Management of Dynamic Networks and Services Olivier Festor, Radu State
Total Page：16
File Type：pdf, Size：1020Kb
Management of Dynamic Networks and Services Olivier Festor, Radu State To cite this version: Olivier Festor, Radu State. Management of Dynamic Networks and Services. Concordia Summer Workshop on Services Engineering and Network Management, Concordia Institute for Information Systems Engineering (CIISE), Aug 2003, Montréal, Canada, 23 p. inria-00107670 HAL Id: inria-00107670 https://hal.inria.fr/inria-00107670 Submitted on 19 Oct 2006 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Preamble : Management • All one does to keep the subject : Management of Dynamic – Healthy (up and running efficiently) – Evolving positively Users Networks & Services • Fundamental operations Customers – Monitoring A sample of candidate technologies and a case study • Observe the subject at work ☺ – Control Objectives •Act Commerciaux Olivier Festor, Radu State (e.g. Gold • change or constrain the behavior of the Service) Ph.D., Hab. Ph.D. managed subject – Plan its evolution Services •Why ? (e.g. Virtual The MADYNES Research Team Private Network) LORIA – INRIA Lorraine – Service quality 615, rue du Jardin Botanique – Reduce costs 54602 Villers-lès-Nancy Network – With limited risks Elements France (e.g. PHD • Keeping the right level of quality configuration, {Olivier.Festor,Radu.State}loria.fr …) © MADYNES 2003 -1- © MADYNES 2003 -2- Preamble : “management” wording in the presentation Tutorial Outline • Growing Dynamics and their • SyncML-DM : an Approach for Management Impact on Management (20 minutes) Managing Dynamic Devices (60 min) – Dynamics – Representation Protocol for Device – Collapsing time scales Management Device Management Network Management – The need for automation and – Device Management Protocol autonomy – Standardized Objects Service Management Application Management – Management solutions for the – Device Management Tree dynamic world – Security for Device Management – Standardized approaches • Case studies (40 min) – Emerging alternatives Configuration Management – Ad hoc networks management • JMX : Dynamic Management Fault Management • Conclusion & discussion (15 min) for/through the Java World (60 min) Accounting Management – Basic Concepts Performance Management – Dynamic components • Attribute, method & notification Security Management level • Managed Object level – Remoting – Unless explicitly stated otherwise. – Implementations – Application domains & instrumentation patterns © MADYNES 2003 -3- © MADYNES 2003 -4- Dynamic systems we are interested in : Dynamics Everywhere in Networks and Services ! • Topology related dynamics • In general : – Ad hoc networks : connectivity, reachability, autonomy – P2P networks & overlays : availability, presence, contribution, « A system that changes its state over time » topology [Jakobson 2001] – Mobile networks : mobility, power, usage • From a management plane point of view : • Capacity related dynamics – Constrained power environments « A system that changes its state over time in such a – Vertical handoff & Associated QoS variation way that changes cannot be detected or processed • Services & Service Infrastructure level dynamics – Software radio, extensible terminals efficiently by the management plane » – P2P networks & overlays – Because they occur too fast – Active networks : protocols change, functionality extension, … – Service platforms : service life cycle, hot deployment, update, … – Because they occur too often – Service usage : discovery, trading, late binding, lease, subscribtion – Because they occur too slowly • Device, Network & Service Management Dynamics – Because their processing requires changes in the – Online SLA parameters (re)-negociation management plane – Management interfaces tailoring –… © MADYNES 2003 -5- © MADYNES 2003 -6- 1 Collapsing Time scales Management solutions for the dynamic world System • Static Management name status – Manager/MIB/Agent 1. Change to a static world and keep the … – Manual configuration of participating entities – Homogeneous management domains management frameworks as they are ! – Pre-sharing of management information models • 1-6 years to build a management interface – Size : 1 to a few hundred – Time Scale • Accept that you will at some time end in state #2 ! • Few minutes up to a few month • 1-8 years for information models (standards) 2. Consider management as a useless interface and service and stop working on it ! Jain, J2EE, … • Dynamic Management • Start thinking about other ways to QoSing – Infrastructures, services, usage & mobility • Trust webs, in-band probing, signalling, … – Multiple, shared evolving administrative domains – Variable information models and management interfaces 3. Make management as dynamic as the entities – Negociable & evolutive management functions – Size : 1 to several millions/billions it manages ! – Time Scale • Rethinking everything • ms up to a few seconds New management solutions are needed! © MADYNES 2003 -7- © MADYNES 2003 -8- Mgmt & dynamics (1): the static option Mgmt & dynamics (2): the empty chair option • Networks and services do not need management anymore ! • « Many aspects of dynamic systems are often – They adapt themselves to the provided conditions perceived as static » [Jakobson 2001] – 90’s ALF philosophy with an extension to the active networks community • So lets consider them as static • As long as theses dynamic aspects do not alter the external enveloppe. • Keep the management frameworks as they are • But « Nature abhors a vacuum » • Mapping in the networking world – Benedict Spinoza (1632-1677) – Focus on the service or application level only – End2End monitoring & performance measurment – As soon as something becomes too dynamic : ignore It • Thus traditional management will be replaced by smart solutions • History tells us that this is not a good approach – From management to signalling • The case of an evolution : ATM VPC, SVC – Often micro-dynamics highly impact the world – DHCP was an evolution in IP address management • « Butterfly Effect » by E. Lorentz MIT (1961) • but not a sufficient one since DHCP needs to be managed and DNS is • Maybe not the best example … required too for the translation (Dynamic DNS is a partial solution) – In highly dynamic environments, even management of – Zeroconf, zerouter are other « management free » efforts within IETF the static parts becomes almost impossible using the • No management entity needed usual paradigms • Initially focused on IP address (unicast, multicast) and name assignment • Because Unmanaged entities are part of the Chaos ! © MADYNES 2003 -9- © MADYNES 2003 -10- Mgmt & dynamics (3): make management dynamic Standardized approaches support • Divide ut imperes! (Divide & conquer !) • Mostly focused on Delegation & Distribution – Julius Cesar (102-40 BJC) – RMON & RMON2 – Manage smaller worlds to manage each better – SNMP’s INFORM Service • Get compliant with each degree of encountered dynamicity – Build an adequate hierarchy to manage the interactions among the worlds – SNMP DISMAN initiative • This is NOT an easy task! • Script MIB, Event MIB, Schedule MIB, Remote Operations MIB, • Foster Management Autonomy & Automation • Adaptation to emerging constraints – Avoid human interference – Dynamic MIB extensibility support • Self-Anything for the management plane • AgentX & previous efforts like SNMP-DPI or SMUX • E.g. self-configuration of the management plane • CIM Provider in the WBEM Architecture – Avoid time costly operations – Asynchronous messaging • Like long interface standardization efforts • MOM’s gain acceptance as a management communication channel • New dedicated management languages & models learning, •… • Current limits (self-ing) • Equivalent to the consequence of the previous option – Self-configuration • Accept & Integrate new constraints in the management plane • Service, network, device & MANAGEMENT PLANE ! – Low connectivity – Self-provisioning – Shared / cooperative management, new security models, negociation services – Self-instrumentation of services, – Unstable domains & entities – Self monitoring – Evolving and changing management interface – Self-healing – Self-optimizing © MADYNES 2003 -11- © MADYNES 2003 -12- 2 Emerging alternatives Summary • Delegation support and management interactions remain a • Automation is the focus of many efforts Self- Self- Configuring Healing strong focus – Autonomic Computing by IBM – Active networks based management is one approach Self- Self- – N1 for Farmed resources by Sun Optimizing Protecting – Mobile agents are another solution – CISCO CNS Configuration Engine & Intelligence Engine … – Dynamic Managed objects loading & distribution become common – SelfCon initiative (Nortel & Waterloo U.) • Dynamic instrumentation is growing fast – Self-aware networks (Alcatel) – CORBA interceptors –… – Web Services Proxies • Promising future – J2EE management introspection – Several technologies cover parts of the needs –OSGi service interface monitoring • But do not cover all needs at the same degree • Automation is increasing • We will examine 2 different ones with different