networks research Bob Briscoe & Ben Bappu Dec 2004 networks - PDF document

networks research Bob Briscoe & Ben Bappu Dec 2004 networks research centre - themes • networks of x • x: computers, devices, people, physical phenomena • new science • framework for reasoning, proof, architecture, language, implementation • necessity of relaying infrastructure? • design for tussle: end to end design then modify to edge to edge commercial • commercial service vs. collaborative self-supply viability • incentives secure • for collaborative self-supply responsibility scalability • for commercial service (viability) simple • social costs (e.g. congestion, loss-of-privacy) freedom evolvable • regulation & public policy • collaborative vs. polity

computational networking process model • Can implement typical network processes routing ⊕ congestion pricing = congestion routing – • Provides an alternative view on (distributed) computing – convergence vs. transition distributed information flow – interaction vs. input/output ? • Connects with category (graph) process calculi theory rewrite – Economics, game theory ambients systems π π π -calculus π – Constraint and functional programming, agents objects message passing Research! grid graph Turing m/c λ λ λ -calculus λ – Dynamic systems, control theory, networking computing optimisation (information dependence) (routing + flow) – Electronic circuits, thermodynamics • Offers a framework within which to address various research questions within networking – layer interaction, feedback, time scales, network economics Internet naming, addressing and routing Aims • Now major commercial environment upon which many businesses depend • An unforeseen role that has forced the Internet to increasingly face evolutionary challenges • The Internet architecture • The set of fundamental design principles that give direction to the evolution of the Internet • Mostly irrelevant and hence ignored by the communications industry • Many Internet extensions have violated the original architecture • Thus reduced the evolvability of the Internet • Today, many scientists believe that it is time to re-think the Internet architecture to: • Absorb within a new architecture the chaotic evolutionary patches that have modified today’s Internet • Make it relevant to the industry by introducing the missing sensitivity to commercial viability of its extensions Our ultimate aim is to re-think, within today’s economic and societal context, the fundamental aspects of the Internet architecture that relate to addressing, naming and routing. Research focus Previous results showed that the Internet addressing, naming and routing architectures have been violated in different points and by different technological extensions. These extensions (shown in green arrows) represent point solutions that the communications industry has issues with, created by the new business-oriented role of the Internet, that the Internet architecture did not address.

Internet naming, addressing and routing Achievements – a meta-architecture for Research approach Internet identification To analyse the existing extensions (as point solutions) and draw from these the design A meta-architecture for Internet identification, principles that, once generalised, are captured which defines a logical abstraction of the by the overall addressing, naming and routing Internet, creating a class of architectures architectures in order to naturally cope with which capture a number of desirable these extensions. properties of identification systems. It is used as a design tool to capture existing and new design principles for Internet identification by (1) showing that two properties are contradictory or cannot be both achieved, and (2) suggesting design rules or mechanisms that satisfy some property or set of properties. Motivational Issues in Peer-to-Peer Edge computers (Peers) Minimal use of central servers Communicating directly Sharing information or system resources or (more interesting!) peoples’ time and effort.

Economic Architecture In a community context Other Peers Consumer Provider Make the social control Social Control explicit Aggregated Information Accounting and Distribution Tokens Peer Control Peer Control Underlying atoms token Service Usage of consumption and contribution service www.mmapps.org MASCAPONE MA ximising S pectrum CAP acity O ver mobile N etworks Simple propagation models understate gain AND overstate interference in a cooperative network What is the transport capacity? Source Useful Target relay Less interference here • How will next generation MAC layers perform in ad-hoc scenarios? • What are the metrics for ad-hoc hybrid networking? • What is the real value of co-operative MIMO approaches? • MIT’s RF relaying – is it really practical? • How much cross layer interaction do we need?

shared control; shared value 1) downstream knowledge upstream (re- feedback) propagation time congestion hop count etc 0 R 1 13 -7 7 7 0 8 control 8 3 & info -5 2 -1 S 1 9 control control 0 -2 & info & info R 2 control S 2 control control & info & info & info shared control; shared value 2) contractual mobility • User-selected route vs. user-selected edge-provider – Can the user selection of the edge provider (e.g. access provider) boost competition of the wide area market? – Or does the Internet need a ‘carrier pre-selection’ type of functionality (wide area providers selected by users)? e2 user user

contractual mobility - prototype provider1 provider2 provider3 provider1 provider2 Offer Offer Offer Offer Offer disseminator disseminator disseminator disseminator disseminator 1 1 4 selector roamer 3 2 Offer repository handler Shared control; shared value 3) Deep Packet Inspection (DPI) Project Inspect and analyse data packets beyond the IP headers and deeper into the payload. Research Motivation • Analyse potential applications of DPI for BT • Propose an architectural framework for DPI applications • Propose a control structure for managing these DPI applications

DPI Applications • Security - Firewalls, Intrusion Detection & Intrusion Prevention BT’s Network Services Customer • Policing - Monitoring traffic to enforce policies & contractual SLAs BT BT p2p traffic p2p traffic ISP traffic ISP traffic VoIP traffic VoIP traffic DPI Applications • Service Operations - Shaping, Capping, Load-balancing, Charging, QoS VoIP DoS VoIP http Public BT’s DPI http Network Network p2p p2p • Internal Operations - Revenue assurance, Quality assessment, Information gathering, SLA monitoring • Legal Requirements - Legal Intercept and Call records

IP End to End Quality of Service App App Core IP QoS L3 L3 L2 L2 L2 L2 L2 L2 L2 L2 L2 L1 L1 L1 L1 L1 L1 L1 R R R R R R R GQS Access End Home Core Core Core Access Home End End to End QoS Coordination GQS region - DSCP aware, strict priority queuing for guaranteed data - ECN marking - No per flow state WLAN – 802.11e Congestion Collapse & DCCP 2 QoS Modes: - Priority (EDCA) – simple but non efficient for small packets - Polling (HCCA) – complex and still Congestion collapse as non effected by interfance rate adaptive UDP traffic Little info on how to set parameters or increases integration into e2e QoS system Cambridge Uni BT Research Microsoft Research DIY QoS target rate a inelastic a (stream a media) a a a (shadow) price a target rate congestion marking a = (shadow) price target rate ultra-elastic max (p2p) ave. TCP util/ (shadow) price % 100 (shadow) price www.m3i.org

guaranteed QoS synthesis IP routers Data path processing Reservation 1 Reserved flow processing enabled � intensive BT research 2 Policing flow entry to CP RSVP/ECN � potential v. low cost PSTN replacement gateway 4 Meter congestion per peer � identical guarantees, even televotes ECN only � core: classic Internet with minor mods 3 Bulk ECN marking CP prioritised over BE � neither overprovisioning nor QoS smarts reservation signalling 1 2 congestion 3 3 pricing congestion pricing 3 3 4 1 congestion pricing guaranteed best effort e2e QoS signalling co-ordination Applicati Applicati on on SS SS SS SS HG IR IR IR HG IR IR R R R R R R R R R R R R R R Access Core Core Home Core Home Access