Joachim Lepping

Development of scheduling strategies with Genetic Fuzzy systems

This paper presents a methodology for automatically generating online scheduling strategies for a complex objective defined by a machine provider. To this end, we assume independent parallel jobs and multiple identical machines. The scheduling algorithm is based on a rule system. This rule system classifies all possible scheduling states and assigns a corresponding scheduling strategy. Each state is described by several parameters. The rule system is established in two different ways. In the first approach, an iterative method is applied, that assigns a standard scheduling strategy to all situation classes. Here, the situation classes are fixed and cannot be modified. Afterwards, for each situation class, the best strategy is extracted individually. In the second approach, a Symbiotic Evolution varies the parameter of Gaussian membership functions to establish the different situation classes and also assigns the appropriate scheduling strategies. Finally, both rule systems will be compared by using real workload traces and different possible complex objective functions.

Prospects of collaboration between compute providers by means of job interchange

This paper empirically explores the advantages of the collaboration between different parallel compute sites in a decentralized grid scenario. To this end, we assume independent users that submit their jobs to their local site installation. The sites are allowed to decline the local execution of jobs by offering them to a central job pool. In our analysis we evaluate the performance of three job sharing algorithms that are based on the commonly used algorithms First-Come-First-Serve, EASY Backfilling, and List-Scheduling. The simulation results are obtained using real workload traces and compared to single site results. We show that simple job pooling is beneficial for all sites even if the local scheduling systems remain unchanged. Further, we show that it is possible to achieve shorter response times for jobs compared to the best single-site scheduling results.

On advantages of scheduling using genetic fuzzy systems

In this paper, we present a methodology for automatically generating online scheduling strategies for a complex scheduling objective with the help of real life workload data. The scheduling problem includes independent parallel jobs and multiple identical machines. The objective is defined by the machine provider and considers different priorities of user groups. In order to allow a wide range of objective functions, we use a rule based scheduling strategy. There, a rule system classifies all possible scheduling states and assigns an appropriate scheduling strategy based on the actual state. The rule bases are developed with the help of a Genetic Fuzzy System that uses workload data obtained from real system installations. We evaluate our new scheduling strategies again on real workload data in comparison to a probability based scheduling strategy and the EASY standard scheduling algorithm. To this end, we select an exemplary objective function that prioritizes some user groups over others.

Robust Load Delegation in Service Grid Environments

In this paper, we address the problem of finding well-performing workload exchange policies for decentralized Computational Grids using an Evolutionary Fuzzy System. To this end, we establish a noninvasive collaboration model on the Grid layer which requires minimal information about the participating High Performance and High Throughput Computing (HPC/HTC) centers and which leaves the local resource managers completely untouched. In this environment of fully autonomous sites, independent users are assumed to submit their jobs to the Grid middleware layer of their local site, which in turn decides on the delegation and execution either on the local system or on remote sites in a situation-dependent, adaptive way. We find for different scenarios that the exchange policies show good performance characteristics not only with respect to traditional metrics such as average weighted response time and utilization, but also in terms of robustness and stability in changing environments.

Decentralized Grid Scheduling with Evolutionary Fuzzy Systems

In this paper, we address the problem of finding workload exchange policies for decentralized Computational Grids using an Evolutionary Fuzzy System. To this end, we establish a non-invasive collaboration model on the Grid layer which requires minimal information about the participating High Performance and High Throughput Computing (HPC/HTC) centers and which leaves the local resource managers completely untouched. In this environment of fully autonomous sites, independent users are assumed to submit their jobs to the Grid middleware layer of their local site, which in turn decides on the delegation and execution either on the local system or on remote sites in a situation-dependent, adaptive way. We find for different scenarios that the exchange policies show good performance characteristics not only with respect to traditional metrics such as average weighted response time and utilization, but also in terms of robustness and stability in changing environments.

Designing multi-objective variation operators using a predator-prey approach

In this paper, we propose a new conceptual method for the design, investigation, and evaluation of multi-objective variation operators for evolutionary multi-objective algorithms. To this end, we apply a modified predator-prey model that allows an independent analysis of different operators. Using this model problem specific operators can be combined to more complex operators. Additionally, we review the simplex recombination, a new rotation-independent recombination scheme, and examine its impact concerning our design method. We show exemplarily as a first attempt the advantageous combination of several standard variation operators that lead to better results for selected test functions.

Exploring the behavior of building blocks for multi-objective variation operator design using predator-prey dynamics

In this paper, we utilize a predator-prey model in order to identify characteristics of single-objective variation operators in the multi-objective problem domain. In detail, we analyze exemplarily Gaussian mutation and simplex recombination to find explanations for the observed behaviorswithin this model. Then, both operators are combinedto a new complex one for the multi-objective case in order to aggregate the identified properties. Finally, we show that (a) characteristic properties can still be observed in the combination and (b) the collaboration of those operators is beneficial for solving an exemplary multi-objective problem regarding convergence and diversity.

Discovering performance bounds for grid scheduling by using evolutionary multiobjective optimization

In this paper, we introduce a methodology for the approximation of optimal solutions for a resource allocation problem in the domain of Grid scheduling on High Performance Computing systems. In detail, we review a real-world scenario with decentralized, equitable, and autonomously acting suppliers of compute power who wish to collaborate in the provision of their resources. We exemplarily apply NSGA-II in order to explore the bounds of maximum achievable benefit. To this end, appropriate encoding schemes and variation operators are developed while the performance is evaluated. The simulations are based upon recordings from real-world Massively Parallel Processing systems that span a period of eleven months and comprise approximately 100,000 jobs. By means of the obtained Pareto front we are able to identify bounds for the maximum benefit of Grid computing in a popular scenario. For the first time, this enables Grid scheduling researchers to rank their developed real-world strategies.

Towards A Standards-Based Grid Scheduling Architecture

The definition of a generic Grid scheduling architecture is t he concern of both the Open Grid Forum’s Grid Scheduling Architecture Research Group and a CoreGRID research group of the same name. Such an architecture should provide a blueprint for Grid system and middleware designers and assist them in linking their scheduling requirements to diverse existing solutions and standards. Based on work executed within OGF related to scheduling use cases and requirements, which tackles the problem from a more theoretical point of view, we approach in this paper the problem practically by evaluating the teikoku scheduling framework in the light of standards-compliance. The results of this evaluation and the existing Grid Scheduling Architecture proposal are set into context, existing ga ps are described and potential solutions to bridge them are int roduced. In doing so, we concentrate on the interoperability of schedulers and the necessity of a Scheduling Description Language to achieve it.

Genetic Fuzzy Systems applied to Online Job Scheduling

This paper presents a comparison of three different design concepts for genetic fuzzy systems. We apply a symbiotic evolution that uses the Michigan approach and two approaches that are based on the Pittsburgh approach: a complete optimization of the problem and a cooperative coevolutionary algorithm. The three different genetic fuzzy systems are applied to a real-world online problem, the generation of scheduling strategies for massively parallel processing systems. The genetic fuzzy systems must classify different scheduling states and decide about a corresponding scheduling strategy within each scheduling state. The main challenge arise in the delayed reward given by a critic. Therefore, it is impossible to directly evaluate the assignment of scheduling strategies to scheduling states. In our paper, the three design concepts are evaluated with real workload traces considering result quality, computational effort, convergence behavior, and robustness.