Artificial life techniques for load balancing in computational grids

It is our great pleasure to welcome you to BADS 2011, the third Workshop on Bio-Inspired and Self-* Algorithms for Distributed Systems, co-located with ICAC 2011, the 8th IEEE/ACM International Conferenceon Autonomic Computing, Karlsruhe, Germany, June 14-18, 2011. BADS 2011 aims to provide a forum for exploring bio-inspired and self-* algorithms to build autonomic distributed systems; i.e., self-managing distributed systems that can organize, configure, optimize, scale, protect and/or heal themselves with minimal human intervention. Computing inspired by biology has a long history since 1950s. Its common objectives are to formulate theoretical models that are faithfully designed after biological principles, phenomena and processes and to exploit the models for developing computational tools (e.g., algorithms) and systems. Well-known bio-inspired algorithms include artificial neural networks, evolutionary algorithms, swarm intelligence, artificial immune systems, reaction-diffusion algorithms, epidemic algorithms, cellular automata algorithms, gene expression/regulatory algorithms and Lindenmayer system. They have been successfully used as alternative and/or superior solutions to the problems that traditional algorithms cannot solve satisfactorily. In recognition of their achievements and potential, bio-inspired algorithms were named one of Scientific American magazine's 10 "World Changing Ideas 2010." The BADS workshop series focuses on distributed computing and examines the opportunities, current state-of-the-art, challenges and future outlook to create autonomic distributed systems with bio-inspired algorithms. We cover a broad range of aspects in distributed computing, including communication protocols, resource clustering, allocation and management, topology management, parallelism and concurrency, overlay construction and maintenance, decentralized information search and delivery, load distribution, synchronization, service selection, composition and deployment, labor division and task allocation, anomaly and misbehavior detection, malware propagation and detection. Following the success of its former editions in 2009 and 2010, BADS 2011 continues to focus on bio-inspired, autonomic distributed systems. In addition, it expands its themes to cover a general class of algorithms that exhibit self-* properties (e.g., self-configuring, self-organizing, self-managing, self-healing, self-scaling and selfadaptive properties) in the context of distributed computing. BADS 2011 has seven research papers selected through a rigorous review process. Each paper was reviewed and judged by at least three PC members on merits including correctness, originality, technical strength, quality of presentation and relevance to the workshop themes. We would like to sincerely thank the Program Committee members who have greatly contributed to the success of BADS 2011. BADS 2011 is organized in two sessions and one keynote presentation. The first session includes papers that present bio-inspired approaches for distributed computing. The second session focuses on solving optimization problems with distributed bio-inspired and self-* algorithms. The keynote presentation, entitled "Methods for Self-organizing Distributed Software", is given by Elisabetta Di Nitto, and focuses on the applications of self-organization in distributed systems. In nature, self-organization occurs such that a global behavior emerges from simple and local decisions made autonomously by each element of the system. This keynote overviews the benefits and challenges to apply the notion of self-organization for the development of distributed systems. It also presents several example applications such as energy efficiency and configuration optimization in cloud computing. The first session, "Bio-inspired Approaches for Distributed Computing", comprises four papers. The first paper, "Memetic Algorithm for Web Service Selection" by Ludwig, studies the service selection process, whose aim is to locate services of interest in service-oriented environments. The author proposes two approaches based on a genetic algorithm and a memetic algorithm to match consumers with services based on Quality of Service attributes. Both approaches are compared with an optimal assignment algorithm, called the Munkres algorithm, as well as with random approaches. The next paper, "OZMOS: Bio-Inspired Load Balancing in a Chord-based P2P Grid" by Brocco, proposes a load balancing mechanism, called Ozmos, that follows the principle of osmosis to relocate tasks between nodes in a P2P-based Grid. Bio-inspired agents are deployed to share information about the status of the Grid as well as to reschedule tasks between nodes. The efficacy of the proposed algorithm is illustrated in achieving system-wide load balance in Grids of different scales, with both homogenous and heterogeneous resources. The third paper, "Model-Driven Performance Engineering for Wireless Sensor Networks with Feature Modeling and Event Calculus" by Boonma and Suzuki, proposes an application development framework for wireless sensor networks. The proposed framework allows developers to graphically configure their applications and generate application code. With event calculus, the framework also estimates each application's performance such as per-node power consumption and network lifetime without running it on simulators and real networks. The proposed framework currently targets a shortest-path routing; however, it is a promising next step to estimate the performance of bio-inspired routing protocols such as pheromone-based chemotaxis routing. The last paper of the first session, "Description and Composition of Bio-Inspired Design Patterns: the Gradient Case" by Fernandez-Marquez, Arcos, Di Marzo Serugendo, Viroli and Montagna, investigates a group of bioinspired mechanisms in terms of design patterns. The authors show the relations between the different patterns and identify the boundaries of each mechanism. The mechanisms are organized into different levels. In basic level, Spreading, Aggregation and Evaporation are studied. Gradient is investigated in mid-level, and top-level mechanisms are Chemotaxis, Morphogenesis, and Quorum sensing. The second session, "Solving Optimization Problems with Distributed Evolutionary Algorithms", opens with a paper entitled "Protein Structure Prediction Using Particle Swarm Optimization and a Distributed Parallel Approach" by Kondov and Berlich. It demonstrates the efficiency of the standard Particle Swarm Optimization (PSO) algorithm to find the folded state of two proteins of different sizes starting from completely extended conformations. They illustrate that the predicted structure of the larger protein is in good agreement with the structure from the Protein Data Bank within the experimental resolution. The parallelization of PSO is shown to speed up the simulation linearly with the number of workers and it reduces the prediction time dramatically without loss of accuracy. The next paper, "Discrete Optimization Problem Solving with three Variants of Hybrid Binary Particle Swarm Optimization" by Singh V., Singh D., Tiwari and Shukla, presents three variants of the binary PSO algorithm in order to solve discrete optimization problems. This is achieved by introducing an additional step of crossover that varies in all the three variants. These hybrid algorithms show competitive results compared to other state-of-theart metaheuristics. The third paper, "Self-organized Invasive Parallel Optimization" by Mostaghim, Pfeiffer and Schmeck, proposes a new parallelization framework for optimization algorithms. In this approach, the self-organized resources are represented as a unified resource to the user who only specifies the optimization problem and his/her preferences to the system. The invasive approach starts with one resource and automatically divides the optimization task stepwise into smaller tasks, which are assigned to more resources. The job assignment pattern is decided on demand, i.e., the number of required resources is estimated during the optimization process. The authors examine their generic framework on multi-objective problems.

BackgroundProtein structure prediction (PSP) has important applications in different fields, such as drug design, disease prediction, and so on. In protein structure prediction, there are two important issues. The first one is the design of the structure model and the second one is the design of the optimization technology. Because of the complexity of the realistic protein structure, the structure model adopted in this paper is a simplified model, which is called off-lattice AB model. After the structure model is assumed, optimization technology is needed for searching the best conformation of a protein sequence based on the assumed structure model. However, PSP is an NP-hard problem even if the simplest model is assumed. Thus, many algorithms have been developed to solve the global optimization problem. In this paper, a hybrid algorithm, which combines genetic algorithm (GA) and tabu search (TS) algorithm, is developed to complete this task.ResultsIn order to develop an efficient optimization algorithm, several improved strategies are developed for the proposed genetic tabu search algorithm. The combined use of these strategies can improve the efficiency of the algorithm. In these strategies, tabu search introduced into the crossover and mutation operators can improve the local search capability, the adoption of variable population size strategy can maintain the diversity of the population, and the ranking selection strategy can improve the possibility of an individual with low energy value entering into next generation. Experiments are performed with Fibonacci sequences and real protein sequences. Experimental results show that the lowest energy obtained by the proposed GATS algorithm is lower than that obtained by previous methods.ConclusionsThe hybrid algorithm has the advantages from both genetic algorithm and tabu search algorithm. It makes use of the advantage of multiple search points in genetic algorithm, and can overcome poor hill-climbing capability in the conventional genetic algorithm by using the flexible memory functions of TS. Compared with some previous algorithms, GATS algorithm has better performance in global optimization and can predict 3D protein structure more effectively.

To improve the global throughput of computational grid, effective and efficient load balancing algorithms are fundamentally important. A computational grid differs from traditional high-performance computing system in the heterogeneity of computing nodes, as well as the communication links that connect the different nodes together. In this paper, we propose a resource queue length based solution to the grid load balancing problem. The proposed algorithm will balance the load in the grid based on the queue length of each resource and transfer the job to the resource having minimum queue length. The simulation results show that our algorithm can achieve a better load balancing performance than its counterpart provided in simulation.

In object-based image analysis of high-resolution images, the number of features can reach hundreds, so it is necessary to perform feature reduction prior to classification. In this paper, a feature selection method based on the combination of a genetic algorithm (GA) and tabu search (TS) is presented. The proposed GATS method aims to reduce the premature convergence of the GA by the use of TS. A prematurity index is first defined to judge the convergence situation during the search. When premature convergence does take place, an improved mutation operator is executed, in which TS is performed on individuals with higher fitness values. As for the other individuals with lower fitness values, mutation with a higher probability is carried out. Experiments using the proposed GATS feature selection method and three other methods, a standard GA, the multistart TS method, and ReliefF, were conducted on WorldView-2 and QuickBird images. The experimental results showed that the proposed method outperforms the other methods in terms of the final classification accuracy.

The combination of principal component analysis, genetic algorithm and tabu search in 3d molecular similarity

Hybrid genetic algorithm and tabu search for finite capacity material requirement planning system in flexible flow shop with assembly operations

Two heuristic techniques, the genetic algorithm (GA) and Tabu search (TS), both with an embedded linear programming routine for earthwork allocation, were compared to a manually designed forest road profile. The manually designed road length was 345.7m and its average gradient was 14.1%. The best costs of the profiles designed by GA and TS, without changing the placement of control points, were less than that designed manually. The best cost found by GA was almost the same as the global optimum solution. While TS could not find a better solution than GA, it usually found a good solution in less time. It was not possible to search all alternatives by changing the placement of control points and find the global optimum solution within a reasonable time. However, it can be concluded from the results that both GA and TS found good solutions within a reasonable time. Since it is not possible to manually evaluate many alternatives, road designers should find heuristic techniques helpful for design of the road profile. Moreover, the effect of the number of control points on construction costs was examined. The results indicated that increasing the number of control points reduces the construction costs. However, driving safety and comfort might be decreased.

The traditional single intelligent algorithm has slow convergence speed in the frequency assignment, and the effect cannot meet the increasing frequency equipment in the naval battlefield. Based on the single intelligent frequency assignment algorithm and the frequency conflict analysis model between systems, this paper proposes three heuristic frequency assignment algorithms based on hybrid genetic algorithms, namely, Genetic Algorithm and Tabu Search (GATS), Hybrid Genetic Simulated Annealing Algorithm (HGSAA) and Genetic Algorithm-ant Colony Algorithm (GAA). The simulation results show that the hybrid algorithm can quickly converge to a better allocation result than the single intelligent algorithm in the frequency assignment problem, and the optimal value is better than the single intelligent algorithm. Among them, the convergence speed and optimal value of the GAA are the best of the three algorithms. Therefore, this algorithm can be applied to frequency assignment with more frequency-using equipment within a fixed range.

Load balancing is a key concern when developing parallel and distributed computing applications. The emergence of computational grids extends this problem, where issues of cross-domain and large-scale scheduling must also be considered. In this paper an agent-based grid management infrastructure is coupled with a performance-driven task scheduler that has been developed for local grid load balancing. Each grid scheduler utilises predictive application performance data and an iterative heuristic algorithm to engineer local load balancing across multiple processing nodes. At a higher level, a hierarchy of homogeneous agents are used to represent multiple grid resources. Agents cooperate with each other to balance workload in the global grid environment using service advertisement and discovery mechanisms. A case study is included with corresponding experimental results to demonstrate that both local schedulers and agents contribute to overall grid load balancing, which significantly improves grid application execution performance and resource utilisation.

The performance of SPMD programs is strongly affected by dynamic load imbalancing factors. The use of a suitable load balancing algorithm is essential for overcoming the effects of these imbalancing factors. In this work, we evaluate the performance of a scientific SPMD parallel application when executed on a computational grid, with different kinds of load balancing strategies. The developed SPMD application computes the macroscopic thermal dispersion in porous media. A set of experiments was conducted on a computational grid composed by two geographically separated clusters. The main contribution of this work is the performance evaluation and comparison of a large variety of load balancing techniques under dynamic environment conditions. The experimental results showed the importance of choosing appropriate load balancing strategies when developing SPMD applications on a grid environment.

Aiming at the disadvantage of premature convergence of basic genetic algorithm, an adaptive simulated annealing genetic tabu search algorithm is proposed. This algorithm fully combines the global convergence and adaptability of simulated annealing algorithm and the strong climbing ability and high efficiency of tabu search strategy. It has strong convergence and adaptability. The simulation results of the adaptive simulated annealing genetic tabu search algorithm are given and compared with the basic genetic algorithm and simulated annealing algorithm. The simulation results show that the algorithm has better convergence and optimization performance, and can better solve the combinatorial optimization problem.

Objectives: In this paper various task scheduling algorithm along with load distribution techniques investigated to ensure efficient mapping of tasks to resources and for coherent resource utilization in heterogeneous environment. Statistical Analysis: A detailed comparative approach is conferred for different job scheduling and load balancing algorithm available in the literature. Considered parameters for comparisons are scheduling approaches, techniques, findings, benefits, pros and cons. This study facilitates to identify the efficient scheduling and load balancing approach which optimizes resource use, minimizes completion time as well as task waiting time and balanced load among resources. Findings: Make span time and flow time are two important parameters to compute and minimize job execution time. This paper explores an intense survey of minimization of task execution time with minimum make span time and flow time. Apart from this a joint review on job scheduling, load balancing and minimizing job completion time using Expected Time to Compute (ETC) matrix is presented. At the end of this paper a research model is proposed which is a novel and hybrid approach with a combination of the efficiency of scheduling and load balancing. Application: The Load Balancing (LB) is an important feature of the grid environment. The efficient LB algorithm assures cost effective resource usage by the provisioning of resources to users on demand schedule in pay-as-you-say-manner.

A lot of scheduling problems of flexible manufacturing processes can be described as combinatorial optimization problems. The sequence optimization problem, which asks for an optimal sequence of products, is a well known example. Although many different scheduling techniques were developed in the past, the application of them is very difficult. In our opinion the optimization of scheduling only can be solved satisfactorily in combination with methods of simulation which guarantee the required high flexibility of the optimization system. We describe in this paper an approach for an open optimization system, where the optimization cycle consists of two independent parts - the simulation and the search algorithm. The easy replacement of these parts is the advantage of the open architecture. So it is possible to combine several simulation systems and optimization algorithms in the same system and to use it as a experimental environment for examining new optimization strategies. Some optimization algorithms were already examined e.g. Genetic Search Algorithms and Tabu Search. The practical test will be carried out in the field of electronic production.

One of the challenging issues in computational grid is load balancing. In many approaches load balancing is done only at the local scheduler level, which is applicable to small application and leads to more communication overhead between the resources. For the large scale application load balancing at the local scheduler level will not provide the feasible solution. So the novel Load Balancer algorithm is proposed, which provides the load balancing at the meta-scheduler level. To initiate the load balancing triggering policy is used, which determines the appropriate time period to start the load balancing operation by using Boundary value approach. This approach increases the performance by reducing the waiting time of jobs and by maximizing the utilization resource which is least loaded.KeywordsLoad BalancingLoad informationTriggering policyBoundary value approach

Computational grids provide a massive source of processing power, providing the means to support processor intensive applications. The strong burstiness and unpredictability of the available resources raise the need to make applications robust against the dynamics of grid environment. The two main techniques that are most suitable to cope with the dynamic nature of the grid are load balancing and job replication. In this work, we develop a load-balancing algorithm by juxtaposes the strong points of neighbor-based and cluster-based load-balancing methods. We then integrate the proposed load-balancing approach with fault-tolerant scheduling namely MinRC and develop a performance-driven fault-tolerant load-balancing algorithm or PD_MinRC for independent jobs. In order to improve system flexibility, reliability, and save system resource, PD_MinRC employs passive replication scheme. Our main objective is to arrive at job assignments that could achieve minimum response time, maximum resource utilization, and a well-balanced load across all the resources involved in a grid. Experiments were conducted to show the applicability of PD_MinRC. One advantage of our approach is the relatively low overhead and robust performance against resource failures and inaccuracies in performance prediction information.