Workshop on Adaptive Methods in Autonomic Computing Systems (AMACS)
2nd International Workshop on Engineering Emergence in Decentralised Autonomic Systems (EEDAS 2007)
2nd Workshop on Hot Topics in Autonomic Computing (HotAC II)
1st International Workshop on Policy-Based Autonomic Computing (PBAC 2007)
Workshop on QoS in Autonomic Communication Networks (ACN 2007) (Canceled)
Autonomic Networking: Theory and Practice
John Strassner, Motorola Research Lab
Schedule: 8:30AM – 12:00PM, 1:30PM – 5:00PM, June 11 2007
The increasing complexity of computing systems is beginning to overwhelm the capabilities of software developers and system administrators to design, evaluate, integrate, and manage these systems. Autonomic networking is a collection of technologies and mechanisms that enable systems and components to govern their behavior in accordance with policies. This enables business needs to drive the services and resources available from the network.
This tutorial is aimed at giving the participant a reasonably deep understanding of the motivation for autonomic networking, concentrating on the semantic and behavioral aspects of network management. After defining autonomic computing and networking, this tutorial will first describe different architectural styles of autonomics, and then elaborate on how to build autonomic elements. Then, a novel autonomic networking architecture will be examined in detail, including examining the motivation for this architecture, the role that each of its components play, and up to date progress in implementing this architecture. This theory will be reinforced with use cases and practical examples, including a demonstration of ongoing research work in Motorola Labs.
Adaptive Middleware for Autonomic Communications
Simon Dobson, Dave Lewis, and Paddy Nixon, University College Dublin
Schedule: 8:30AM – 12:00PM, June 11 2007
Traditional systems engineering has concentrated on the correctness and robustness of systems
within a well-defined and clearly delineated environment. Recent trends in open systems, ad hoc
networking, self-management and pervasive computing are leading to a situation in which a system’s
configuration and behaviour must adapt to a more dynamic environment in order to continue to
deliver its required services.
Autonomic computing and communications form an active area of current research aiming to address
these issues. From a systems perspective, autonomic systems aim to reduce the costs of ownership
by devolving some responsibility for a system’s administration onto the system itself. From a
programming perspective, this leads to a whole new range of concerns whereby programmers must
address self-configuration, self-management, self-optimisation and a range of other self-* properties.
This poses significant challenges for developers.
The traditional approach to managing complexity in distributed systems has been through middleware
providing a platform which abstracts-away from the details of the underlying communications
system. While traditional middleware does not directly address self-* issues, recent trends
in middleware design are leading to platforms that are themselves significantly self-managing and
self-optimising, and which lend themselves very well to the development of highly autonomic applications
The purpose of this tutorial is to introduce attendees to the issues and tools – both existing and
emerging – in the area of adaptive and autonomic middleware. The emphasis is on understanding
the available tools and techniques, and how these may be used together to build systems with
self-* properties. These techniques include decentralisation, peer-to-peer networking, gossipping,
knowledge-based decision-making, ontologies and content-directed networking. Such techniques are
to a large extent technology-neutral, and so may be applied across a range of scales from sensor networks to enterprise information systems.
Attendees will acquire a familiarity with the business and technological drivers that affect adaptive
autonomic systems development. They will acquire an high-level view of the application of these
techniques to existing and emerging tools, platforms and standards.
The target audience is developers wanting to understand how to develop self-managing applications.
It would also be of interest to systems architects and other strategic managers wanting to understand
the impact that this technology will have on their core business systems.
The tutorial will be fully supported by slides on a web site, as well as a comprehensive reading list
for those wishing to dig deeper into the topics covered.
About the Presenters:
Simon Dobson and Paddy Nixon are with UCD Dublin’s Systems Research Group, while Dave
Lewis is a researcher in the School of Computer Science and Statistics at Trinity College Dublin.
Between them they have over forty years’ experience in large-scale distributed systems, advanced
communications and programming languages within both academia and industry.
Design and Implementation of Intelligent Agents for Autonomic Computing
Benno Overeinder, Vrije University Amsterdam
Schedule: 1:30PM – 5:00PM, June 11 2007
Agent systems have been used for various applications in complex domains. The agent approach decomposes
a problem in small autonomous entities, while the agents manage and schedule their activities to achieve a
goal. Multiple agents can coordinate their activities to achieve a common goal; to facilitate this, a number of
cooperation and coordination schemes exist.
By the autonomous behavior of multi-agent systems, many similarities exist with autonomic computing.
In autonomic computing (in its most general sense) entities can be both software (applications, services,
etc.) as well as hardware (host, network, hard disk, etc.). In general, both software and hardware entities
are monitored by software components. The software components monitoring and managing the entities in
an autonomic system can be considered a multi-agent system using cooperation and coordination schemes
well-known in AI research.
In the tutorial, an introduction to intelligent agents will be given. Typically issues like weak and strong
notion of agency will be introduced, defining the autonomous behavior of agents in both system terms and
their translation into high-level (human) notions used in reasoning.
After a general introduction to agents, an in-depth presentation of system requirements for multi-agent
systems will be given. General principles of agent runtime environments (middleware) will be introduced and
the specific implementation in the AgentScape middleware discussed. Multi-agent systems also set a number
of requirements to agent platforms, e.g., life-cycle management, (optionally) mobility, security and resource
management. All these aspects will be discussed and examples in the AgentScape middleware will be given.
In the third part, a practical example will be presented for a resource negotiation scenario where agents
need to acquire resources to complete a task. Resources are also autonomous entities and agents have to
negotiate to settle the terms of resource access and usage.
Optionally, the tutorial can include a number of practical hands-on exercises. In these exercises, different
aspects of agents, e.g., cooperation, coordination, negotiation, etc., can be exemplified, resulting in a deeper
understanding of these concepts. The agents will be programmed in the Java programming language and run
on the AgentScape middleware.
The tutorial is targeted to any researcher and/or professional that is active in autonomic computing and has
an interest in the design and implementation of autonomic computing architectures. The tutorial presents a general agent-based approach to autonomic computing rather than a specific topic in autonomic computing.
With the introduction of intelligent agents, some common techniques from Artificial Intelligence, such as
cooperation, coordination, and/or diagnosis, will be presented in context of the autonomic computing.
Duration of the Tutorial:
Half-day tutorial, or a full-day tutorial including a practical hands-on session (about 2 + 3 hours sessions).
No prerequisite knowledge is required, other than some general knowledge of distributed systems (which is
considered to be known to the autonomic computing audience). For the practical hands-on, a basic knowledge
of the Java programming language is useful.
Schedule: 3:00PM - 3:30PM, June 12 2007
Policy-Driven Autonomic Management in Enterprise-Scale Information
Vibhore Kumar, Brian F. Cooper, Greg Eisenhauer, Srihari
Govindharaj, Chaitanya Karlekar, Mohamed Mansour, Karsten Schwan, Sangeetha
Seshadri, Balasubramaniam Seshasayee
(Georgia Tech and Yahoo! Research)
A Standards Based Self-managing Resource Framework
Peter Brittenham, Balan Subramanian
(IBM Tivoli Autonomic Computing)
Conventional WSDM: Plumbing for Autonomic Computing
Andrew Eberbach, Balan Subramanian
(IBM Tivoli Autonomic Computing)
Virtualized Electronic Trading with EMS
Duncan Johnston Watt
Adaptive HP Infrastructure for SAP
(HP Labs/SAP Research)
SLA Decomposition: Translating Service Level Objectives to System Level Thresholds
IPOP: A Self-Organizing Virtual Network for Wide-Area Environments
Arijit Ganguly, David Wolinsky, P. Oscar Boykin, Renato J.
(University of Florida)