Henan Zhao

A hybrid heuristic for DAG scheduling on heterogeneous systems

Summary form only given. This paper is motivated by the observation that different methods to compute the weights of nodes and edges when scheduling DAGs onto heterogeneous machines may lead to significant variations in the generated schedule. To minimize such variations, we present a novel heuristic for DAG scheduling, which is based upon solving a series of independent task scheduling problems. A novel heuristic for the latter problem is also included. Both heuristics compare favourably with other related heuristics.

Scheduling workflows with budget constraints

Grids are emerging as a promising solution for resource and computation demanding applications. However, the heterogeneity of resources in Grid computing, complicates resource management and scheduling of applications. In addition, the commercialization of the Grid requires policies that can take into account user requirements, and budget considerations in particular. This paper considers a basic model for workflow applications modelled as Directed Acyclic Graphs (DAGs) and investigates heuristics that allow to schedule the nodes of the DAG (or tasks of a workflow) onto resources in a way that satisfies a budget constraint and is still optimized for overall time. Two different approaches are implemented, evaluated and presented using four different types of basic DAGs.

Scheduling multiple DAGs onto heterogeneous systems

The problem of scheduling a single DAG onto heterogeneous systems has been studied extensively. In this paper, we focus on the problem of scheduling more than one DAG at the same time onto a set of heterogeneous resources. The aim is not only to optimize the overall makespan, but also to achieve fairness, defined on the basis of the slowdown that each DAG would experience as a result of competing for resources with other DAGs. Two policies particularly focussing to deliver fairness are presented and evaluated along with another four policies that can be used to schedule multiple DAGs.

A low-cost rescheduling policy for efficient mapping of workflows on grid systems

Workflow management is emerging as an important service in Grid computing. A simple model that can be used for the representation of certain workflows is a directed acyclic graph. Although many heuristics have been proposed to schedule such graphs on heterogeneous environments, most of them assume accurate prediction of computation and communication costs. This limits their direct applicability to a dynamically changing environment, such as the Grid. In this environment, an initial schedule may be built based on estimates, but run-time rescheduling may be needed to improve application performance. This paper presents a low-cost rescheduling policy, which considers rescheduling at a few, carefully selected points during the execution. This policy achieves performance results, which are comparable with those achieved by a policy that dynamically attempts to reschedule before the execution of every task.

Scheduling Data-IntensiveWorkflows onto Storage-Constrained Distributed Resources

In this paper we examine the issue of optimizing disk usage and of scheduling large-scale scientific workflows onto distributed resources where the workflows are data- intensive, requiring large amounts of data storage, and where the resources have limited storage resources. Our approach is two-fold: we minimize the amount of space a workflow requires during execution by removing data files at runtime when they are no longer required and we schedule the workflows in a way that assures that the amount of data required and generated by the workflow fits onto the individual resources. For a workflow used by gravitational- wave physicists, we were able to improve the amount of storage required by the workflow by up to 57 %. We also designed an algorithm that can not only find feasible solutions for workflow task assignment to resources in disk- space constrained environments, but can also improve the overall workflow performance.

An Experimental Investigation into the Rank Function of the Heterogeneous Earliest Finish Time Scheduling Algorithm

This paper considers the Heterogeneous Earliest Finish Time (HEFT) algorithm for scheduling the tasks of an application, represented by a directed acyclic graph, onto a bounded number of heterogeneous machines. We focus on the appropriate selection of the weight for the nodes and edges of the graph, and experiment with a number of different schemes for computing these weights. Our findings indicate that the length of the schedule produced may be affected significantly by the scheme used, and suggest that the mean value based approach used by HEFT may not be a particularly good choice.

Advance reservation policies for workflows

Advance reservation of resources has been suggested as a means to provide a certain level of support that meets user expectations with respect to specific job start times in parallel systems. Those expectations may relate to a single job application or an application that consists of a collection of dependent jobs. In the context of Grid computing, applications consisting of dependent tasks become increasingly important, usually known as workflows. This paper focuses on the problem of planning advance reservations for individual tasks of workflow-type of applications when the user specifies a requirement only for the whole workflow application. Two policies to automate advance reservation planning for individual tasks efficiently are presented and evaluated.

Optimizing workflow data footprint

In this paper we examine the issue of optimizing disk usage and scheduling large-scale scientific workflows onto distributed resources where the workflows are data-intensive, requiring large amounts of data storage, and the resources have limited storage resources. Our approach is two-fold: we minimize the amount of space a workflow requires during execution by removing data files at runtime when they are no longer needed and we demonstrate that workflows may have to be restructured to reduce the overall data footprint of the workflow. We show the results of our data management and workflow restructuring solutions using a Laser Interferometer Gravitational-Wave Observatory (LIGO) application and an astronomy application, Montage, running on a large-scale production grid-the Open Science Grid. We show that although reducing the data footprint of Montage by 48% can be achieved with dynamic data cleanup techniques, LIGO Scientific Collaboration workflows require additional restructuring to achieve a 56% reduction in data space usage. We also examine the cost of the workflow restructuring in terms of the applications runtime.

A Low-Cost Rescheduling Policy for Dependent Tasks on Grid Computing Systems

A simple model that can be used for the representation of certain workflows is a directed acyclic graph. Although many heuristics have been proposed to schedule such graphs on heterogeneous environments, most of them assume accurate prediction of computation and communication costs; this limits their direct applicability to a dynamically changing environment, such as the Grid. To deal with this, run-time rescheduling may be needed to improve application performance. This paper presents a low-cost rescheduling policy, which considers rescheduling at a few, carefully selected points in the execution. Yet, this policy achieves performance results, which are comparable with those achieved by a policy that dynamically attempts to reschedule before the execution of every task.

Mapping Workflows on Grid Resources: Experiments with the Montage Workflow

Scientific workflows have received considerable attention in Grid computing. This paper is concerned with the issue of scheduling scientific workflows and, by considering a commonly used astronomy workflow, Montage, investigates the impact of different strategies to schedule the workflow graph. Our experiments suggest that the rather regular and symmetric nature of the Montage graph allows rather simple to implement scheduling heuristics that do not take into account the whole structure of the graph, such as Min-min, to deliver competitive performance in most cases of interest. The results support the view that sophisticated graph scheduling heuristics may not be always a prerequisite for good performance in workflow execution. Instead, mechanisms to deal with uncertainties in execution time may be of comparatively higher importance.