Characterizing Virtual
Population in Games
1

World of Warcraft
2

Use an add-on written in Lua
3

monitor
1100 random players
over 71 days
4

median num of players per poll
vs time of the day polled
5

CDF of all session length, and
median session length of each player
6

sufficiently long session length
good for P2P games
7

8

can use level to
predict session length?
9

CDF of median downtimes and
median session length of each player
10

Availability of players (time online/total time)
in February 2007
11

12

13

monitor
3 locations
Stormwind City
Stanglehorn Vale
Burning Steppes
14

15

can use zone to
predict session length?
16

density: num of players* in a
zone
*: out of 1100 players tracked
17

Mean Location density of 117 locations
18

Difference in Density between Dec and Feb
19

20

dynamic load balancing not
needed?
21

“My life is so great that I literally wanted a second one!”
- Dwight Schrute,The Office
22

256x256 m regions.
23

24

25

26

27

avatar mobility:
who is where, when
28

why do we care?
29

research in systems
support for NVE
30

31

How to partition a world into regions
and assign regions to servers considering
- communication cost
- hand-over rate
- balancing server load
:
32

33

How to predict avatar movement (end
therefore what a user will see next)?
34

35

AoI-based scheme
36

How many connections?
How stable are the connections?
37

supernode-based scheme
38

How to pick supernodes?
How stable are the supernodes?
39

how to simulate
avatar mobility?
40

random walk
random waypoint
clustered movement
:
41

or,
small-scale
implementation
42

no large-scale NVE
available until recently
43

482,594
residents logged in between 2-9 June 2008
44

secondlife.com/whatis/economy-graphs.php
45

• collect mobility traces of
avatars in Second Life
• what it means w.r.t. systems
design for NVEs?
46

collecting traces
47

how do avatars move
inside a distributed
virtual environment?
48

how are avatars
distributed within a
region?
49

how long do they stay
at a location?
50

do they move in
groups?
51

etc.
52

FPS MMORPG NVE
53

Linden, can we get access to
the server traces?
No.
54

• Wrote our own client
• Parses packets using libsecondlife
• Insert bots into regions
• Log positions of avatars every 10s
55

difficulties
56

running out of memory
57

anti-bots policy
58

over crowded region
59

inter-region tracking
60

• Wrote our own client
• Parses packets using libsecondlife
• Insert bots into regions
• Log positions of avatars every 10s
61

who is
where,
when
(doing what)
62

63

Freebies
64

The Pharm
65

Isis
66

Ross
67

Mobility Patterns
68

Freebies: number of visits to a cell
69

Freebies: average pause time in a cell
70

Freebies: average speed in a cell
71

Isis: number of visits to a cell
72

caching/prefetching
based on popularity of
locations?
73

Isis: average pause time in a cell
74

pick supernodes from
sticky location?
75

Isis: average speed in a cell
76

mobility model:
random walk +
pathway ?
77

churn rate
78

79

80

Reasonably high churn (up to 6/min)
81

1 min 10 min
1 hr 2 hr
Highly skewed. Some stay for hours.
82

cannot pick
supernodes uniformly
83

clustering of
avatars
84

meeting: encounter
between two avatars
(within each other AoI)
85

Meet many different avatars.
86

1 min
1 hr
10 min 2 hr
Most meetings are short.
87

Meeting size is large.
88

high overhead in
maintaining AoI
neighbors
89

meeting stability:
avg meeting size
over
num of avatars met
90

Wide range of stability
91

other tidbits
92

little temporal variations
can use historical information
to predict future
93

rotate 18% of the time
Second Life’s prefetching is
wasteful
94

25-35% revisits the same
region in a day
region-based caching?
95