Mobile Broadband Working Group Open Internet Advisory Committee

Mobile Broadband Working Group Open Internet Advisory Committee Jennifer Rexford Princeton University http: //www. cs. princeton. edu/~jrex

Mobile Broadband Working Group • Harvey Anderson, Mozilla • Brad Burnham, Union Square Ventures • Alissa Cooper, Center for Democracy & Technology • Jessica Gonzalez, National Hispanic Media Coalition • Charles Kalmanek, AT&T • Matthew Larsen, Vistabeam • Dennis Roberson, IIT and representing T-Mobile • Marcus Weldon, Alcatel-Lucent 2

Mobile Broadband • Mobile broadband – Increasingly crucial part of Internet access – Yet, still at an early stage of development • Open Internet Order – Network practice transparency – Certain “no blocking” requirements – Wider latitude for differentiated service • Working group – Review state of mobile broadband – Assess how Open Internet principles are working 3

Our Initial Approach • Discuss several case studies – Focus on concrete, real-life scenarios – Capture the relevant facts, issues, and viewpoints – Identify trade-offs, principles, and areas needing study – Not articulating specific policy recommendations • Three specific examples – AT&T limiting the Face. Time application – Mobile apps overloading signaling resources – Carriers limiting use of Google Wallet 4

AT&T and Face. Time • Apple Face. Time – High-quality video chat service on i. Phone/i. Pad/Mac – Originally available only over Wi. Fi on i. Phone and i. Pad – Phone call upgraded to Face. Time by tapping a button • Face. Time over 3 G networks – Jun’ 12: Apple announced Face. Time over 3 G in i. OS 6 … though carrier restrictions may apply – Aug’ 12: AT&T limits 3 G Face. Time to Mobile. Share users … with data cap shared across multiple devices • Other mobile providers – Sprint and Verizon confirm that FT works on all plans 5

Arguments in the AT&T/FT Debate • Some advocates and press denounce the decision – AT&T is violating FCC’s Open Internet Order – AT&T is blocking an application competing with its own voice or video telephony services – Reasonable network management practices do not include favoring one data plan over another – Suspicion that Apple is cooperating with the scheme • AT&T responds in a blog posting – AT&T’s policy regarding Face. Time is fully transparent – AT&T does not have a competitive video chat app – FCC rules don’t regulate availability of preloaded apps – All users can continue to run Face. Time over Wi. Fi 6

AT&T/Face. Time Issues • Pre-loaded application, tightly integrated with OS – Available to all users of popular phone w/o downloading – Accessed via the core calling functions of the device • Symmetric bandwidth usage – Access network capacity is asymmetric – Single Face. Time user consumes 33 -50% of sector uplink • Limited adaptation to congestion – Many multimedia apps reduce rate during congestion – … but, FT doesn’t seem to adapt as much as other apps • Staged deployment to understand the effects – Initially limit number of users accessing an app 7

Apps With High Signaling Traffic • Signaling channel – Keeps track of mobile devices and their locations – Notifies network when a device wants to send traffic • Overloaded signaling channel – Prevents new requests from reaching the network – Can become congested before the network bandwidth • Unique issue in cellular networks – Due to Radio Resource Control function – … and the shared, constrained resources 8

Chatty Applications • Periodic transfers – Keep-alive messages (e. g. , push services, NATs) – Polling (i. e. , has something happened? ) – Ad updates – Measuring user behavior • Skewed usage of signaling resources – Up to 30% resource usage for <2% traffic volume – In some cases, 90% signaling usage by one application – Also drains the battery on the phone • Machine-to-machine traffic could make this worse 9

Managing Signaling Load is Hard • End device – Strong incentives, and (some) app-level knowledge – But, incomplete control over application behavior • Application developer – Complete knowledge of own application, but not others – But, limited knowledge of the network state • Network – Sees all traffic and controls resource scheduling – But, incomplete knowledge of applications, or ability to infer app before harm has been caused 10

Signaling Management Challenges • Application-specific management – Adjusting timers for periodic polling – Piggybacking of requests on data traffic – Application-level signaling control • Huge number of applications – Average lifetime of 30 days and revenue of $700 – E. g. , AT&T has worked with ~100 app developers to better optimize their applications • Complex optimizations – Cross-application and cross-device • Joint management of bandwidth and signaling load 11

Going Forward… • Network management is challenging – Large mix of rapidly evolving applications – Growing number of mobile devices – Limited bandwidth and signaling capacity • The technical details matter – Capabilities of today’s equipment – Best-common practices for network management • Perhaps enlist help from other groups – To capture the relevant technical context – E. g. , Broadband Internet Technical Advisory Group • Review ongoing standards work (e. g. , 3 GPP) 12