logo

Intelligent Interactive Distributed Systems



  • Ibis integration in AgentScape
    • The current communication layer in AgentScape depends on creation and deletion of communication endpoints (similar to Unix sockets) and provides secure end-to-end communication between hosts running AgentScape middleware. This approach provides a basic communication mechanism, but it is not very flexible or extensible. The Ibis project, also developed in the Computer Systems group at the VU, provides a uniform Java-based communication layer (the Ibis Portability Layer) that is specifically designed for usage in a grid and other distributed environments.
    • Research proposal:
      Investigate the feasibility and then (if feasible) replace the current communication infrastructure of AgentScape with the Ibis' IPL. Make sure that the communication layer provides authentication, confidentiality and integrity.

  • Dynamic and Robust Tree Overlays
    • Hierarchical structures support a wide range of network management applications. However, trees suffer from the effect of failures, as a node disconnection results in the disconnection of a branch from the main body of the tree. Minimizing the effect of failures means decreasing the number of nodes in the disconnected branch. This can be achieved by building reliability-driven trees in which more reliable nodes are placed closer to the root leaving the unreliable ones underneath. Moreover, in the case of failures, the system must be able to react and reconnect the disconnected nodes in the tree on the fly (dynamically).
    • Research proposal:
      This project focuses on the dynamic building and maintenance of tree overlays. The candidate must be able to explore the feasibility of existing mechanisms and potentially propose and implement other effective ones. The evaluation of the system will be based on simulations, thus good programming skills are required.

  • Web service composition, comparison of description languages using planning algorithm
    • Service Oriented Architectures are becoming more prevalent, with Web service technology as their main instantiation. Automated Web service composition, using planning algorithms, is used to create complex functionality, and to make it accessible on the fly. The properties used for planning are defined in web service description languages. Explore planning using three different web service description languages: WSDL, OWL-S and WSDL-S. WSDL is commonly used in workflow- and GRID based environments, describing the interface. OWL-S is a knowledge- based representation, describing semantic information. WSDL-S acts as a merger between the two previous approaches.
    • Research project:
      Explore configuration by planning, using these three languages. Evaluate and compare their applicability within the domain of configuration.

  • AgentScape Mobile
    • Mobile devices are traditionally unable to perform complex tasks or execute complex applications. Recently, with the advent of such devices as Apple's iPhone, Google's Android technology and the widespread adoption of Windows mobile, the abilty to execute complex applications is more readily available.
    • Research project:
      Investigate and develop an implementation of AgentScape for mobile devices. Initially it is proposed to develop a version of AgentScape for a simulator, but the goal is to have a running version on existing mobile hardware.

  • Security for AgentScape
    • Securing hosts from malicious agents entails monitoring every action that the agent attempts on the host. Whenever an agent makes a call to the middleware API, it is first mediated upon by a security manager. The security manager checks the system policy to determine if an action, such as migration and resource access, should be allowed or denied. For example, a host could decide that it is not willing to allow an agent to migrate anywhere except directly back to the owner's host, or, perhaps, to another host that has been properly authenticated to the current host.
    • Research project:
      The goal of this project is to implement a robust security manager in AgentScape. This security manager should monitor all the security relevant actions that Agents perform on a secured host.


  • Support Service Level Agreements in AgentScape
    • A Service Level Agreement (SLA) is an electronic contract between a service user and a provider, and specifies the service to be provided, Quality of Service (QoS) properties that must be maintained by a provider during service provision (generally defined as a set of Service Level Objectives (SLOs)), and a set of penalty clauses specifying what happens when service providers fail to deliver the QoS agreed. Although significant work exists on how SLOs may be specified and monitored, not much work has focused on actually identifying how SLOs may be impacted by the choice of specific penalty clauses. A trusted mediator may be used to resolve conflicts between the parties involved.
    • Research Proposal:
      The objectives of this work is to extend the AgentScape environment with an extension of WS-Agreement to support monitoring of violations and to enforce penalty clauses based on this monitoring of SLAs.

  • Support multiple and new languages for AgentScape agents
    • AgentScape currently supports Java and C agents. Support for Python agents is underway. In future, other languages can be supported, such as, lisp, ruby or C#. Determining the consequences of multiple binary agents and managing multiple programming languages form the main research questions in this project
    • Research project:
      Develop support for agents written in a new language and analyze the consequences thereof

  • Autonomic AgentScape Deployment on Distributed Networks
    • AgentScape should support automatic installation. Suppose that an agent could migrate to a host that has not yet installed AgentScape? An agent could first transfer all the necessary components of the AgentScape environment to the remote host and start the environment and finally migrate to it. Such a scheme would allow for easy and rapid deployment of AgentScape in a network.
    • Research project:
      Develop a rapid and user-friendly deployment program for AgentScape



login
For comments on this web-site, please contact: Michel Oey