TOWARD A POLYMORPHIC FUTURE INTERNET: AN INTERNET SCIENCE APPROACH Kavé Salamatian, LISTIC, Université de Savoie Once upon a time, …, in the beginning And now ! Happy birthday !!!! Three anniversaries in 2009 20 years of WWW 30 years of USENET 40 years of Internet Internet: a global media The “Hourglass” as the model From clear layers to spaghetti … ICA/Citrix L2TP KaZaa DNS ICMP IRC UDP TCP IPv4 PPP PPPoE Ethernet XMLRPC HTTP SMTP ICMPv6 IPv6 802.1q SOAP The Real Internet looks like 8 The R&D « Learning » Curve Solution complexity Solutions proposed Solutions in use Understanding of problem area early middle late New Cycle Industry transfer time 9 Solution complexity New challenges in Networking Opposite forces Applications Regulation Economy Management Complexity Are We’re here or there or both! early middle late time 10 Nowadays … Pocket switched networks I have 100M bytes of data, who can carry for me? Give it to me, I have 1G bytes phone flash. Thank you but you are in the opposite direction! Don’t give to me! I am running out of storage. I can also carry for you! Reach an access point. There is one in mySearch La pocket… Bonheme.mp3 Internet for me Finally, it arrive… Search La Bonheme.mp3 for me Search La Bonheme.mp3 for me Go back to basics A network is build of components Local in a node or distributed Sitting in one layer or crossing layers Tightly or lightly coupled Cooperating to transmit information from point to point. Actual layered architecture is just one specific type of collaboration Autonomous networking idea Collaborating through protocols Moving from pile view to puzzle view Need for new theoretical framework Cooperation ? Full Cooperation Do the best possible behavior to reach a goal Assuming full collaboration Is the goal achievable ? How to achieve the goal ? Multi-user information Theory Non–cooperative Selfish behavior Different rational goal How to mitigate conflicting rational goal ? Game theory Malicious behavior Harmful goal How to contain irrational objectives ? Behavioral inference Cooperation framework • Each Node implement a forwarding function • The forwarding function implement the cooperation (Y t' 1 , Y2t ' ,… , YNt ' )= f i (X 10:t , X 20:t ,… , X N0:t ) M it X 1t X f i( t 2 X Nt ) Y1t ' , Y2t ' ,… YMt ' Forwarding function Examples Flooding Y jt +e = X it "i, j, t Routing Y t +e j ì X it =í îf t i if cond ( X ) else Distributed computation Yot + kT = f (Xit , Xit +T ,… , Xit + kT ) Network coding Any other ? Y t +T j = åa j,t j <t +T tj j Xi CLEAN SLATE VS EVOLUTIONARY? Exploring the future Internet What is broken and need to be fixed? What is the right approach? Is the clean slate approach appropriate? Any transition methodology? How difficult do you expect that a new architecture would be adopted? We have a few examples on how difficult it is to change the current architecture: multicast, mobile IP, Intserv/Diffserv, and IPv6. Any enabling factor? Explore the possible Future(s) of the Internet Realistic view Continuous evolution and change Usage and Services will become predominant User-centric approach to system design On-line social networks Greening the Internet Other factors than technology will be instrumental Economics, Social behaviors, Entry cost, Regulation A single architecture ? The future Internet might be Polymorphic Various research projects, scientists and “people” will propose new ideas GENI, FIND, FIRE, etc… Building blocks Architectures Flexibility is the major constraint of future Internet Flexibility or the internet contortionist to accommodate Future Internet polymorphism Enable different cooperation paradigms in parallel. Enable easy deployment of new application deployment. Without raising routing and addressing to the application As Peer to Peer and overlay networks A network where one could implement, and deploy its new network protocols or cooperation schemes without disturbing other running protocols Virtualization, the enabler Properties Isolation Encapsulation Fault, software and performance isolation All VM state can be captured into a file Interposition` All guest actions go through monitor Monitor can inspect, modify, deny operations PlanetLab, OneLab, XORP, etc. Security, Internet Achille Heel’s What should be in future Internet security strong encryption and authentification ? VPN’s Proxies ? Secure routing ? Future Internet minimal security A secure execution environment A code should be executed the way it is designed and not differently An authentication mechanism Local of global authentication ? Monitoring Interposition layer ?` Scalability or the delusion of grandeur Active BGP entries (FIB) All BGP entries (RIB) 349893 10775196 Compact space routing y d-dimensional space with n zones (x,y) Peer Q(x,y) Query/ Resource 2 zones are neighbor if d-1 dim overlap Routing path of length: Algorithm: Choose the neighbor nearest to the destination Fundamental Q(x,y) key question Is it possible to design a compact space routing for any application ? Enabling continuous deployment of New architectures Policies Monitoring First applied to Test-beds federation Security Federation IP world Virtualized substrate ••• 26 Vision – The « Facility » framework 27 Experimentally driven research Why? Research driven by Intuition & Experimentation + validation A Playground for enabling new architectures Shorter time to System/Market Complementary to Network Science What? …Enforce a Federation effort on building a large-scale facility for “Internet” research Building, Operating, Maintaining a testbed is costly Diversity and scale Lower the entry cost for experimentation 28 Conclusions Two main positions Future Internet should be polymorphic Enable the coexistence of different networking paradigms in the same framework Virtualisation is a key enabler Future To Internet needs a networking science have strong foundations rooted in Physics, information theory, economics, sociology, ethnology, etc… Cooperation is a strong concept Conclusions Issues To develop this networking science Flexibility Virtualization is key We are moving toward a “no architecture” Security Sandbox, authentification and monitoring should be integrated to architecture Scalability Is it possible to construct infinitely scalable addressing and routing schemes Clean Slate approach should enable the coexistence of different addressing/routing schemes.