes a mathematical formalisation for exploring the suitability of a schedule. It is proposed to use this formalisation in scheduling algorithms, due to be developed as ongoing work. the work also looks briefly at research in connected areas, such as scheduler architecture and scheduler simulation, as these areas will be of importance to future work on this project.
   View as PPT