(FC5260) Instructor: Venkat Padmanabhan 1 What this course is and - PowerPoint PPT Presentation

Special Topics in Mobile Systems (FC5260) Instructor: Venkat Padmanabhan 1

What this course is and is not • Learn about the frontiers of mobile systems research • Experience research yourself by way of a mini project • It is NOT about getting a high grade or boosting your GPA! 2

Course Evaluation • Class participation: 15% • Written paper critiques: 15% • Project proposal: 15% • Project checkpoint: 15% • Final project presentation: 20% • Final project report: 20% 3

Course Schedule • Monday, Aug 19: Lecture 1 + project discussion • Saturday, Aug 31: Lecture 2 + project proposals due • TBD: Lecture 3 • TBD: Lecture 4 • TBD: Lecture 5 + project checkpoint • TBD: Lecture 6 • TBD: Lecture 7 • TBD: Lecture 8 • Saturday, Nov 30: Final project presentation & report 4

Mobile Systems 5

What Defines a Mobile System • Mobile • Lightweight • Battery-powered • Wirelessly-connected • Always-on • Personalized • Sensor-rich 6

Issues & Challenges • Mobile • Energy efficiency • Lightweight • Adaptive communication • Battery-powered • Context • Wirelessly-connected • Location • Always-on • Privacy • Personalized • Cloud augmentation • Sensor-rich 7

General Strategies • Optimize the common case (e.g., connected standby) • Energy proportionality (e.g., DVFS) • Scheduled operation (e.g., WiFi PSM) • Complementarity (e.g., WiFi vs. cellular) • Triggered operation (e.g., location change) • Offload (e.g., cloud offload) • Crowdsourcing (e.g., maps) 8

Thought Experiment #1 • Design a mobile social network service with proximity-based alerts 9

Thought Experiment #2 • Design an energy-efficient background transfer service 10

Mini-Research Project • Project on some mobile systems problem • Defining the problem is key! • Would involve experimentation and implementation • Documenting and presenting the work is as important as doing it 11

Some Project Suggestions 1. Stabilizing mobile display for rough roads 2. Energy-optimized display based on gaze tracking 3. Mobile-based tracking of TV ad context 4. Personalized voice commands 5. Robust mobile transactions with intermittent connectivity 6. Driver scorecard 7. WiFi privacy leak detection and mitigation 8. User tracking using weighing scales 9. Spatial meeting recorder 12

#1: Stabilizing Mobile Display • Problem: reading for long or viewing video while on a rough road can be tiring • Hypothesis: tiredness results from the relative movement between the displayed content and the viewer’s eyes, requiring the eyes to constantly adjust • Idea: – use accelerometer and front-facing camera to estimate the relative movement – adjust rendering of content to eliminate it • Research question: – effectiveness – energy efficiency 13

#2: Energy-Optimized Display • Problem: display consumes a lot of energy even when the user is only focusing on a part of it or not looking at it at all • Approach: – dim display or darken the pixels based on what the user is looking at or interacting • Research questions: – energy savings 14

#3: TV Ad Context • Problem: viewing of content, including ads, on TV is disconnected from online experience • Goal: bridge the two – allow user to bookmark TV ads – profile user based on TV ads viewed • Approach: – use smartphone to listen in on TV ad – match against database of ad fingerprints • Research questions: – speed & accuracy – energy efficiency 15

#4: Personalized Voice Commands • Problem: voice-based interaction is prone to failure because of accent and language • Goal: enable personalized voice commands for arbitrary actions • Approach: – record macro for specified user actions – associate voice command with macro • Research question: – flexibility and customization 16

#5: Robust Transactions with Intermittent Connectivity • Problem: intermittent connectivity (e.g., for a mobile cellular client) causes failure of multi- step network transactions • Idea: mask network failures from transaction • Approach: client-side and server-side proxies to mask failures and keep connections alive • Research question: – effectiveness 17

#6: Driver Scorecard • Problem: bad driving abounds --- speeding, excessive honking, violation of traffic rules • Idea: used smartphone to create a driving scorecard • Approach: – use sensors to measure different facets of driving – crowdsource measurements to detect anomalies – create a competitive game • Research questions: – incentives – accuracy – energy efficiency 18

#7: WiFi Privacy Leakage • Problem: WiFi transmisions leak private info by allowing a user to be tracked • Idea: obfuscate the WiFi MAC address • Approach: – set a random MAC address for probe requests – switch between MAC addresses in an unlinkable way • Research question: – effectiveness 19

#8: User Tracking using Weight • Problem: counting and tracking users is key to physical analytics • Idea: leverage the constancy of a user’s weight to track the user • Approach: – install (hidden) weighing scales at key locations – count people going in and out of a space • Research questions: – reliability of weight measurement for a walking user 20

#9: Spatial Meeting Recorder • Problem: meeting recording is typically only indexed by time • Goal: enable indexing by speaker, and more, by annotating recording with relative spatial positions of speakers • Approach: – use acoustic ranging to establish relative positions of users’ phones – establish positions of speakers relative to the phones • Research questions: – accuracy – energy efficiency 21

Project Proposal • Due Saturday, Aug 31 • One-page max, single column, 10 pt font • What problem is being solved • Why is it important • What is the proposed approach • Expected deliverable • Rough plan 22

Next Lecture: Energy • Papers to be critiqued: – “ An Analysis of Power Consumption in a Smartphone ”, Usenix ATC 2010 – “ Fine Grained Energy Accounting on Smartphones with Eprof ”, EuroSys 2012 • Other papers to read: – “ Carat: Collaborative Energy Debugging for Mobile Devices ”, Usenix HotDep 2012 – “ Who Killed My Battery: Analyzing Mobile Browser Energy Consumption ”, WWW 2012 23

Paper Critiques • One-page max, single column, 10 pt font • Summarize the paper in your own words • Call out key contributions • Identify shortcomings • Suggest improvements • Note: would require reading more than the paper! 24