Daniel Millot

An adaptive scheduling method for grid computing

This paper presents an adaptive scheduling method, which can be used for parallel applications whose total workload is unknown a priori. This method can deal with the unpredictable execution conditions commonly encountered on grids. To address this scheduling problem, parameters which quantify the dynamic nature of the execution conditions had to be defined. The proposed scheduling method is based on an on-line algorithm so as to be adaptable to the varying execution conditions, but avoids the idle periods inherent to this on-line algorithm.

Scheduling on Unspecified Heterogeneous Distributed Resources

In this paper, we present an adaptive method for scheduling parallel applications on unspecified distributed memory platforms. The presented method can be used to schedule parallel applications when the total workload and the execution parameters (communication speed, available computing power...) are unspecified. When used to schedule divisible load applications according to a master worker model, this method delivers the workload through multiple rounds. In order to maximize the throughput of each worker, it can prevent both idleness in the use of workers and contentions in the use of the links between master and workers. Before focusing on the proposed scheduling method, the paper recalls the underlying methods on which its development relies. The paper then gives a theoretical analysis of the method before presenting results of simulations obtained with the Sim Grid framework on a limited distributed memory platform.

Dynamically scheduling divisible load for grid computing

This paper presents an improved framework for an existing adaptive method for scheduling divisible load. The proposed framework delimits in a new way the application field of the method. This method can be used to schedule parallel applications whose total workload is unknown a priori, and can deal as well with the unpredictable execution conditions commonly encountered on grids. Some parameters which quantify the dynamic nature of these varying execution conditions are discussed in the paper.

A runtime scheduling method for dynamic and heterogeneous platforms

In this paper, we present a runtime method for scheduling parallel applications on dynamic and heterogeneous platforms. It can be used to schedule parallel applications whose total workload is unknown a priori. It can also handle the heterogeneous and dynamic conditions of execution that are typical of grids. The method delivers the workload through multiple rounds in order to improve communication/computation overlap on an existing on-line algorithm

Some Tests of Adaptivity for the AS4DR Scheduler

This paper presents some tests of adaptivity for the AS4DR (Adaptive Scheduling for Distributed Resources) scheduler. The objective of AS4DR is to maximize the CPU use efficiency when executing divisible load applications on heterogeneous distributed memory platforms. Furthermore, this scheduler can operate when the total workload is unknown and when the execution parameters (available communication speed, available computing speed.) are unspecified or may vary through time. The paper analyzes results obtained when simulating AS4DR scheduling on platforms characterized by such type of execution parameters.

Setting up Clusters of Computing Units to Process Several Data Streams Efficiently

Let us consider an upper bounded number of data streams to be processed by a Divisible Load application. The total workload is unknown and the available speeds for communicating and computing may be poorly a priori estimated. This paper presents a resource selection method that aims at maximizing the throughput of this processing. From a set of processing units linked by a network, this method consists in forming an optimal set of master-workers clusters. Results of simulations are presented to assess the efficiency of this method experimentally. Before focusing on the proposed resource selection method, the paper comes back on the adaptive scheduling method on which it relies.

First Experimental Assessments of the Adaptivity of the Scheduling with AS4DR

The AS4DR (Adaptive Scheduling for Distributed Resources) scheduling method experimented in this paper aims at maximizing the CPU use efficiency when executing divisible load applications on heterogeneous distributed memory platforms. AS4DR adapts the scheduling to: the unawareness of the total workload, both the unspecification and the variation over time of the execution parameters (available communication speed, available computing speed, etc.). This paper presents the first experimental assessments of the adaptivity of the scheduling with this method.

Contention-Free Scheduling in a Dynamic Context

In this paper, we present an adaptive method for scheduling parallel applications on heterogeneous multi-processor platforms, in a dynamic context. When executing divisible load applications according to a master-worker model, this method delivers the workload through multiple rounds and can avoid contentions in the use of the network. Such contentions avoidance is a means to avoid idle time, thus to maximize the throughput. Before presenting the proposed scheduling method, the paper revisits a runtime method which motivated its development. The method presented in this paper can be used to schedule parallel applications whose total workload is large but unknown a priori.

Progressive Introduction of Security in Remote-Write Communications with no Performance Sacrifice

In a framework where both security and performance are crucial, cluster users must be able to get both at the desired level. Remote-write communications bring high performance but expose physical addresses. In this paper, we present an approach which allows the user to secure remote-write while deciding the cost of that securization.

Models for dynamically placed concurrent processes

We define a formal framework to model executions and dynamic placement of programs on networks of processors. It provides an equational characterization of programs with respect to their expansion.