2008’ Future Internet Autonomic Network Management May 2008 Joon-Myung Kang Distributed Processing & Network Management Lab. Dept. of Computer Science and Engineering POSTECH, Korea eliot@postech.ac.kr Autonomic Network Management -1- •DPNM Lab. Contents 1. Introduction 2. Autonomic Computing vs. Autonomic Networking 3. Autonomic Computing Environment 4. Research Challenges 5. Conclusions Autonomic Network Management -2- •DPNM Lab. The Problem of Complexity Autonomic Network Management -3- •DPNM Lab. Moving Early majority CCS Autonomic Network Management -4- •DPNM Lab. Network Environment Billing VHE ISP SIP Proxy Server Signalling Gateway WAPAccounting The Internet Context-aware information Centre Satellite Core Network Broadcast Networks (DAB, DVB-T) CDMA, GSM, GPRS 4G IP-based micro-mobility Bluetooth Zigbee WiBro, HSDPA Wireless LANs Autonomic Network Management -5- •DPNM Lab. A Traditional OSS/BSS Autonomic Network Management Autonomic Network Management -6- •DPNM Lab. The Problem – Managing Complexity The Complexity of system design and management keeps increasing – Stovepipe systems: best-of-breed functionality but integration nightmares – Increased technology overwhelms users and administrators • Different devices have different programming models and interaction models • Different management tasks and integration types require different skill levels The complexity of business is also increasing – People are demanding a pervasive presence – Many types of business LOSE MONEY if they can’t react fast enough – Varieties of threats, problems, and non-optimized behavior keeps increasing Behavioral complexity is also increasing – Everything is interconnected, requiring different policies and functions – Too complex to predict, needs too high a skill level, not enough people! Autonomic Network Management -7- •DPNM Lab. Shortcomings - Infrastructural Architectural issues – – – – Integration issues Data redundancy Synchronization problems Application authorization issues Vendor and Application “lock in” Autonomic Network Management – Isolated Data Silos – Administrative nightmare – Integration/customization nightmare – Transition from legacy systems to a new OSS -8- •DPNM Lab. Stovepipes are Everywhere! Autonomic Network Management -9- •DPNM Lab. Some Effects of Complexity Expensive – Cost of management by administrators is increasing (CAPEX, OPEX) CAPEX(Capital Expenditure): expenditures creating future benefits. OPEX (Operational Expenditure): on-going cost for running a product, business, or system Fragile – Complex interdependencies make it hard to diagnose and fix problems – More prone to human error (additional cost) – Upgrades, performance tuning, re-purposing all suffer Inflexible – Reluctance to change infrastructure once it is working – Does not support agile business (new software, business processes) Worsening – Technology innovations typically exacerbate the problem, preventing product innovations from being deployed Solution: Self-managing systems Autonomic Network Management - 10 - •DPNM Lab. More Effects – Constituency Separation Different constituencies have different terms, grammars, and needs – Service Level Agreement meaning changes – Business “speak” vs. networking commands – Different representations (e.g., use of UML) Relating network services and resources to business needs – Not reflected in EMS and NMS design – Lack of policy controlling allocation – Lack of ability to • Incorporate new knowledge • React in a timely manner to changes Autonomic Network Management - 11 - •DPNM Lab. Motivation of Autonomic Management Complexity is growing – – – – Telecommunications industry has changed dramatically Explosive growth of the Internet The proliferation of mobile technologies Fixed mobile convergence Autonomic Network Management – Simplify network management process by automating and distributing the decision making processes involved in optimizing network operation – Enable expensive human attention to focus more on business logic and less on low level device configuration processes Autonomic Network Management - 12 - •DPNM Lab. Autonomic Network Management - 13 - •DPNM Lab. What is Autonomic Computing? Autonomic – Pertaining to operating system that responds automatically to problems or system failures. Autonomic Nervous System – handles many crucial (but what we'd consider mundane) functions without requiring any conscious though on our part. – When we run, it increases our heart and breathing rates. If we get too hot, it redirects blood flow to "cool us". If we turn a light on or off, it adjusts our pupils for maximum visual activity. – This leads us to describe an approach we call "autonomic computing.“ Autonomic computing – A computing environment with the ability to manage itself and dynamically adapt to change in accordance with business policies and objectives. Autonomic Network Management - 14 - •DPNM Lab. Future Vision of Autonomic Computing? Machines will take over all management tasks, rendering humans superfluous. Wrong Machines will free system administrators to manage system at a higher level. Right Autonomic Network Management - 15 - •DPNM Lab. A Misconception About Autonomics Autonomics is not the four (in)famous self-functions (self-configure, protect, -heal, and –optimize) – These do not define an autonomic system – These are benefits resulting from an autonomic system Autonomics is rooted in the following capabilities – Self-knowledge • We can’t configure what we don’t know! – Ability to understand what is happening to our surroundings • Learn from and reason about sensed data – Inspiration from biology, sociology, economics, … • New ways to build and organize management functionality • Notion of maximization of “social welfare” of network service – Link to business rules • Network services and resources adapt to change Autonomic Network Management - 16 - •DPNM Lab. Autonomic Computing Attributes Increased Responsiveness Business Resiliency Discover, diagnose, and act to prevent disruptions Adapt to dynamically changing environments Operational Efficiency Secure Information and Resources Tune resources and balance workloads to maximize use of IT resources Autonomic Network Management Anticipate, detect, identify, and protect against attacks - 17 - •DPNM Lab. Self-Configuration Configuration is governed by high-level policies – BUSINESS objectives that specify WHAT is to be accomplished, but not HOW New way of component interaction – New component adapts itself to how other components in the environment are working – Existing components adapt to the presence of the new component – This cannot be done unless the component “knows itself” and its environment DEN-ng models this interaction using the concepts of capabilities and constraints – DEN-ng: new version of Directory Enabled Networks (DEN) – Common information model to translate business rules into device configuration Autonomic Network Management - 18 - •DPNM Lab. Self-Healing Error detection and correction is HARD – Network management is a good (well, bad) example of how NOT to do this – Predictive failure analysis is still magic. Why? Because there is no selfknowledge! • Multiple incompatible knowledge sources • Self-healing affects ALL phases of the control loop Autonomic computing detects errors – Based on self-knowledge (critical role of DEN-ng) – Once an error is known, it can be repaired – or can it? (critical role of DEN-ng) • The problem is more difficult than this – component and system behavior can change • How do we know if the change is desired? Autonomic Network Management - 19 - •DPNM Lab. Self-Optimization Current research efforts are – Oriented mostly on optimizing system performance – Just the aspect of tuning is complex • Hardware and software dependencies, backwards compatibility, unclear semantics, etc. • Tuning one component can adversely affect others It’s also about – – – – Using the right resources for a given task Ensuring that tasks with higher business importance get the resources they need Adapting to recognized environmental and service usage patterns Learning through action Autonomic Network Management - 20 - •DPNM Lab. Self-Protecting Current problems that need to be addressed – Detecting threats and malicious operations – Prevent the cascading of uncorrected errors – Transitioning from reactive to proactive systems It can NOT be done without using an information model to represent the behavior – Capabilities and constraints – Patterns and roles Interaction between hardware and software to protect a system Autonomic Network Management - 21 - •DPNM Lab. Any Self-Function Can Change The system must have models pre-defined – Characteristics and behavior of itself – Needs of the user – Environmental constraints Self-configuration can change – Based on different user needs – Based on environmental conditions Decision-making requires policy to govern the system and its interactions – Reasoning enables deduction of cause and effect – Learning functions enable the system to improve Autonomic Network Management - 22 - •DPNM Lab. Just Add Water and Stir The integration of self-configuration, -healing, -protection, and – optimization is critical – If that happens, these separate concepts will merge into more powerful concepts – For example, self-maintenance is the holistic combination of all four of these principles • Using anti-virus software as an example, the system will proactively and upgrade its functionality • Adjustment of workload in response to changing conditions (e.g., component failures) and environment (e.g., new users running new apps) Self-management is more than the sum of the self-management of its individual components Autonomic Network Management - 23 - •DPNM Lab. Autonomic computing architecture concepts Autonomic computing system – A computing system that senses its operating environment, – models its behavior in that environment, – and takes action to change the environment or its behavior. - Autonomic Computing reference architecture Autonomic Network Management - 24 - •DPNM Lab. Autonomic computing architecture concepts Managed resource – An entity that exists in the run-time environment of an I/T system and that can be managed. Touchpoint – The interface to an instance of a managed resource, such as an operating system or a server. – A touchpoint implements sensor and effector behavior for the managed resource. – And it maps to the sensor and effector interfaces to existing interfaces Touchpoint autonomic managers – An autonomic manager that works with managed resources through their touchpoints. Autonomic Network Management - 25 - •DPNM Lab. Autonomic computing architecture concepts Orchestrating autonomic managers – An autonomic manager that works with other autonomic managers to provide coordination functions. Integrated solutions console – A technology that provides common, consistent user interface, based on industry standards and component reuse, and can host common system administrative functions. – The IBM Integrated Solutions Console is a core technology of the IBM Autonomic Computing initiative that uses a portal-based interface. Autonomic Network Management - 26 - •DPNM Lab. Autonomic computing architecture details Autonomic manager internal structure – Knowledge • Standard data shared among the monitor, analyze, plan and execute functions of an autonomic manager, such as symptoms and policies. Autonomic Network Management - 27 - •DPNM Lab. Autonomic computing architecture details Knowledge types Solution Topology Knowledge • Captures knowledge about components and constructions and configuration for business system. • For AM to install or configure components, installation and configuration knowledge is captured in a common installable unit. Policy Knowledge • A policy is a knowledge that is consulted to determine whether or not changes need to be made in the system. • Autonomic computing system requires a uniform method for defining the policies that govern decision-making for autonomic managers Problem Determination Knowledge • It includes monitored data, symptoms and decision trees. • As the system responds to actions, newly learned knowledge can be collected within AM. • Autonomic Computing System requires a uniform method for representing problem. Autonomic Network Management - 28 - •DPNM Lab. Autonomic computing architecture details Autonomic manager is a component that implements the control loop. – Monitor Function • the function that collects, aggregates, filters and reports details (e.g. metrics, topologies) – Analyze Function • the function that models complex situations to understand current system state. – Plan Function • the function that structures the actions needed to achieve goals and objectives. – Execute Function • the function that changes the behavior of the managed resource using effectors. Autonomic Network Management - 29 - •DPNM Lab. Autonomic computing architecture details Managed Resource – A controlled system component. ex) a server, a router, a cluster or business application etc. Manageability Interface – A service of the managed resource that includes the Sensor and the Effector used by an autonomic manager. – This is for autonomic manager to monitor and control the managed resource. Autonomic Network Management - 30 - •DPNM Lab. Autonomic computing architecture details Sensor – A set of “get” operations that retrieve information about the current state of a managed resource. – A set of management events (unsolicited, asynchronous messages or notifications) that can occur. Effector – A collection of “set” operations that allow the state of the managed resource to be changed in some important way. – The operations that managed resource can use to make request. Autonomic Network Management - 31 - •DPNM Lab. Autonomic computing architecture details Evolving towards Autonomic Computing Systems Autonomic Network Management - 32 - •DPNM Lab. Autonomic computing architecture details An evolution, not a revolution Autonomic Network Management - 33 - •DPNM Lab. Autonomic Network Management - 34 - •DPNM Lab. Current Network Management Deficiencies Aggregates of elements may exhibit behavior not predictable from knowledge of individual behaviors – “The whole may be greater than the sum” Causal determinacy still limited by simple statistical analysis and rudimentary correlation approaches – Precompiled diagnostic processes required to guide approaches No ability of the system to “go beyond” precompiled knowledge and procedures – “Human-in-the-loop” is still the order of the day All current techniques require “human-in-the-loop” back-end analysis – Extensive system, deployment, and technology knowledge – Drives us CAPEX and OPEX Autonomic Network Management - 35 - •DPNM Lab. Autonomic Networking Biology, Sociology, and Economics can Inspire Better Networks! – – – – Technical complexity: human body technology, devices Business complexity: macro-economics e- and m-Commerce Behavioral complexity: social interaction service composition Operational complexity: healing anti-virus, configuration management Autonomic Network Management - 36 - •DPNM Lab. Business to System Interactions Autonomic Network Management - 37 - •DPNM Lab. Autonomic Control Loop Autonomic Network Management - 38 - •DPNM Lab. Autonomic Management System Conceptual Representation Autonomic Network Management - 39 - •DPNM Lab. Autonomic Management System The control loop – Controlled by an autonomic manager that influences the deployment of the policies that effect decision making within the loop DEN-ng (Directory Enabled Network new generation) – Finite state machines to model behavior and augmenting with ontological models that embody semantic information that cannot be represented in the UML – A comprehensive information model for telecommunications, capturing everything from business concepts (products, service level agreements, and customers) to low-level device functionality (packet marking, forwarding, and queuing) Model-based policy processing component – Incorporate policy conflict analysis algorithms that (1) elaborate newly defined/modified policies so that conflicts are easier to detect, (2) detect sets of policies that will or potentially could conflict, given certain network context, and (3) resolve conflicts by modifying or removing policies based on separate resolution policies or by referring back to the appropriate policy author for a decision Autonomic Network Management - 40 - •DPNM Lab. Autonomic Management System Autonomic Management Architecture Autonomic Network Management - 41 - •DPNM Lab. Autonomic Management System FOCALE – Foundation Observation Comparison Action Learn rEason – Based on the observation that business objectives, user requirements, and environmental context all change dynamically – Two control loop • Maintenance control-loop – Used when no anomalies are found • Adjustment control loop – Used when one or more policy reconfiguration actions must be performed, and/or new policies must be codified and deployed – It is unreasonable to assume that a single entity can maintain all the information required to realize the FOCALE control loops for large scale networks containing large numbers of heterogeneous devices – FOCALE must be a distributed architecture, to the degree that even individual network devices may incorporate autonomic management software, implementing the maintenance and adjustment control loops Autonomic Network Management - 42 - •DPNM Lab. FOCALE Autonomic Network Management Architecture Autonomic Network Management - 43 - •DPNM Lab. FOCALE Autonomic Network Management Architecture Main functional components – AM • Independent of the vendor-specific functionality/data of the underlying managed resource(s), which facilitates easier communication between AMEs for coordination of management decision making – MBTL • Indepth knowledge of the managed resource(s) to enable it to translate normalized vendor-specific data gathered from the managed resource(s) into DEN-ng compliant vendor-neutral data to pass to the policy analyer/PDP and vice versa for configuration commands Autonomic Network Management - 44 - •DPNM Lab. Basis of MBTL Use of ontologies to identify cognitive equivalence across heterogeneous data models Autonomic Network Management - 45 - •DPNM Lab. Prototype Implementation Implementation – Single FOCALE AME that targets aspects of traffic conditioning in a simulated IP-based network of an ISP, over which customers are offered a small number of communications services – Simulated network is very loosely coupled to the AME implementation – Plan to replaced the simulation with real routers that will be configured by CLI commands generated by the AME and that will provide context information to the AME via SNMP – OPNET based simulation is configured to emit information relating to network events and read and apply new router configurations generated by the AME Autonomic Network Management - 46 - •DPNM Lab. Prototype Implementation Autonomic Network Management - 47 - •DPNM Lab. Autonomic Networking Scenario Different networks, technologies, and business rules Conflicting resource and service requirements Different capabilities Maximize network services according to business rules across all customers Autonomic Network Management - 48 - •DPNM Lab. Autonomic Computing and Autonomic Networking Autonomic Computing – Coined by IBM as an analogy to the autonomic nervous system – Attempts to manage the operation of individual pieces of IT infrastructure through the introduction of an autonomic manager that implements an autonomic control loop in which the managed element and the environment in which it operates is monitored – Autonomic control loop: monitor, analyze, plan, and execute components The vision of autonomic computing – Self-managing IT infrastructure • Self-configure, self-optimize, self-heal, self-protect (self-* behavior) Autonomic Networking – Burgeoning research area that seeks to integrate results from disciplines ranging from telecommunications network management to artificial intelligence and from biology to sociology Autonomic Network Management - 49 - •DPNM Lab. Autonomic Computing and Autonomic Networking Focus of research in autonomic network management – On the development of highly distributed algorithms that seek to optimize one or more aspects of network operation and/or performance – Investigating the potential use of biologically-inspired algorithms and processes – Although work on the development of decentralized, self-management algorithms is crucial, the deployment of these algorithms will not be sufficient – Equally important will be the flexible specification and enforcement of the goals these algorithms collectively seek to achieve Policy-based Network Management – Appropriate management paradigm to facilitate higher-level, human-specified cognitive decision making Little work has been done to date on integrating distributed selfmanagement algorithms with policy-based management Autonomic Network Management - 50 - •DPNM Lab. Autonomic Computing/Networking People express at a high level what they want the system to achieve – Level could be business, or IT The system strives to manage its own behavior to optimally satisfy these multiple criteria, given resource constraints – Resources: Hardware, software, cost – Tradeoffs among multiple criteria must be clear – Self-{configuration, healing, optimization, protection, …} are general classes of behavioral criteria, but don’t define Autonomic Computing – The AC challenge is to develop the right technologies and architecture People and self-managing systems will work together iteratively, in partnership with one another – People will do what they’re best at – Systems will gradually assume more management burden • As they become more competent to do so • As people become more comfortable with this Autonomic Network Management - 51 - •DPNM Lab. Autonomic Network Management - 52 - •DPNM Lab. Autonomic Computing Element (ACE) An ACE is an abstraction that enables an autonomic system to manage the functionality of a new or legacy managed resource ACEs manage their own behavior in a standard way Does not have autonomic capabilities Autonomic Network Management - 53 - •DPNM Lab. How to Build an Autonomic Element AC toolkit: Autonomic Integrated Development Environment (AIDE) AIDE is part of IBM “Build To Manage” initiative – WSDM (Manageability) support Eclipse tool generates java ‘stubs’ for multiple runtimes – Apache Muse (default) – OSGi, Eclipse 3.1 – WebSphere App Server Currently available at – http://www.alphaworks.ibm.com In use by several IBM business partners and customers Autonomic Network Management - 54 - •DPNM Lab. IBM’s Autonomic Control Loop ACEs are building blocks that are arranged to provide higher-level system behavior Autonomic Network Management - 55 - •DPNM Lab. Autonomic Networking in the ACF Autonomic Network Management - 56 - •DPNM Lab. Autonomic Computing Element (ACE) Autonomic Network Management - 57 - •DPNM Lab. IBM Touchpoint vs. FOCALE MBTL Autonomic Network Management - 58 - •DPNM Lab. Autonomic Computing Scalability Autonomic Network Management - 59 - •DPNM Lab. Motorola Labs Autonomic Element Autonomic Network Management - 60 - •DPNM Lab. Scalability in an Autonomic Network Autonomic Network Management - 61 - •DPNM Lab. Autonomic Computing vs. Autonomic Networking Autonomic COMPUTING – – – – – – Assumes homogeneous elements Control loop is straightforward System adaptation based on ITIL (or eTOM) Focused on self-config, -heal, -optimize, -protect Policies determine what actions to take Goal is to orchestrate static behavior Autonomic NETWORKING – – – – Assumes heterogeneous elements Control loop adapts based on network, task, … System adaptation based on DEN-ng & ontologies Focused on self-governance and –knowledge, which drive all other self-* functions – Policies determine overall behavior of the ACE – Goal is to orchestrate dynamic behavior Autonomic Network Management - 62 - •DPNM Lab. Autonomic Network Management - 63 - •DPNM Lab. Research Challenges Key Challenges – Providing self-* functionality • Mapping management information and commands in the legacy programming models to an autonomic programming model • Orchestrate behaviour in a way that legacy devices can understand – The complexity of applying various techniques may preclude their use in certain platforms • • • • Model-Driven Engineering Aspect-Oriented Software Development Reverse Engineering Generative Programming Autonomic Network Management - 64 - •DPNM Lab. Challenge: Architecture AE: How to coordinate multiple threads of activity? – AE’s live in complex environments – Multiple task instances and types Define set of fundamental architectural principles from which self-* emerges S – concurrent, asynchronous E Autonomic Manager – Multiple interacting expert modules Analyze AE: How to detect/resolve conflicts arising from – Internal decisions by independent expert modules – External directives (possibly asynchronous) – Internal policies vs. external directives Plan Monitor Knowledge S Execute E Managed Element System-level: Enable more flexible, service-oriented patterns of interaction An Autonomic Element – As opposed to traditional top-down, hierarchical systems management – Multi-agent architecture – Communication – Representing and reasoning about needs, capabilities, dependencies 65 Autonomic Network Management - 65 - •DPNM Lab. Challenge: Policy Human interface – Authoring and understanding policies – Avoiding or ameliorating specification errors Policy: “Set of guidelines or directives provided to autonomic element to influence its behavior” Developing a universal representation and grammar – Many different application domains, disciplines – Many different flavors of policy – Covers service agreements too? S E Autonomic Manager Analyze Monitor Algorithms that operate upon policies (and agreements?) Plan Knowledge Execute S E Managed Element – Automated derivation of actions (e.g. planning, optimization) – Automated derivation of lower-level policies from high-level policies • E.g. “Maximize profit from this set of service contracts” Conflict resolution – Both design time and run time – Need to establish protocols, interfaces, algorithms 66 Autonomic Network Management - 66 - •DPNM Lab. Three flavors of (policy = “decision-making guide”) Possible State s1 Action rule – If (S) then do a2 – Results implicitly in desired state s2 a1 Goal Current State S – Achieve a most desired state s2 – Compute a2 most likely to result in s2 – Assumes that most desired state can be determined a priori Possible State s2 a2 a3 Possible State s3 Utility function – Achieve state s with maximal net value V(s) – C(aSds) – Benefit and burden of being explicit about value – States have intrinsic value; value of policy is a derived quantity [More levels of code hierarchy] Machine code Workflows Programming Rules Decisiontheoretic Planning Element Optimization Adapters, Translators Element Actions Generative Goals Planning utility functions Modeling, Optimization System utility functions Higher-level specifications 67 Autonomic Network Management - 67 - •DPNM Lab. Challenge: Human-System Interface Develop new languages, metaphors and translation technologies that enable humans to monitor, visualize, and control AC systems – Specify goals and objectives to AC systems, and visualize their potential effect – Techniques must be – Sufficiently expressive of preferences regarding cost vs. performance, security, risk and reliability – Sufficiently structured and/or naturally suited to human psychology and cognition to keep specification errors to an absolute minimum – Robust to specification errors Autonomic Network Management - 68 - •DPNM Lab. Challenge: Learning Single element level Establish theoretical foundation for understanding and performing learning and optimization in multiagent systems. – AE needs to learn a model of itself and environment quickly; environment is noisy, and dynamic in both state and structure – On-line, so exploration of the space can be costly and/or harmful – May be several hundreds of tunable parameters! – Maybe only a few dozen are relevant, but which ones? – Some of them can only be changed upon reboot – is it worthwhile? System level – Multi-agent system: several interacting learners – What are good learning algorithms for cooperative, competitive systems? – What are conditions for stability? – What is sensitivity to perturbations? – Opportunities for layered learning Autonomic Network Management - 69 - •DPNM Lab. Challenge: Negotiation Develop and analyze – Methods for expressing or computing preferences – Negotiation protocols – Negotiation algorithms Establish theoretical foundation for negotiation – Explore conditions under which to apply – Bilateral – Multi-lateral (mediated, or not) – Supply-chain – Study how system behavior depends on mixture of negotiation algorithms in AE population Autonomic Network Management - 70 - •DPNM Lab. Challenge: Control and Harness Emergent Behavior Understand, control, and exploit emergent behavior in autonomic systems – How do self-*, stability, etc. depend on – Behaviors and goals of the autonomic elements – Pattern and type of interactions among AEs – External influences and demands on system – Invert relationship to attain desired global behavior – How? – Are there fundamental limits? Develop theory of interacting feedback loops – Hierarchical – Distributed 71 Autonomic Network Management - 71 - •DPNM Lab. Conclusions Autonomic Computing is a grand challenge, requiring advances in several fields of science and technology – Policy, planning, learning, knowledge representation, multi-agent systems, negotiation, emergent behavior – Human-system interfaces Integrating these technologies to support self-management in complex, realistic environments is a research challenge in itself – What are the best architectures and design patterns? Role of (multi-)agent systems? – Building system prototypes is key to developing and validating AC technology and architecture Technology to Manage Technology – Autonomic Computing/Networking does not replace people, it empowers them – Context-awareness is achieved by enabling the system to reconfigure itself, based on • Changing business goals • Changing user needs • Changing environmental conditions – Reconfiguration is model-driven, and uses ontologies to reason to build its conclusions – This is slowly becoming a reality, but should not be a panacea Autonomic Network Management - 72 - •DPNM Lab. How to realize Autonomic Network Management? Thank you for your attention Autonomic Network Management - 73 - Joon-Myung Kang •DPNM Lab. eliot@postech.ac.kr References 1. 2. IBM Corporation, “An architectural blueprint for autonomic computing,” White paper, 2003 J.O. Kephart and D.M. Chess, “The vision of autonomic computing,” Computer, vol. 36, no. 1, Jan. 2003, pp. 41-52 3. J.O. Kephart., “Research Challenges of Autonomic Computing,” Proceedings of the 27th International Conference on Software Engineering, 2005 4. John Strassner, “Autonomic Systems and Networks: Theory and Practice”, NOMS 2008 Tutorial, 2008 5. R. Sterritt, “A concise introduction to autonomic computing,” Journal of Advanced Engineering Informatics, Engineering Applications of Artificial Intelligence, Elsevier Publishers, Vol. 19, pp. 181-187, 2005 6. Brendan Jennings, “Towards Autonomic Management of Communications Networks”, IEEE Communications Magazine, Vol. 45, Issue 10, pp. 112-121, 2007 7. http://www.autonomiccomputing.org/ 8. http://www.research.ibm.com/autonomic/ 9. http://www.autonomic-communication-forum.org/ 10. http://dnac.org/autonomic-networking/ Autonomic Network Management - 74 - •DPNM Lab.