Ensembles of distributed, heterogeneous resources, also known as Computational Grids are emerging as critical platforms for high-performance and resource-intensive applications. Such platforms provide the potential for applications to aggregate enormous bandwidth, computational power, memory, secondary storage, and other resources during a single execution. However, achieving this performance potential in dynamic, heterogeneous environments is challenging. Recent experience with distributed applications indicates that adaptivity is fundamental to achieving application performance in dynamic Grid environments. The AppLeS (Application Level Scheduling) project provides a methodology, application software, and software environments for adaptively scheduling and deploying applications in dynamic, heterogeneous, multi-user Grid environments. In this paper, we discuss the AppLeS project and outline our results.

There is a current need for scheduling policies that can leverage the performance variability of resources on multiuser clusters. We develop one solution to this problem called stochastic scheduling that utilizes a distribution of application execution performance on the target resources to determine a performance-efficient schedule. In this paper, we define a stochastic scheduling policy based on time-balancing for data parallel applications whose execution behavior can be represented as a normal distribution. Using three distributed applications on two contended platforms, we demonstrate that a stochastic scheduling policy can achieve good and predictable performance for the application as evaluated by several performance measures.

In heterogeneous and dynamic environments, efficient execution of parallel computations can require mappings of tasks to processors whose performance is both irregular (because of heterogeneity) and time-varying (because of dynamicity). While adaptive domain decomposition techniques have been used to address heterogeneous resource capabilities, temporal variations in those capabilities have seldom been considered. We propose a conservative scheduling policy that uses information about expected future variance in resource capabilities to produce more efficient data mapping decisions. We first present techniques, based on time series predictors that we developed in previous work, for predicting CPU load at some future time point, average CPU load for some future time interval, and variation of CPU load over some future time interval. We then present a family of stochastic scheduling algorithms that exploit such predictions of future availability and variability when making data mapping decisions. Finally, we describe experiments in which we apply our techniques to an astrophysics application. The results of these experiments demonstrate that conservative scheduling can produce execution times that are both significantly faster and less variable than other techniques. 1

The computational grid is becoming the platform of choice for large-scale distributed data-intensive applications. Accurately predicting the transfer times of remote data les, a fundamental component of such applications, is critical to achieving application performance. In this paper, we introduce a performance prediction method, ARM (Adaptive Regression Modeling), to determine data transfer times for network-bound distributed dataintensive applications. We demonstrate the eectiveness of the ARM method on two distributed data applications, SARA (Synthetic Aperture Radar Atlas) and SRB (Storage Resource Broker) , and discuss how it can be used for application scheduling. Our experiments demonstrate that applying the ARM method to these applications predicted data transfer times in wide-area multi-user grid environments with accuracy of 88% or better. 1 Introduction Ensembles of distributed computational, storage, and other resources, also known as computational grids [12, 14], are...

Prediction is a critical component in the achievement of application execution performance. The development of adequate and accurate prediction models is especially difficult in local-area clustered environments where resources are distributed and performance varies due to the presence of other users in the system. This paper discusses the use of stochastic values to parameterize cluster application performance models. Stochastic values represent a range of likely behavior and can be used effectively as model parameters. We describe two representations for stochastic model parameters and demonstrate their effectiveness in predicting the behavior of several applications under different workloads on a contended network of workstations.

Current distributed parallel platforms can provide the resources required to execute a scientific application efficiently. However, when these platforms are shared by multiple users, performance prediction becomes increasingly difficult due to the dynamic behavior of the system. This paper addresses the use of stochastic values, represented by intervals, to parameterize performance models. We describe a method for using upper and lower bound information to parameterize application prediction models in order to make better predictions about the application's behavior in a contentious environment. We demonstrate this technique for a set of 3 applications under different workloads on a production network of workstations. 1 Motivation and Outline In order to achieve performance in multi-user distributed environments, it is critical to provide performance models which accurately represent the execution behavior of programs in contended systems. In [15], we explored the accuracy of struct...

This paper address the use of practical histogram stochastic values to parameterize

One of the compelling reasons for developing the Information Power Grid (IPG) is to provide a platform for more rapid development and execution of simulations and other resource-intensive applications. However, the IPG will ultimately not be successful unless users and application developers can achieve execution performance for their codes. In this paper, we describe a performance-efficient approach to scheduling applications in dynamic multiple-user distributed environments such as the IPG. This approach provides the basis for application scheduling agents called AppLeS. We describe the AppLeS methodology and discuss the lessons learned from the development of AppLeS for a variety of distributed applications. In addition, we describe an AppLeS-in-progress currently being developed for NASA's INS2D code, a distributed "parameter sweep" application. 1 Introduction The NASA Information Power Grid (IPG) project provides an important opportunity to improve the efficiency and effectiven...

In the ever increasing world of distributed systems, different middleware implementations can be compared qualitatively or quantitatively. Existing evaluation techniques are often not satisfying. In this contribution we apply Design of Experiments (DoE) to evaluate and model the performance of JavaSpaces, a Tuple Spaces implementation for distributing tasks through a virtual space. DoE is a statistical technique for identifying relevant inputfactors from a (large) set of inputfactors. The setup of the experiments is determined using Experimental Design (ED) and a selection of the relevant inputfactors is based on Analysis of Variance (ANOVA). Then Regression Analysis (RA) is used to obtain a multivariate representation. Some extra experiments are performed to validate this approach.

Abstract: The Historical Trace Manager is a task duration predictor module embedded in the agent of a Problem Solving Environment relying on the client-agent-server. The HTM is introduced in [CJ02a] and [CJ04]. In this paper, we explain some improvements built into the HTM and NetSolve, the Problem Soving Environment we use for our tests, in order to synchronize the HTM to the reality. We also introduce two new scheduling heuritics relying on the HTM information: Advanced HMCT and Minimum Length. We study the scheduling of several scenarios, including the simultaneous submissions of DAGS and independent tasks, on a real heterogeneous platform. The excellent behavior of the HTM validates its estimations of the duration of each task concurrently running in the system. It can consequently predict the contention tasks may have on each other if scheduled and executed concurrently on the same computing resource. Heuristics performances show the relevancy of the HTM information through the experiments: their ability of behaving with a constant quality between two executions of the same experiment as well as the quality of their respective scheduling choices to optimize several criteria at the same time. We also show that heuristics which rely on minimizing the contention give generally the best results regardless the criterion.