The Case for Viral Broadband Decentralizing broadband Internet - PowerPoint PPT Presentation

The Case for Viral Broadband Decentralizing broadband Internet access Dimitris Vyzovitis vyzo@media.mit.edu Viral Communications, MIT Media Laboratory Viral Communications Workshop, October 20, 2004

Outline • Background and Context • Broadband Access and Peer-to-Peer Systems • Pricing Broadband • Viral Broadband Dimitris Vyzovitis 2 Viral Communications Workshop, October 20, 2004

Background Past research in Viral Communications Group: decentralize content distribution • Collaborative content distribution protocols multicast, stream aggregation, localized distribution • DiVA: Distributed Video Architecture tuner and disk space sharing Dimitris Vyzovitis 3 Viral Communications Workshop, October 20, 2004

Collaborative Content Distribution Node N 5 Local Network Router L 4 Data Flow R 4 R 2 R 3 L 3 R 1 N 6 N 4 L 1 L 2 N 2 N 3 N 1 Dimitris Vyzovitis 4 Viral Communications Workshop, October 20, 2004

Context Viral Broadband decentralizes broadband access. • Base technology is already there (WiFi) • Grass-roots (Roofnet etc) and city-wide (Philadelphia etc) projects, commercial offers (Nortel etc) with viral characteristics. Impetus • Solve the asymmetry problem. • Enable real-time programming. • Provide better and cheaper service for end users by leveraging wired broadband infrastructure. Dimitris Vyzovitis 5 Viral Communications Workshop, October 20, 2004

Context (contd) From decentralization to innovation: allow ideas to evolve, centralize and legitimize. • Client-server Internet: web, online banking, etc. • Online content distribution: file sharing → iTunes, Raphsody. • Peer-to-peer realtime communication: Skype • Other industries: video, telephony, airline pricing systems, variable tolls, etc. Dimitris Vyzovitis 6 Viral Communications Workshop, October 20, 2004

The State of Broadband Broadband Network A network in which the bandwidth can be divided and shared by multiple simultaneous signals (as for voice or data or video) [Wordnet] State of affairs • Basic Internet service • Asymmetric access • Flat rate pricing What’s coming • Multiple priorities, integrated services • Usage-based pricing Dimitris Vyzovitis 7 Viral Communications Workshop, October 20, 2004

Network Asymmetry Upstream capacity 4-10 times less than downstream . Ok for client- server computing, but not ok for peer-to-peer applications. 500 residential business 450 400 350 Service cost (USD, monthly) 300 250 200 150 100 50 0 0 200 400 600 800 1000 1200 1400 1600 Upstream capacity (Kbps) Upstream capacity cost. Source: sample prices from leading ISPs Dimitris Vyzovitis 8 Viral Communications Workshop, October 20, 2004

Growth of Broadband: Global Exponential increase since 1999 to 70 million worldwide in 2002 Source: ITU Dimitris Vyzovitis 9 Viral Communications Workshop, October 20, 2004

Growth of Broadband: U.S. Market penetration over 50% in September 2004 Source: Bandwidth Report Dimitris Vyzovitis 10 Viral Communications Workshop, October 20, 2004

Peer-to-peer Systems and Broadband Peer-to-peer systems is a killer application for broadband. • 2001: 30 million Napster users, 40 million broadband users worldwide. • Now: peer-to-peer is the single larger consumer of data in ISP’s networks. • Flat rate pricing: no actual cost, users encouraged to file share. • Usage-based pricing: resource sharing is problematic. • Network asymmetry: symmetric high bandwidth applications and real-time programming are problematic. Dimitris Vyzovitis 11 Viral Communications Workshop, October 20, 2004

Peer-to-peer Traffic CacheLogic analysis of tier-1 and tier-2 ISPs traffic: Over 80% in last mile . Source: CacheLogic Dimitris Vyzovitis 12 Viral Communications Workshop, October 20, 2004

Pricing Broadband Layered pricing model (Stiller et al ) : Content Pricing Usage-based Pricing Transport Connection Pricing Related Price Access Pricing Dimitris Vyzovitis 13 Viral Communications Workshop, October 20, 2004

Pricing Broadband: Usage Constraining Pricing Constraining resource usage: bit counting charges with volume caps. � c = c access + ( c service,i + c excess,i · ( b i − U i ) · u ( b i − U i )) + c content i fixed network access cost c access c service,i i th service fixed cost c excess,i i th service excess cost i th service bit count b i U i i th service bit cap u ( · ) unit step function end-to-end content cost c content Dimitris Vyzovitis 14 Viral Communications Workshop, October 20, 2004

Pricing Broadband: Illustration Plot of c for a single service Cost c excess c access c service + U Volume Dimitris Vyzovitis 15 Viral Communications Workshop, October 20, 2004

Viral Broadband Architecture Elements of architecture • Wireless ad-hoc network, static end user nodes provide infrastructure. Extant architecture: wireless mesh networks. • A subset of end users provide Internet access via direct ISP connection. • Peer-to-peer interactions. • Turn WiFi inside-out. Modest extrinsic requirements • Base technology already there: WiFi. • Reasonable ISPs: open network access. • Reasonable regulation: open spectrum. Dimitris Vyzovitis 16 Viral Communications Workshop, October 20, 2004

Wireless Mesh Networks: ISP-centric User Node Transit Access Point T T 1 2 Gateway A T 3 Internet TAP-based mesh (Nortel etc) Dimitris Vyzovitis 17 Viral Communications Workshop, October 20, 2004

Wireless Mesh Networks: Community-based A User Node 1 A 2 A User Gateway Node 3 Gateway Internet MIT Roofnet Dimitris Vyzovitis 18 Viral Communications Workshop, October 20, 2004

Pure Viral Broadband Architecture A User Node A 1 A User Gateway Node 2 A 3 A 4 A 5 Internet Dimitris Vyzovitis 19 Viral Communications Workshop, October 20, 2004

Mixed Viral Broadband Architecture A User Node 1 A 1 User Gateway Node A A 2 6 ISP Transit Access Point A 3 ISP Gateway T A 1 4 A 5 T 2 Internet Dimitris Vyzovitis 20 Viral Communications Workshop, October 20, 2004

Parasitic Peer-to-peer Content Distribution B A User Node User Gateway Node User Content Cache C Internet Dimitris Vyzovitis 21 Viral Communications Workshop, October 20, 2004

Economics of Viral Broadband Observations • Higher capacity for users by aggregation. • “Symmetric”, peer-to-peer access. • Some users incur actual cost . Compensation by cost distribution among users according to fair share of resources! Dimitris Vyzovitis 22 Viral Communications Workshop, October 20, 2004

Economics of Viral Broadband (contd) Amortized access cost � A i ∈ A c access,i c ∗ access = n Per-user viral network overhead ∆ c = c excess · (( b V + b − U ) · u ( b V + b − U ) − ( b − U ) · u ( b − U )) User share b i · � A j ∈ A ∆ c j ∆ c ∗ i = B number of users n subset of users with direct access A b V viral network excess bit count i th user excess bit count b i global viral network excess bit count B Dimitris Vyzovitis 23 Viral Communications Workshop, October 20, 2004

Benefits • End users: symmetric access, better service, better prices. • Base technology providers: economies of scale. • Service providers: accounting, cost distribution. • Content providers: efficient peer-to-peer content distribution. • Entrepreneurs: innovation at the edge, application opportunities. Dimitris Vyzovitis 24 Viral Communications Workshop, October 20, 2004

Challenges • Fair share cost distribution: accounting, payments, efficient implementation. • Routing: multi-gateway. • Transport: multi-path. • Multicast. Dimitris Vyzovitis 25 Viral Communications Workshop, October 20, 2004

More Information and Progress Viral Broadband Project Viral Communications, MIT Media Laboratory http://web.media.mit.edu/ ~ vyzo/vbb Dimitris Vyzovitis 26 Viral Communications Workshop, October 20, 2004