Rethinking the Internet and the telecommunications market today Why the Internet just works � Why the Internet just works � Departure from the e2e � M. Handley, BT Technology journal, 2006 � Network neutrality � Trilogy consortium � Designing for tussle � The market today The history of changes The failures No significant change since 1993 in core protocols � TCP separated from IP (1983) � MPLS and VPNs works around the limitations of these protocols, � � DNS (1982) below IP layer � Congestion control in TCP (mid 80s) Several attempts to change core protocols failed materialize: � � BGP (policy routing) 1990s � ECN standardized but not widely deployed � CIDR (1993) � Integrated services � Differentiated services (used intra-ISP) The growth of the routing tables � Mobile IP � IP multicast (no end-to-end service) Why? Not immediately pressing problems, enhancements instead of � fixes No money making by ISPs: changes in the core need � interoperability, but interoperability creates no differentiation from competitors. Costas Courcoubetis The Future of the Internet - 3 Costas Courcoubetis The Future of the Internet - 4
The future Architectural problems Short term problems � Key issue: decrease of transparency end-to-end � � SPAM (SPIT) � Security concerns -> firewalls everywhere! � Security � NAT: modify e2e flows, not explicit part of Internet � Denial-of-service attacks architecture. Basic reason: ISPs charge for IP � Application deployment addresses, differentiate between home and biz users Medium term problems � � Hard to develop new applications and transport � Congestion control protocols! Vicious circle! – 1Gbps link, 150ms RTT: 20’ for filling the pipe! Fairness… � No substantial change at layers 3 (10y) ,4 (20y) � BGP: slow to converge, error prone, easy to misconfigure, difficult to debug, no privacy. It mostly works… � Ossification?? � Mobility � Convergence? – DHCP provides a solution � Is the internet up the challenge of video? – Solve it at layer 4 or above instead of layer 3? Costas Courcoubetis The Future of the Internet - 5 Costas Courcoubetis The Future of the Internet - 6 Architectural problems (cont.) Some of the new challenges � Medium term (cont) � inter-domain routing & addressing: � Multi-homing � multi-homing by most businesses and residential users – Robustness vs cost of routing tables � for their own resilience (otherwise when access fails, � Fast path through router forwarding engines: packets they find they can’t work, play, communicate, etc) with IP options are forwarded slower � different access providers specialise in different � IPv4 lost its use of extension mechanisms! applications � Long term � businesses and residential users switch connectivity � Address space depletion regularly between providers � IPv6: no economic incentive to deploy � networks frequently re-route subsets of routes to the � Mobile devices: create walled gardens instead of cheapest neighbours offering e2e connectivity Costas Courcoubetis The Future of the Internet - 7 Costas Courcoubetis The Future of the Internet - 8
Challenges (cont.) Challenges (cont.) � multicast routing � Mobility � currently globalisation is shrinking the distance between � combining the Internet of things with mobility (e.g. humans, but this will extend to things... sensing devices in cars) � the Internet of things will create a more virtualised world, � frequent re-connection to new providers where physical things can be present virtually in other � frequent creation of VPNs parts of the planet � continual re-location of large numbers of devices relative � layered multicast for congestion control of multimedia to network topology distribution � instantaneous authentication, accounting (AAA) & � huge number of disjoint routing groups within core of dynamic config network � why does it take so long to set up a VPN tunnel? or to � huge churn in subscription to multicast routing groups complete connectivity negotiation following powerup? Costas Courcoubetis The Future of the Internet - 9 Costas Courcoubetis The Future of the Internet - 10 Challenges (cont.) Challenges (cont.) � disasters & unrest � Traffic management ‘click and it’s there’ video � major flooding (e.g. global warming) � channel hopping (assuming not all channels served from the same � � frequent power outages cache) and generally many more much larger flows starting & � traffic surges due to mobile comms during major civil stopping (youTube, video calling, 3-D remote presence) unrest (both co-ordinating the unrest, co-ordinating the Internet of things � between innocent bystanders and coordinating the time-division multiplexed dedicated use of high capacity link � official response) technologies, co-ordinated end-to-end � survival of remaining infrastructure, perhaps with affected large amounts of sudden increases of traffic along different paths � machines coming back on and off line frequently a traffic mix dominated by huge numbers of short transfers (eg. � � protection of inelastic flows in progress and prioritisation single packet events) of authority-authority and authority-people generally much higher flow data rates � communications Costas Courcoubetis The Future of the Internet - 11 Costas Courcoubetis The Future of the Internet - 12
Challenges (cont.) Challenges (cont.) � co-ordinated attacks � disconnectivity zero day worm infects large proportion of devices (TVs, PCs, cell- � � resilience during extended periods of disconnectedness phones, PDAs, iPods) � opportunistic connectivity large proportion of routers infected � � fast exploitation of capacity immediately when available infrastructural services (DNS, certification, search, location etc) � infected � graceful cache/cookie expiry access network has to cut off nearly all hosts; how do they get online � to be cleaned healthy routers have to detect who to isolate and simultaneously � prioritise remaining service perhaps combined with attacks on caching infrastructure intended to � insulate against loss of connectivity to services Costas Courcoubetis The Future of the Internet - 13 Costas Courcoubetis The Future of the Internet - 14 Perspective E2e principle � Internet’s strength: general-purpose net � Application-level functions should not be implemented in the core protocols � Provides 80% of capability for 20% of the cost � Implications: � New challenge: how to add an extra 10% of functionality? � Complexity of core network reduced � No good track record! � Easy to add new apps, no dependence on network � It only just works! � Applications on the net instead than in the net Costas Courcoubetis The Future of the Internet - 15 Costas Courcoubetis The Future of the Internet - 16
Moving away from e2e How will the Internet evolve? New Internet requirements not compatible with e2e � � ISPs control the evolution, have their own incentives to Operation in an untrustworthy world � place new mechanisms, invest, control customer traffic � make the network more trustworthy instead of the edges -> add more complexity in the core � Global connectivity with local trust? Where will be the More demanding apps for QoS � new mediation mechanisms? � Current solution: intermediate storage sites -> apps depend on � Ossification? A new network will eventually evolve?? intermediate 2-stage delivery ISP service differentiation � � Promote enhanced service models within ISP boundary, weaken incentives for open e2e services (inter ISP) � Value-based pricing The rise of 3d party involvement � Less sophisticated users � � More functionality in the network, becomes part of the execution model Innovation by larger players backed by more money? � Costas Courcoubetis The Future of the Internet - 17 Costas Courcoubetis The Future of the Internet - 18 Network neutrality The big debate: Net neutrality We don't want to create a world like television where in order to get � your voice heard, you really need to have the backing of a major corporate network When companies have both strong business incentives and the � technical ability to interfere with Web content, it doesn't take a crystal ball to predict that they will do so the government must ensure that [broadband] operators do not � encumber relationships between their customers and destinations on the network many corporations are "speaking out of both sides of their mouths," � claiming to support consumer choice, but promoting a deregulatory agenda that would clear the way for more consolidation of ISPs, less competition and fewer choices Competitive broadband distribution would allow us to rely upon � market forces, rather than government regulation, to govern market structure and service provision (!!) Costas Courcoubetis The Future of the Internet - 19 Costas Courcoubetis The Future of the Internet - 20
Recommend
More recommend
