A linear programming-driven genetic algorithm for meta-scheduling on utility grids

Abstract

In Grids, single user-based brokers focus on meeting individual user job's quality of service requirements such as minimising the cost and time without considering demands from other users. This results in contention for resources and suboptimal schedules. Meta-scheduling in Grids aims to address this scheduling problem, which is NP-hard due to its combinatorial nature. Thus, many heuristic-based solutions using genetic algorithm (GA) have been proposed, apart from traditional algorithms such as greedy and first come first serve. In this paper, we propose the need for a 'metabrokering system' and present a Meta-Broker which schedules multiple jobs on utility Grids. First, we present the architecture of our Meta-Broker and discuss the requirements and functionalities of the Meta-Broker. We, then, propose a linear programming (LP)/integer programming model for scheduling user jobs to multiple resources. We also propose a novel algorithm LP-driven GA which combines the capabilities of LP and GA. The aim of this algorithm is to obtain the best meta-schedule that minimises the combined cost of all users in a coordinated manner. Simulation results show that in comparison to single user-based brokers such as Gridbus and GRUBER, our proposed integrated algorithm offers the best schedule having the minimum processing cost with negligible time overhead.

Min-heap-based scheduling algorithm: an approximation algorithm for homogeneous and heterogeneous distributed systems

“…In general, those algorithms mimic natural mechanisms such as evolution, colony organisation, etc. Genetic Algorithms [6,14,21,41,43] and Particle Swarm Optimisation techniques [24,27,33] are widespread approaches used to address the scheduling problem.…”

A Hybrid Optimization Approach for Autonomy Enhancement of Nearly-Zero-Energy Buildings Based on Battery Performance and Artificial Neural Networks

“…Imitating biological evolution, GA provides a mechanism capable of solving NP-hard optimization problems [34]. It relies on natural selection processes that enable the production of a population of points, promotes the best solution to the next iteration, and progressively recommends the optimal configuration [35]. In addition, GA utilizes random number generators (rather than deterministic computation), to strengthen the exploration space and repeatedly modify a population of "child" solutions to conclude to an optimal "parent" solution.…”

“…The most important steps in the GA process refer to the population production, successive evaluation, and the best candidate recommendation, crossover, and mutation. This procedure is repeatedly performed until convergence, the criterion of which is commonly reflected as "no change in the solution for n generations" [35]. This way, the exploitation of the best solutions via the exploration of new regions guarantees a large search space, which heuristically provides a high-quality solution.…”

Development of information technology of tasks distribution for grid-systems using the grass simulation environment

“…In this case, the task suppliers can set preferred time frames of use of computing resources. In [12] it is proposed to use a combination of integer linear programming model with a genetic algorithm, which enables composing the distribution plan and taking into account the cost of computing resources for specific groups of suppliers. In [13] a model of distribution of tasks is proposed, which enables re-planning "in the air" by means of information about the actual dynamics of the system load.…”

“…Planning methods described in [4,[9][10][11][12][13] use a number of criteria (the value of the use of resources, the level of load of computing resources, the use of nearby resources for related objectives in the task). In the studies [14][15][16][17] the methods of planning, which focus on the preferences of task suppliers, administrators of virtual organizations or suppliers of computing resources are suggested to be used.…”