VoCCN: Voice Over Content-Centric Networks Paulo Mendes (paulo.mendes@ulusofona.pt) March 3rd, 2011 IAN meeting, SITI, Lisbon Voice Over Content-Centric Networks Agenda 1. Brief Introduction to Named Data Networking 2. Focus on Content-Centric Networking 3. Study of Voice over Content-centric Networking Van Jacobson, D. Smetters, Nick Briggs, Michael Plass, Paulo Stewart, James Thornton, Rebecca Braynard. “VoCCN: Voice over Content-Centric Networks”, in Proc of ACM Reach, Rome, Italy, Dec 2009 V. Jacobson, D. K. Smetters, J. D. Thornton, M. F. Plass, N. H. Briggs, R. L. Braynard (PARC) Networking Named Content, CoNEXT 2009, Rome, December, 2009. 2 Brief View on Networking Historic Appointment Since the invention of the telephone, communications have been related with a “conversation” between two hosts For the Internet generation, the Web just changed everything And now the Internet of the things is placing an extra effort in host-to-host communications The “circuit” model of communications requires Spanning Trees Introduces global dependencies that remove local choice. Makes load near source scale with popularity. Today’s view of communications Wires move information in space. Memories move information in time. What do we need? An architecture that merges both views of the same problem 3 Name Data Networking General Motivation and Design For sharing, packets must describe what you want, not the process of getting it. An address has to name data, not conversation endpoints. For asynchrony (time-shifting, intermittent connectivity, mobility, ...) Memory has to be explicit in the communication model. Have to stop pretending container-based security can work Start securing data. Communication paradigm based on controlling named-content in the network, instead of host-to-host communication as in today's Internet. Note: Publish-Subscribe paradigm is just a possible way to realize a data-centric network. 4 Data-Named Networks Major Efforts Projects Named Data Networking (UCLA, PARC, University of Arizona, University of California, Irvine, University of California, San Diego, Colorado State University, University of Illinois, Urbana-Champaign, University of Memphis, Washington University, Yale University) MobilityFirst Future Internet Architecture Project (Rutgers, UMASS, U. Michigan, U. Duke, MIT, UNC, U. Winconsin, U. Nebraska) - http://mobilityfirst.winlab.rutgers.edu/ Nebula: A Future Internet Architecture that Supports Trustworthy Coud Computing (Cornell Univ, U. Texas, U. Washington, Purdue Univ., U. Illinois, U.Pennsylvania, MIT) - http://nebula.cis.upenn.edu/ Expressive Internet Architecture (Boston Univ., CMU, U. Wisconsin) - http://www.cs.cmu.edu/~xia/ NetInf (http://www.netinf.org/home/home/) Haggle (U. Cambridge, Thompson, Uppsala U., EPFN, CNR, .. ) - http://code.google.com/p/haggle/ PSIRP – Publish/subscribe Interdomain Routing Paradigm: system based on brokers to match cryptographicbased flat identifiers to application data names (Data must be publish forehand) People: Scott Shenker, Deborah Estrin, Ion Stoica, Lixia Zhang, Van Jacobson, Jim Kurose, Don Towsley, Dipankar Raychaudhuri. Peter Steenkiste 5 Data-Named Networks CCN aims to …. (provably) optimal content distribution Painless mobility Same efficiency as TCP/IP Simple, secure, robust configuration An easy, incremental, evolutionary path Much better security Greater scalability and a more dynamic network topology. Fundamentals: Replace Packets with Data Objects and Interests Replace Addresses with Object Names 6 Data-Named Networks Host vs Data communications Path determined by global routing, not local choice. Structural asymmetry precludes market mechanisms and encourages monopoly formation. Packets say ‘what’ not ‘who’ (no src or dst) Forwarding decision is local Upstream performance is measurable. 7 VoCCN Motivation and Major Goals Fact: Architectures based on data-oriented abstractions provide a good fit to the massive amounts of static content exchanged via the World Wide Web and various P2P overlay networks. Doubts: How well they fit more conversational traffic such as email, e-commerce transactions or VoIP? Goal: To investigate Voice Over CCN in terms of simplicity, security and more scalability. Requires Separate signalling and data paths. Requires regular updates of topologic locations VoIP is secured by: Encrypted form RTP (SRTP) Tunneling RTP with secure protocol (e.g., DTLS). Encryption keys are set up via: encrypted signaling path or in-band in the media path (ZRTP). 8 VoCCN Architecture and Advantages Motivation: VoIP signalling and data paths result from a mismatch between the user’s goal and the network’s means of achieving it. Alice simply wants to talk to Bob but the network requires that the communication be addressed to the IP address of Bob’s phone. Goal: Remove the “translation” from real life to network control. Challenges: Support service rendezvous: request a connection and get a confirmation response. On-demand publishing is needed: request data that was not published yet. Transition phase: from rendezvous to a bi-directional conversation. Constructable names are needed: construct name a priori: Deterministic algorithm to get to the same routable name (providers & consumers) Consumer retrieve content based on partial names. 9 VoCCN Implementation and Security Inline message security: Caller: encrypt and authenticate SIP invite using random-Inline message encriptation generated symmetric key (sk) Components of Interest Caller: encrypt sk using callee’s public key (pk B) Calle: decrypt interest using his/her private key Calle: uses sk to verify and decrypt SIP invite, which includes first MIKEY key exchane message 10 VoCCN (and CCN) Advantages and Major Limitations Proof-of-concept of real-time CCN Extension of Linux VoIP phone (libeXoSIP, liboRTP) Open CCN toolkit (CCNx): routers run on endpoints. Support multi-point routing (mobility and multi-homing) “Identity” of an endpoint represented by credentials Easy to build advanced services (e.g. conference calls) Performance for a 10 minutes conversation: More packets below the expected inter-packet interval Small number of long-interval packets. No packets loss for both cases Constructable names: flexibility may require partial names that are not unique Pipeline of Interest packets: not proved to perform Working scenarios: The use of hierarchical names poses difficulties in dynamic graphs Performance studies based on fast workstations and high-speed networks (100 Mbps and 1Gbps) 11 12