Klaus H. Ecker

Scheduling Computer and Manufacturing Processes

No abstract

Handbook on Parallel and Distributed Processing

Parallel and Distributed Computing: State-of-the-Art and Emerging Trends.- The Design of Efficient Parallel Algorithms.- Languages for Parallel Processing.- Architecture of Parallel and Distributed Systems.- Parallel Operating Systems.- Management of Resources in Parallel Systems.- Tools for Parallel Computing: A Performance Evaluation Perspective.- Parallel Database Systems and Multimedia Object Servers.- Networking Aspects of Distributed and Parallel Computing.- Parallel and Distributed Scientific Computing.- High-performance Computing in Molecular Sciences.- Multimedia Applications for Parallel and Distributed Systems.

A framework for decision support systems for scheduling problems

Considering a decision support system as a tool where executives judgment can be included along with the mathematical tool kit of the management scientist, this paper shows the need to include problem management as an integral component of the decision support system for scheduling problems. A methodology based on the resolution of conflicts among various constraints in scheduling problems is proposed to implement the problem management system in a decision support system for these problems. The paper concludes with some guidelines to create a workable framework for providing effective decision support to solve scheduling problems and the identification of some fruitful directions for future research.

Scheduling tasks on a flexible manufacturing machine to minimize tool change delays

This paper considers the problem of scheduling a given set of precedence constraint tasks on a flexible machine equipped with a tool magazine where each task requires exactly one of the tools during its execution. Changing from one tool to another requires a certain amount of time that depends on the pair of tools being exchanged. We present a new algorithmic approach for general task precedence relations when it is desired to sequence the tasks in such a way that the total time required for tool changes is minimized. The proposed algorithm is of polynomial time complexity in case of task precedences of limited width w, i.e. for precedence relations where each subset of independent tasks has not more than w elements. Since the task precedences width w could be arbitrary, we describe two heuristic algorithms and empirically evaluate their effectiveness in finding schedules with minimum total time required for tool changes.

An optimization framework for dynamic, distributed real-time systems

The paper presents a model that is useful for developing resource allocation algorithms for distributed real-time systems that operate in dynamic environments. Interesting aspects of the model include dynamic environments, utility and service levels, which provide a means for graceful degradation in resource-constrained situations and support optimization of the allocation of resources. The paper also provides an allocation algorithm that illustrates how to use the model for producing feasible, optimal resource allocations.

Algorithms for dynamic scheduling of unit execution time tasks

Abstract We analyze performance properties of list scheduling algorithms under various dynamic assumptions and different levels of knowledge available for scheduling, considering the case of unit execution time tasks. We focus on bounds for the ISF (immediate successors first) and MISF (maximum number of immediate successors first) scheduling strategies and show the difference from other bounds obtained for the same problem. Finally, we present case studies and experimental results to assess the average behavior.

Management of Resources in Parallel Systems

In this chapter we discuss the parallel processing context of scheduling in computer systems. Only deterministic model of scheduling will be examined. This does not mean that only static problems in which all parameters are known in advance and the solutions are in all ways fixed, are discussed. Also dynamic problems are possible in which some characteristics are not known in advance. In the following, basic classical scheduling approaches will be recalled and then a special attention will be paid to three new models of scheduling in parallel systems: scheduling multiprocessor tasks, scheduling with communication delays and scheduling divisible tasks.

Resource Constrained Scheduling

The scheduling model we consider now is more complicated than the previous ones, because any task, besides processors, may require for its processing some additional scarce resources. Resources, depending on their nature, may be classified into types and categories. The classification into types takes into account only the functions resources fulfill: resources of the same type are assumed to fulfill the same functions. The classification into categories will concern two points of view. First, we differentiate three categories of resources from the viewpoint of resource constraints. We will call a resource renewable, if only its total usage i.e. temporary availability at every moment is constrained (in other words this resource can be used once more when returned by a task currently using it). A resource is called nonrenewable, if only its total consumption, i.e. integral availability up to any given moment is constrained (in other words this resource once used by some task cannot be assigned to any other task). A resource is called doubly constrained, if both total usage and total consumption are constrained. Secondly, we distinguish two resource categories from the viewpoint of resource divisibility: discrete (i.e. discretely-divisible) and continuous (i.e. continuously-divisible) resources. In other words, by a discrete resource we will understand a resource which can be allocated to tasks in discrete amounts from a finite set of possible allocations, which in particular may consist of only one unit per task. Continuous resources, on the other hand, can be allocated in arbitrary a priori unknown amounts less than or equal to some given maximum value.

A note on the complexity of scheduling coupled tasks on a single processor

This paper considers a problem of coupled task scheduling on one processor, where all processing times are equal to 1, the gap has exact length h, precedence constraints are strict and the criterion is to minimise the schedule length. This problem is introduced e.g. in systems controlling radar operations. We show that the general problem is NP-hard.

Scheduling under Resource Constraints

The scheduling model we consider now is more complicated than the previous ones, because any task, besides processors, may require for its processing some additional scarce resources. Resources, depending on their nature, may be classified into types and categories. The classification into types takes into account only the functions resources fulfill: resources of the same type are assumed to fulfill the same functions. The classification into categories will concern two points of view. First, we differentiate three categories of resources from the viewpoint of resource constraints. We will call a resource renewable, if only its total usage, i.e. temporary availability at every moment, is constrained (in other words this resource can be used once more when returned by a task currently using it). A resource is called non-renewable, if only its total consumption, i.e. integral availability up to any given moment, is constrained (in other words this resource once used by some task cannot be assigned to any other task). A resource is called doubly constrained, if both total usage and total consumption are constrained. Secondly, we distinguish two resource categories from the viewpoint of resource divisibility: discrete (i.e. discretely-divisible) and continuous (i.e. continuously-divisible) resources. In other words, by a discrete resource we will understand a resource which can be allocated to tasks in discrete amounts from a given finite set of possible allocations, which in particular may consist of one element only. Continuous resources, on the other hand, can be allocated in arbitrary, a priori unknown, amounts from a given interval.