1. Switchboard: A Matchmaking System for Multiplayer Mobile Games Justin Manweiler, Sharad Agarwal, Ming Zhang, Romit Roy Choudhury, Paramvir Bahl ACM MobiSys 2011 Battling demons and vampires on your lunch break…

2. Breakthrough of Mobile Gaming 2 Windows Phone 7 Top 10+ apps are games Windows Phone 7 Top 10+ apps are games John Carmack (Wolfenstein 3D, Doom, Quake)… “multiplayer in some form is where the breakthrough, platform-defining things are going to happen in the mobile space” iPhone App Store 350K applications 20% apps, 80% downloads iPhone App Store 350K applications 20% apps, 80% downloads 47% Time on Mobile Apps Spent Gaming

3. Mobile Games: Now and Tomorrow 3 Increasing Interactivity Single-player Mobile (mobile today) Multiplayer Turn-based (mobile today) Multiplayer Fast-action (mobile soon)

4. Key Challenge 4 Game TypeLatency Threshold First-person, Racing≈ 100 ms Sports, Role-playing≈ 500 ms Real-time Strategy≈ 1000 ms Challenge: find groups of peers than can play well together Bandwidth is fine: 250 kbps to host 16-player Halo 3 game Bandwidth is fine: 250 kbps to host 16-player Halo 3 game Delay bounds are much tighter Delay bounds are much tighter

5. The Matchmaking Problem 5 Match to satisfy total delay bounds End-to-end Latency Threshold Clients Connection Latency

6. Instability in a Static Environment 6 Due to instability, must consider latency distribution Due to instability, must consider latency distribution

7. End-to-end Latency over 3G 7 First-person Shoot. Racing Real-time Strategy Sports Peer-to-peer reduces latency and is cost-effective

8. The Matchmaking Problem 8 Link Performance P2P Scalability Grouping Targeting 3G: play anywhere Targeting 3G: play anywhere Latency not Bandwidth interactivity is key Latency not Bandwidth interactivity is key Measurement / Prediction at game timescales Measurement / Prediction at game timescales

9. Requirements for 3G Matchmaking ●Latency estimation has to be accurate  Or games will be unplayable / fail ●Grouping has to be fast  Or impatient users will give up before a game is initiated ●Matchmaking has to be scalable  For game servers  For the cellular network  For user mobile devices 9

10. State of the Art ●Latency estimation  Pyxida, stable network coordinates; Ledlie et al. [NSDI 07]  Vivaldi, distributed latency est.; Dabek et al. [SIGCOMM 04] ●Game matchmaking for wired networks  Htrae, game matchmaking in wired networks; Agarwal et al. [SIGCOMM 09] ●General 3G network performance  3GTest w/ 30K users; Huang et al. [MobiSys 2010]  Interactions with applications; Liu et al. [MobiCom 08]  Empirical 3G performance; Tan et al. [InfoCom 07]  TCP/IP over 3G; Chan & Ramjee [MobiCom 02] 10 Latency estimation and matchmaking are established for wired networks Latency estimation and matchmaking are established for wired networks

11. A “Black Box” for Game Developers 11 Internet IP network RNC SGSN RNC SGSN GGSN Link Performance (over time) End-to-end Performance “Black Box” Crowdsourced Measurement

12. Latency Similarity by Time Crowdsourcing 3G over Time 12 Time

13. Crowdsourcing 3G over Space 13 Latency Similarity by Distance

14. Can we crowdsource HSDPA 3G? ●How does 3G performance vary over time?  How quickly do old measurements “expire”?  How many measurements needed to characterize the latency distribution?  … ●How does 3G performance vary over space?  Signal strength? Mobility speed?  Phones under same cell tower?  Same part of the cellular network?  … 14 Details of parameter space left for the paper (our goal is not to identify the exact causes) Details of parameter space left for the paper (our goal is not to identify the exact causes)

15. Methodology ●Platform  Windows Mobile and Android phones  HSDPA 3G on AT&T and T-Mobile ●Carefully deployed phones  Continuous measurements  Simultaneous, synchronized traces at multiple sites ●Several locations  Princeville, Hawaii  Redmond and Seattle, Washington  Durham and Raleigh, North Carolina  Los Angeles, California 15

16. Stability over Time (in a Static Environment) 16 Black line represents phone 1 (all other lines phone 2) Similar latencies under the same tower Performance drifts over longer time periods Live characterization is necessary and is feasible

17. Stability over Space (at the same time) 17 Similarity at the same cell tower Divergence between nearby towers Substantial variation Substantial variation

18. Switchboard: crowdsourced matchmaking 18 Game Network Testing Service Latency Data Latency Estimator Measurement Controller Grouping Agent

19. Scalability through Reuse… ●Across Time  Stable distribution over 15-minute time intervals ●Across Space  Phones can share probing tasks equitably for each tower ●Across Games  Shared cloud service for any interactive game 19

20. Client Matchmaking Delay 20 1 client arrival/sec Switchboard clients benefit from deployment at scale

21. Conclusion ●Latency: key challenge for fast-action multiplayer ●3G latency variability makes prediction hard ●Crowdsourcing enables scalable 3G latency estimation ●Switchboard: crowdsourced matchmaking for 3G 21

22. QUESTIONS? Thank you! 22 cs.duke.edu/~jgm jgm@cs.duke.edu

23. Stability over Time (in a Static Environment) 23 At timescales longer or shorter than 15 minutes: successive interval pairs have less similarity

24. How Many Measurements Req.? 24 Reasonably small number of measurements are required per 15-minute interval

25. End-to-End Performance 25 End-to-end performance predictable as the sum of edge and Internet latencies

26. ICMP Probes by Client Arrival Rate 26 More clients = less probing overhead for each

27. Scalability by Client Arrival Rate 27 After the initial warming period, later clients reuse effort by earlier clients After the initial warming period, later clients reuse effort by earlier clients

28. Group Size by Client Arrival Rate 28 Availability of high-quality matches increases with utilization

29. Timescale Statistical Analysis 29