Research is communication by Smjt2G


									Endless Arguments in Systems Design

      Jon Crowcroft, University of
      Currently IMDEA Networks.
     GENI – Global Environment for
     Network Innovation

   US NSF/Extra Money
   Calls for Intl participation
   Planetlab++ (Intel/300 site)
     FIND – Future Internet Design

   NSF long term research in NeTS
   40+ Projects
   Quite mixed bag
   [EU Trilogy project (which UC3M and
    UCL and others are in) did analysis of
               email WWW phone...
The Internet
               SMTP HTTP RTP...
                   TCP UDP…

(Deering)               IP

                 ethernet PPP…

               CSMA async sonet...

               copper fiber radio...
Putting   email WWW phone...

on        SMTP HTTP RTP...

Weight        TCP UDP…

              IP + mcast

               + QoS +...

            ethernet PPP…
                                  • requires more
          CSMA async sonet...       functionality
          copper fiber radio...     from underlying
Mid-Life   email WWW phone...

Crisis     SMTP HTTP RTP...

               TCP UDP…
                                   • doubles number
               IP4      IP6          of service
             ethernet PPP…         • requires changes
           CSMA async sonet...       above & below

           copper fiber radio...   • major interoper-
                                     ability issues
        Or is GENI losing its bottle?

   Network Innovation is a Hot Topic
       Internet problems abound
       2/3G problems abound
       Convergence driving incremental
       Critical Infrastrure driving radical
       Many US GENI/FIND +EU +Asia projects
       There appear to be some ok ideas
       But also some wrong headedness
   This is a case study on wrong headedness
     Incremental from Convergence

   SCTP (PSTN signaling transport on IP)
   SIP&NAT traversal&VOIP in general
   IPTV (Multicast & P2P)
   BGP fast Convergence
   Traffic Engineering (whether OSPF TE
    or RSVP or MPLS based)
   Neutrality fights
        Long Term
   IP Service Best Effort
   So DDoS not meaningul
   Not helpful – need architectural change
       Fundamentally, Mobile, Multicast, Multihome
       Possibly Multiprotocol (esp. Control Plane)
       DoS Resistence, built in
       Possibly different economic/charging models
   Need shift in model
   Maybe not a stack anymore
   Certainly not IP ++ ++ ++
       i.e. not just IPv6 with current IPv4 knobs and warts
     Two wrongs that don’t make a right(#)

   Virtualisation of Networks
   Next Generation Wireless
   Data Driven Networks

(#) Two parts of FIND/GENI work in US
that I feel they are doing wrong and we
  are doing right + one thing I think the
  US and EU are both doing right.
        Veni, Vidi, Vici, but not Vini?

   Virtualisation is not about VPNs(*)
       Hierarchical Hard Multiplexing won’t scale
       VPNs are an ok idea in the Enterprise
       Its too slow to reconfigure on outages
       Its bad for resource allocation in wireless
       C.f. hard spectrum allocation
       C.f. hard channel allocation

    * Not just VPI/VCI (for those who recall B-ISDN!)
        Virtualisation of Control not Data

   In Xen/Xorp work:
        key is about dynamics and isolation of
        functionality, not fwding performance
       Need to deploy disjoint service ctl
   True, vpn style IP fwding is useful too
       But no unified way of doing this on wireless
       Or on multipath
       Or for netcoded data
       Or swarms…
        Next Gen Wireless

   Cooperative Diversity in all things
       Antennae, coding, routing, power, spectrum,
       Behavioural constraints enforced
       (aside: F# smart phone:-)
   Infrastructure still there for services, e.g.
       To provide radio map (with memory)
       Security/Identity/Location Assistence
       Maybe even payment
   Control plane for virtual wireless communities
        Data Driven, Declarative,
   We’ve seen 1-1, 1-n, n-1 comms patterns
   Now we’re seeing 0-n and n-0
   US loves to re-invent(+) - but in fact
   Pub/sub;event/notify isn’t all there is:
       Think swarm
       Think anysource multicast
 Exaflood is no big deal when
 Think exploding google (aside: haggle:)

(+) architecture is not just creative children’s
    Virtual, Wireless, Declarative Data?

   We’d all like to get rid of wires
        (on the other hand, we’d probably all like to
         get rid of batteries too:)
   So high capacity applications should be
    user contributed content too
   What is the control plane for these to
    co-exist with each other?
   C.f. nano-datacenters and haggle,
        0wning half a PC (or hub or STB)

   Significant fraction of internet access
    via Home Hubs
   Significant fraction of wireless acces
    via smart phones
   Home hub as PVR, smart phone as PC
   Lots of storage + faces 3 ways
       Faces “wide area”; faces home
       Faces neighbours… … …
       Changes economic relationships completely
        Provider coexist w/ virtual

   If provider wants to scale net,
       Run P2P for CDN on home hub
       And do end run on data centers
       Let home user also run P2P CDN
       Let wireless user also run community and ad hoc
   Virtualise STB + Handsets
   With suitable declarative(pub) Ctl Plane
       Paradigm shift now enabled
       Revenue model from services not bitpipes
        Way forward?

   Devil is in detail
       Remove separation of end + intermediate
       All systems may have storage
       Must restore flow, congestion and
        reliability control functions to “network”
       But do not need them “point-to-point”
            So they are going to look very different
       Everyone is a DTN custodian
            & there is no “network”
   GENI is ok for next generation core IP
   Need more fundamental changes if we
    address high speed wireless everywhere
   Especially since application paradigm already
   Europe well placed to contribute:
       both wireless and application architecture,
       which in turn contribute to immediate and long
        term network architecture better than
        incremental core work….
       not that that is a bad thing either!

To top