Switchboard: A Matchmaking System for Multiplayer Mobile Games - PowerPoint PPT Presentation

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

Breakthrough of Mobile Gaming iPhone
App
Store
 350K 
applica,ons
 47%
Time
on

 20%
apps,
80%
downloads 
 Mobile
Apps
 Spent
Gaming
 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” 2


Mobile Games: Now and Tomorrow Increasing
Interac6vity 
 Single‐player
 Mul6player

 Mul6player

 Mobile
 Turn‐based
 Fast‐ac6on
 
 
 
 (mobile
today)

 (mobile
today)

 (mobile
soon)
 
 3


Key Challenge Bandwidth
is
fine: 
 250
kbps 

 to
host
16‐player
Halo
3
game
 Game
Type Latency
Threshold Delay
bounds
are

 First‐person,
Racing ≈
100
ms much
6ghter 
 Sports,
Role‐playing ≈
500
ms Real‐,me
Strategy ≈
1000
ms Challenge:
 find
 groups
of
peers
 than
can
 play
well
together 
 4


The Matchmaking Problem End-to-end Latency Threshold 
 Connection Latency 
 Match 
to
sa+sfy
total
 delay
bounds 
 Clients 
 5


Instability in a Static Environment 310
 Median
Latency
(ms)
 290
 270
 Due
to
instability,
 must
consider
latency
 distribu6on
 250
 230
 210
 190
 170
 150
 9:36
 9:50
 10:04
 10:19
 10:33
 10:48
 11:02
 11:16
 11:31
 11:45
 12:00
 Time
of
Day
(AM)
 6


End-to-end Latency over 3G First‐person
Shoot. 
 Racing 
 Sports 
 Real‐6me 
Strategy 
 1
 0.8
 Empirical
CDF 
 Peer‐to‐peer
 reduces
latency

 0.6
 and
is

 cost‐effec6ve
 AT&T
to
AT&T
Direct
 0.4
 AT&T
to
AT&T
via
Bing
 0.2
 AT&T
to
AT&T
via
Duke
 AT&T
to
AT&T
via
UW
 0
 0
 100
 200
 300
 400
 500
 600
 RTT
(in
ms)
 7


The Matchmaking Problem Targe,ng
 3G :
 Latency 
not
Bandwidth 
 Measurement
/
PredicTon 
 at
game
+mescales
 interac+vity
is
key
 play
anywhere
 Link
Performance
 Grouping
 P2P
Scalability
 8


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


State of the Art ● Latency estimation  Pyxida,
stable
network
coordinates;
Ledlie
et
al.
[NSDI
07]
  Vivaldi,
distributed
latency
est.;
Dabek
et
al.
[SIGCOMM
04]
 Latency
es,ma,on
and
matchmaking
are

 ● Game matchmaking for wired networks established
 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]
  Interac,ons
with
applica,ons;
Liu
et
al.
[MobiCom
08]
  Empirical
3G
performance;
Tan
et
al.
[InfoCom
07]
  TCP/IP
over
3G;
Chan
&
Ramjee

[MobiCom
02]
 10


A “Black Box” for Game Developers IP
network
 Internet 
 GGSN
 GGSN
 “Black

 SGSN
 SGSN
 Box” 
 RNC
 RNC
 End‐to‐end
 Performance
 Crowdsourced
 Link
Performance
 Measurement
 (over
6me)
 11


Crowdsourcing 3G over Time Latency Similarity by Time 
 Time 
 12


Crowdsourcing 3G over Space Latency Similarity by Distance 
 13


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
distribu,on?
 Details
of
parameter
space
lem
for
the
paper

 (our
goal
is
not
to
iden6fy
the
exact
causes)
  …
 ● How does 3G performance vary over space?  Signal
strength?
Mobility
speed?

  Phones
under
same
cell
tower?
  Same
part
of
the
cellular
network?
  …
 14


Methodology ● Platform  Windows
Mobile
and
Android
phones
  HSDPA
3G
on
AT&T
and
T‐Mobile
 ● Carefully deployed phones  Con,nuous
measurements
  Simultaneous,
synchronized
traces
at
mul,ple
sites
 ● Several locations  Princeville,
Hawaii
  Redmond
and
Seanle,
Washington
  Durham
and
Raleigh,
North
Carolina
  Los
Angeles,
California
 15


Stability over Time (in a Static Environment) Redmond,
AT&T,
15m
Intervals
 1
 0.8
 Empirical
CDF
 Live
characteriza,on
is
 necessary 
 and
is
 feasible
 0.6
 Performance
drims
over
 Similar
latencies
under
 0.4
 longer
,me
periods
 the
same
tower
 Black 
line
represents
phone
1

 0.2
 (all
other
lines
phone
2)
 0
 120
 140
 160
 180
 200
 220
 240
 RTT
(Msec)
 16


Stability over Space (at the same time) Similarity
at
the
 Substan,al
 Divergence
between
 same
cell
tower
 varia6on
 nearby
towers
 1
 0.8
 Empirical
CDF
 0.6
 S‐home
 0.4
 Latona
 U
Village
 0.2
 Herkimer
 Northgate
 1st
Ave
 0
 0
 50
 100
 150
 200
 RTT
difference
at
90th
percenTle
(ms)
 17


Switchboard : crowdsourced matchmaking Switchboard
Cloud
Service
 
 on
MSFT
Azure
 Game
 Grouping
Agent
 Network
TesTng
 Service
 Latency
 Data
 Measurement
 Latency
 Controller
 EsTmator
 18


Scalability through Reuse … ● Across Time  Stable
distribu,on
over
15‐minute
,me
intervals
 ● Across Space  Phones
can
share
probing
tasks
equitably
for
each
tower
 ● Across Games  Shared
cloud
service
for
any
interac,ve
game
 19


Client Matchmaking Delay 1
 0.8
 Empirical
CDF
 0.6
 Switchboard
clients
 benefit
 from
deployment
at
scale 
 0.4
 1
client
arrival/sec
 Total
1
client/sec
 0.2
 10
client
arrival/sec
 Total
10
clients/s
 0
 0
 100
 200
 300
 400
 500
 600
 700
 Time
unTl
placed
in
group
(s)
 20


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


k
you!
 cs.duke.edu/~jgm
 jgm@cs.duke.edu