Multiagent and Grid Systems

Many scientific workflow applications are driven by simulation generated data, or data collected from sensors or instruments, and the processing of the data is commonly done at a different location from where the data is stored. Moving large quantities of data among different locations is thus a frequently invoked process in scientific workflow applications. These data transfers often have high quality requirements on the network services, especially when the application requires steering from human interaction. Advanced networks such as hybrid networks make it feasible for high level applications to request network paths and service provisioning. However, current workflow applications tune the execution quality neglecting network resources, and by selecting only optimal software services and computing resources. Including network services in the resource scheduling adds an extra dimension for workflow applications to optimize the runtime performance. In this paper we present a system called NEtwork aware Workflow QoS Planner (NEWQoSPlanner) to complement existing workflow systems on selecting network resources in the context of workflow composition, scheduling and execution when advanced network services are available.

Large-scale agent-based software solutions need to be able to assure constant delivery of services to end-users, regardless of the underlying software or hardware failures. Fault-tolerance of multi-agent systems is, therefore, an important issue. We present two algorithms for an easy and flexible introduction of fault-tolerance to existing agent frameworks. The first algorithm is based on a new type of mobile agent, named Connection Agent, for efficient construction and maintenance of fault-tolerant multi-agent system networks. The algorithm has been experimentally verified, and then significantly optimized by relying on the mobility feature of Connection Agents. Secondly, a robust agent tracking technique based on a special type of agent, named RemnantAgent, is proposed.

Integrating distributed services into workflows comes with its own set of challenges, including security, coordination, fault tolerance and optimisation of execution time. This paper presents an architecture and implementation – nicknamed BeesyBees – that allows distributed execution of workflow applications in BeesyCluster using agents. BeesyCluster is a middleware that allows users to access distributed resources as well as publish applications as services, define service costs, grant access to other users services and consume services published by others. Workflows created in the BeesyCluster middleware are exported to BPEL and executed by BeesyBees agents in a distributed environment. Firstly, the paper demonstrates that engaging several agents to execute a workflow in a distributed fashion is more efficient than a centralised approach. It also discusses negotiation time tradeoffs in case of too many agents assigned to the task. An algorithm was proposed to migrate agents to such locations so that the workflow execution time is minimised. Secondly, it demonstrates that execution in the proposed environment is reliable even in case of failures. If a service fails, a task agent picks a new equivalent service at runtime. If one of task agents fails, another of remaining agents takes over its responsibilities. The communication between the middleware, agents and services is encrypted.

Wireless sensor networks (WSNs) represent a new form of pervasive and ubiquitous computing systems successfully exploited in many different application areas within which they will play an increasingly important role in future. However, the development of applications for WSNs is an extremely challenging and error-prone task, so that the need for high-level, effective programming approaches is quite evident. Among the programming paradigms proposed so far, the agent-based approach can be seen as an effective promising solution on the basis of which a few software platforms for WSNs have been already developed. This paper proposes an in-depth analysis of the only two available Java-based mobile agent platforms for WSNs: Mobile Agent Platform for Sun SPOT (MAPS) and Agent Factory Micro Edition (AFME). In particular, the architecture, programming model and basic performance of MAPS and AFME are described and compared. Moreover, a simple yet effective case study concerning a mobile agent-based monitoring system for remote sensing and aggregation is proposed. This case study is developed both in MAPS and AFME on Sun SPOTs so as to allow both an analysis of efficacy of their programming models and an evaluation of their performances.

We have explored mechanisms for converting organizations to an edge type organization. Beyond structural differences, organizations differ in information flow network and information sharing strategies. We review organizational adaptation. A model of computational organization and reorganization is presented using dynamic roles. In addition to self-organization, our model allows human oversight and guided reorganization. This article lays a foundation for automatic organizational adaptation and human supervision. Our model is exemplified with simulated soccer.

We present a formal model for agent-oriented Virtual Organisations (VOs) for service grids and we study an associated operational model for the creation of VOs. The model is intended to be used for describing different service grid applications based on multiple agents and, as a result, it abstracts away from any realisation choices of service grid applications, the agents involved to support the applications and their interactions. Within the proposed framework VOs are created within societies of agents, where agents are abstractly characterised by goals and roles they can play within VOs. In turn, VOs are abstractly characterised by the agents participating in them with specific roles, as well as the workflow of services and corresponding contracts suitable for achieving the goals of the participating agents. We illustrate the proposed framework with an earth observation scenario, we discuss implementation issues, and we compare our approach with existing work.