SLIDE 1
Avatar Mobility in
Wei Tsang Ooi Mehul Motani Huiguang Liang Ian Tay Ming Feng Neo National University of Singapore
Avatar Mobility in Wei Tsang Ooi Mehul Motani Huiguang Liang Ian - - PowerPoint PPT Presentation
Avatar Mobility in Wei Tsang Ooi Mehul Motani Huiguang Liang Ian - - PowerPoint PPT Presentation
Avatar Mobility in Wei Tsang Ooi Mehul Motani Huiguang Liang Ian Tay Ming Feng Neo National University of Singapore avatar mobility: who is where, when why do we care? research in systems support for NVE how to partition a world into
SLIDE 2
SLIDE 3
why do we care?
SLIDE 4
research in systems support for NVE
SLIDE 5
SLIDE 6
how to partition a world into cells and assign cells to servers considering
- interaction across cells
- movement across cells
- avatar density in each cell
:
SLIDE 7
SLIDE 8
how to predict avatar movement (and therefore what a user will see next)?
SLIDE 9
SLIDE 10
AoI-based scheme
SLIDE 11
how many connections? how stable are the connections?
SLIDE 12
supernode-based scheme
SLIDE 13
how to pick supernodes? how stable are the supernodes?
SLIDE 14
how to simulate avatar mobility?
SLIDE 15
random walk random waypoint clustered movement :
SLIDE 16
- r,
small-scale implementation
SLIDE 17
no large-scale NVE available until recently
SLIDE 18
“My life is so great that I literally wanted a second one!”
- Dwight Schrute, The Office
SLIDE 19
256x256 m regions.
SLIDE 20
SLIDE 21
SLIDE 22
SLIDE 23
SLIDE 24
470,478
residents logged in between 12-19 June 2008
SLIDE 25
secondlife.com/whatis/economy-graphs.php
SLIDE 26
- collect mobility traces of
avatars in Second Life
- what it means w.r.t. systems
design for NVEs?
SLIDE 27
Trace Collection
SLIDE 28
FPS MMORPG NVE
SLIDE 29
Linden, can we get access to the server traces?
No.
SLIDE 30
- Wrote our own client
- Parses packets using libsecondlife
- Insert bots into regions
- Log positions of avatars every 10s
SLIDE 31
who is where, when (doing what)
SLIDE 32
SLIDE 33
Isis: number of visits to a cell
SLIDE 34
caching/prefetching based on popularity of locations?
SLIDE 35
Isis: average pause time in a cell
SLIDE 36
pick supernodes from sticky location?
SLIDE 37
Isis: average speed in a cell
SLIDE 38
mobility model: random walk + pathway ?
SLIDE 39
Reasonably high churn (up to 6/min)
SLIDE 40
1 min 10 min 1 hr 2 hr
Highly skewed. Some stay for hours.
SLIDE 41
cannot pick supernodes uniformly
SLIDE 42
meeting: encounter between two avatars (within each other AoI)
SLIDE 43
1 min 10 min 1 hr 2 hr
Most meetings are short.
SLIDE 44
Meeting size is large.
SLIDE 45
high overhead in maintaining AoI neighbors
SLIDE 46
- very little temporal variations
(predictable)
- avatars rotates 18% of time
(SL’s prefetching is wasteful)
- 25-35% revisits regions within
a day (region-based caching?)
SLIDE 47