POMI Stanford University 2020 Programmable Open Mobile Internet - PowerPoint PPT Presentation

POMI Stanford University 2020 Programmable Open Mobile Internet POMI 2020 pomi.stanford.edu Expeditions in Computing PI Meeting May 14-16, 2013 PI Nick McKeown nickm@stanford.edu Presenters Monica Lam lam@cs.stanford.edu Guru Parulkar parulkar@stanford.edu

Trends of 2007-08 2

Emergence of yet another computing paradigm Closed proprietary solutions taking over computing? 3

Our Focus: Enable Innovation with Openness and Choice Openness Mobile Handheld Computing/Data Choice & Substrate Competition Choice of Wireless Nets Innovation Internet Infrastructure 4

Example: Internet Infrastructure Hundreds of protocols Specialized Features 6,500 RFCs Specialized Control Tens of Millions of lines of code Plane Closed, proprietary, outdated Billions of gates Specialized Packet Power hungry and bloated Forwarding Hardware Vertically integrated, complex, closed, proprietary Not good for network owners and users

Support Virtual Infrastructure on Demand? VM VM VM VM OS OS Forwarding Forwarding OS Forwarding Everyone needs their own infrastructure on demand How to support dynamic virtual infrastructure on physical networks that are not programmable

Software Defined Network (SDN): Fundamental Elements Single control plane controls several forwarding devices Control plane physically separate from data plane Packet Forwarding Packet Forwarding Packet Forwarding Packet Forwarding Packet Forwarding

SDN with Virtualization Control Control Control Programs Programs Programs Abstract Network View Global Network View Packet Forwarding Packet Forwarding Packet Forwarding Packet Forwarding Packet Forwarding

SDN Abstractions Benefits: Network Trouble Shooting Control Control Control Programs Programs Programs (3) Network Debugger (1) Static Checking Abstract Network View Finding bugs, and Is the “ policy ” Network Virtualization their root cause, correctly compiled in an operational to the forwarding Global Network View network rules? Network OS 1. <Match, Action> (2) Automatic Checking 2. <Match, Action> Is the datapath working 3. <Port == 22, Drop> 1. <Match, Action> 2. <Match, Action> as specified? 1. <Match, Action> 3. <Match, Action> 1. <Match, Action> 2. <Match, Action> 4. <Match, Action> 4. <Match, Action> 2. <Match, Action> 3. <Match, Action> 5. <Match, Action> 3. <Match, Action> Packet 5. <Match, Action> 4. <Match, Action> 6. … 4. <Match, Action> 6. <Match, Action> 5. <Match, Action> Forwarding 5. <Match, Action> Packet 6. … 6. … Forwarding Packet Packet 1. <Match, Action> 2. <Match, Action> Forwarding Forwarding 3. <Match, Action> 4. <Match, Action> Packet 5. <Match, Action> 9 6. … Forwarding

Research Agenda Abstraction Logical Crossbar Virtual Network OF Switch Network Map OF OpenRadio Switch OF Switch OF Flow Space Switch Systems Scalability Open Network Virtue STS Optimized Hassel Debuggability NetSight Radio OS VM OF Switch (SDN Troubleshooting) Reliability Placement Capabilities

SDN Deployment Nation-wide SDN Infrastructure in GENI (2010) Part of NSF ’ s GENI

Products Announced/Available Many Start-Ups Most active area for VC investment!

SDN Early Adopters

Full Duplex Radios (Sigcomm 13, Mobicom 12,11) “ It is generally not possible for radios to T receive and transmit on the same frequency band because of the RF Cancellation Circuit interference that results. ” d 1 R+aT - Andrea Goldsmith, “ Wireless Communications, ” Cambridge Σ Σ fixed Press, 2005. delays R+iT d N We have invented In-Band Full Duplex control algorithm Radios • Allows radios to simultaneously TX/RX by cancelling TX RF RX RF Frontend Frontend self interference Digital Cancellation • 110dB of cancellation over 40MHz BW at 2.4GHz, Eliminates 2 nd + Order Σ 22dbM TX power Non-Linearities (Intermod Products, Phase Distortion, …) R Phil Levis, Sachin Katti, and their students

POMI Research Agenda Infrastructure Applications Handheld UI Data & Computing Substrate Economics Secure mobile browser Cinder: Energy aware, secure OS Network Substrate HW Platform Radio technology

The Big Picture of Social Proprietary Proprietary Proprietary Social App Platform Messaging Network

Mobile Can Disrupt Computer Pendulum MainFrame PCs Cloud Phone/Em Friends in a ail Party Mobile More available! Largest ego net! More cycles!

Open Social Mobile (OSM) Messaging-Based Network Dropbox LinkedIn FB messaging messaging skydrive gdrive OSM delivers messages based on user identities. Leverages personal phones & abundant cloud providers. (Musubi version: end-to-end encryption)

The Big Picture Proprietary Proprietary Proprietary Social App Platform Messaging Network Open Open Open Social App Platform Messaging Network PRIVACY MONOPOLY SCALABILITY But kids don ’ t care.

shallow, narcissistic genuine collaborative sharing Contextual sharing extensible with 3rd party apps Playing a multi-party game Identity-firewall: On-the-spot sharing before the other has even Sharing without disclosing downloaded the game. identities to 3rd party app.

1. proprietary Open Social Mobile MUSUBI (OSM) 2nd best student paper WWW, 2012 Open-source MIGO High-school research program Video competition to break 2plus nerdy CS stereotype DISPATCH Startup : Citizen journalism MobiSocial Inc. protection Available in Joint research with iTunes store Columbia University

2. centralized, Gatherings faceless global local social networks enhanced with programmable Wedding community displays Individual Voice Community Building Stanford Dining Hall 4 Seasons Hotel, San Francisco DIY Social Networks for Kids

3. big data for marketeers MUSE: for personal use Memories USing Email Gamification of Experience-Infused memory exercises software Analysis of email archives Physicians, browser, Libraries : psychologists search-engine Stanford, (Alzheimer ’ s) Smithsonian, NYPL, Columbia, Oxford

Conclusions ✦ Software defined networks ✦ Distributed social networks ✦ Commercial and social impact