Vesna Sesum-Cavic

Applying Swarm Intelligence Algorithms for Dynamic Load Balancing to a Cloud Based Call Center

Load-Balancing is a significant problem in heterogeneous distributed systems. Nowadays we face an extreme growth of computer systems and their complexities requiring advanced intelligent solutions for load-balancing that lead to autonomic self-organizing infrastructures. There is still a need to prove that real use cases can benefit from self-* approaches. We developed a pattern, called SILBA, for such an infrastructure based on decentralized control, intelligent and exchangeable policies for load-balancing, and black-board based communication mechanisms. Different types of algorithms (both intelligent and unintelligent) were plugged into SILBA. In this paper, we present one particular use-case – a Call Center that is operated in a Cloud environment.

A space-based generic pattern for self-initiative load clustering agents

Load clustering is an important problem in distributed systems, which proper solution can lead to a significant performance improvement. It differs from load balancing as it considers a collection of loads, instead of normal data items, where a single load can be described as a task. Current approaches that treat load clustering mainly lack of provisioning a general framework and autonomy. They are neither agent-based nor configurable for many topologies. In this paper we propose a generic framework for self-initiative load clustering agents (SILCA) that is based on autonomous agents and decentralized control. SILCA is a generic architectural pattern for load clustering. The SILCA framework is the corresponding implementation and thus supports exchangeable policies and allows for the plugging of different algorithms for load clustering. It is problem independent, so the best algorithm or combination of algorithms can be found for each specific problem. The pattern has been implemented on two levels: In its basic version different algorithms can be plugged, and in the extended version different algorithms can be combined. The flexibility is proven by means of nine algorithms. Further contributions are the benchmarking of the algorithms, and the working out of their best combinations for different topologies.

Chapter 8 Self-Organized Load Balancing through Swarm Intelligence

The load balancing problem is ubiquitous in information technologies. New technologies develop rapidly and their complexity becomes a critical issue. One proven way to deal with increased complexity is to employ a self-organizing approach. There are many different approaches that treat the load balancing problem but most of them are problem specific oriented and it is therefore difficult to compare them. We constructed and implemented a generic architectural pattern, called SILBA, which stands for “self-initiative load balancing agents”. It allows for the exchanging of different algorithms (both intelligent and unintelligent ones) through plugging. In addition, different algorithms can be tested in combination at different levels. The goal is to ease the selection of the best algorithm(s) for a certain problem scenario. SILBA is problem and domain independent, and can be composed towards arbitrary network topologies. The underlying technologies encompass a black-board based communication mechanism, autonomous agents and decentralized control. In this chapter, we present the complete SILBA architecture by putting the accent on using SILBA at different levels, e.g., for load balancing between agents on one single node, on nodes in one subnet, and between different subnets. Different types of algorithms are employed at different levels. Although SILBA possesses self-organizing properties by itself, a significant contribution to self-organization is given by the application of swarm based algorithms, especially bee algorithms that are modified, adapted and applied for the first time in solving the load balancing problem. Benchmarks are carried out with different algorithms and in combination with different levels, and prove the feasibility of swarm intelligence approaches, especially of bee intelligence.

Instantiation of a Generic Model for Load Balancing with Intelligent Algorithms

In peer-to-peer networks, an important issue is the distribution of load having an impact on the overall performance of the system. The answer could be the application of an intelligent approach that leads to autonomic self-organizing infrastructures. In this position paper, we briefly introduce a framework model for load balancing that allows various load-balancing algorithms to be plugged-in, and that uses virtual shared-memory-based communication known to be advantageous for the communication of auto nomous agents in order to enable the collaboration of load-balancing agents. As the main contribution, we show how the biological concepts of bees can be mapped to the load-balancing problem, explain why we expect that bee intelligence can outperform other (un)intelligent approaches, and present an instantiation of the model with the bee intelligence algorithm. This load-balancing scheme focuses on two main policies: a transfer and a location policy for which we suggest some improvements.

Comparing Configurable Parameters of Swarm Intelligence Algorithms for Dynamic Load Balancing

A main challenge on today’s distributed systems is to cope with huge amounts of load. An important research issue is to distribute load across enterprise boundaries in highly heterogeneous environments. However, intelligent and adaptive load balancing problem is a complex problem and requires both intelligent algorithms and approaches. In this paper, we present the first findings of a novel approach towards load balancing, based on bee intelligence. The approach defines a generic architecture called SILBA (self initiative load balancing agents) which allows the exchange of different algorithms through simple plugging techniques. Six algorithms were developed, both unintelligent and intelligent ones, compared and performance benchmarks on both a cluster of virtual nodes and the Amazon EC2 cloud demonstrate promising benefits of the bee based algorithm.

A Swarm Intelligence Appliance to the Construction of an Intelligent Peer-to-Peer Overlay Network

Location and manipulation of complex data is a difficult, challenging task in nowadays extremely complex IT systems and on the Internet, overwhelmed with a huge amount of information that develops and grows rapidly. In this paper, we propose a self-organizing approach that combines purely decentralized unstructured P2P with space based computing in order to locate effectively and filter (retrieve) information from a network. The inspiration came from swarm intelligence that possesses distributive and autonomous properties and represents a self-organizing biological system. The goal is to achieve a good query capability, which we prove by means of benchmarks.

Bio-inspired search algorithms for unstructured P2P overlay networks

Abstract Efficient location and manipulation of complex and often incomplete data is a difficult, challenging task in nowadays extremely complex IT systems and on the Internet, overwhelmed with a huge amount of information. The problem itself is present in numerous different practical use-cases (e.g., in P2P streaming applications that rapidly gain more attention) and refers to the selection of the proper, efficient search algorithm. Research and commercial efforts resulted in a prolific offer of different algorithms that try to address this problem in the best possible way. Due to the huge complexity, intelligent algorithms are the most promising ones. However, everyday changing conditions impose finding even more advantageous approaches that will better cope with the problem, or at least address some “corner cases” better, than previously realized ones. In this paper, we propose a self-organizing approach inspired by bio-intelligence of slime molds that possesses distributive and autonomous properties with the goal to achieve a good query capability. A slime mold mechanism is adapted for search in an unstructured P2P system, and compared with Antnet and Gnutella search mechanisms. The benchmarks cover parameter sensitivity analysis, and comparative analysis. To validate the obtained results, a statistical analysis is performed. The obtained results show good scalability of slime mold algorithm and point to the selected “corner” cases where the slime mold algorithm has a total good performance (measured by different metrics).

Algorithms and Framework for Comparison of Bee-Intelligence Based Peer-to-Peer Lookup

Peer-to-peer has proven to be a scalable technology forretrieval of information that is widely spread among distributed sites and that is subject to dynamic changes. However, selection of a right search algorithm depends on many factors related to actual data content and application problem at hand. A comparison of different algorithms is difficult, especially if many different approaches (intelligent or unintelligent ones) shall be evaluated fairly and possibly also in combinations. In this paper, we describe a generic architectural pattern that serves as an overlay network based on autonomous agents and decentralized control. It supports plugging of different algorithms for searching and retrieving data, and thus eases comparison of algorithms in various topology configurations. A further novelty is to use bee intelligence for the lookup problem, spot optimal parameters’ settings, and evaluate the bee algorithm by using the architectural pattern to benchmark it with other algorithms.

Peer-to-Peer Overlay Network Based on Swarm Intelligence

As the number of information in the Internet constantly increases and the complexity of systems rapidly grows, locating and manipulating complex data has become a difficult task. We propose a self-organizing approach that combines purely decentralized unstructured peer-to-peer (P2P) with space based computing in order to effectively locate and retrieve information from a network. The approach is inspired by swarm intelligence, is distributive and autonomous. As the scalability is a common open issue both for unstructured P2P networks and for coordination models, our approach successfully copes with that by using a biologically inspired multi-agent system. Benchmarks demonstrate powerful query capabilities with a good scalability.

An Open Event-Driven Architecture for Reactive Programming and Lifecycle Management in Space-Based Middleware

Highly dynamic distributed applications often require flexible coordination among several autonomous components. Space-based middleware provides a suitable, data-driven coordination paradigm for such scenarios, where distributed peers exchange data and commands in a scalable and decoupled way using shared tuple spaces. In its basic form, such a middleware supports access to a data storage and (blocking) queries on the stored tuples. However, in many cases the injection of additional server-side logic would ease the development of complex applications, as the semantics of the tuple space can be adapted to domain-specific requirements.This paper introduces reactive programming features for XVSM, a space-based middleware that enhances the tuple space concept with powerful coordination mechanisms. We present a comprehensive extension mechanism that supports the execution of application logic in reaction to composite and time-based events. As an example for the feasibility of the approach, we provide a bootstrapped solution for a leasing mechanism that manages the lifetime of data in the space.