新加坡國立大學 黃瑋璨 Avatar Mobility in 1
Wei Tsang Ooi Mehul Motani Huiguang Liang Ian Tay Ming Feng Neo National University of Singapore 2
“My life is so great that I literally wanted a second one!” - Dwight Schrute, The Office 3
256x256 m regions. 4
avatar mobility: who is where, when 5
why do we care? 6
research in systems support for NVE 7
8
How to partition a world into regions and assign regions to servers considering - communication cost - hand-over rate - balancing server load : 9
10
How to predict avatar movement (end therefore what a user will see next)? 11
12
AoI-based scheme 13
How many connections? How stable are the connections? 14
supernode-based scheme 15
How to pick supernodes? How stable are the supernodes? 16
how to simulate avatar mobility? 17
random walk random waypoint clustered movement : 18
or, small-scale implementation 19
no large-scale NVE available until recently 20
482,594 residents logged in between 2-9 June 2008 21
secondlife.com/whatis/economy-graphs.php 22
• collect mobility traces of avatars in Second Life • what it means w.r.t. systems design for NVEs? 23
collecting traces 24
how do avatars move inside a distributed virtual environment? 25
how are avatars distributed within a region? 26
how long do they stay at a location? 27
do they move in groups? 28
etc. 29
FPS MMORPG NVE 30
Linden, can we get access to the server traces? No. 31
• Wrote our own client • Parses packets using libsecondlife • Insert bots into regions • Log positions of avatars every 10s 32
difficulties 33
running out of memory 34
anti-bots policy 35
over crowded region 36
inter-region tracking 37
• Wrote our own client • Parses packets using libsecondlife • Insert bots into regions • Log positions of avatars every 10s 38
who is where, when (doing what) 39
40
Freebies 41
The Pharm 42
Isis 43
Ross 44
Mobility Patterns 45
Freebies: number of visits to a cell 46
Freebies: average pause time in a cell 47
Freebies: average speed in a cell 48
Isis: number of visits to a cell 49
caching/prefetching based on popularity of locations? 50
Isis: average pause time in a cell 51
pick supernodes from sticky location? 52
Isis: average speed in a cell 53
mobility model: random walk + pathway ? 54
churn rate 55
56
57
Reasonably high churn (up to 6/min) 58
1 min 10 min 1 hr 2 hr Highly skewed. Some stay for hours. 59
can not pick supernodes uniformly 60
clustering of avatars 61
meeting : encounter between two avatars (within each other AoI) 62
Meet many different avatars. 63
1 min 1 hr 10 min 2 hr Most meetings are short. 64
Meeting size is large. 65
high overhead in maintaining AoI neighbors 66
meeting stability: avg meeting size over num of avatars met 67
Wide range of stability 68
other tidbits 69
little temporal variations can use historical information to predict future 70
rotate 18% of the time Second Life’s prefetching is wasteful 71
25-35% revisits the same region in a day region-based caching? 72
proxy-based texture caching 73
why textures? 74
62 - 81% of traffic are textures 75
316 MB of textures in Isis 76
64m 80 o 77
64m 80 o 78
403 TB of textures retrieved in Isis in a day 79
clients SL servers texture proxy 80
what caching algorithm to used? 81
2Q 82
FIFO LRU 83
cache miss FIFO LRU 84
cache hit FIFO LRU 85
cache hit FIFO LRU 86
scan resistant FIFO LRU 87
88
3Q 89
FIFO LRU Victim Buffer 90
cache hit FIFO LRU Victim Buffer (sorted by popularity) 91
how to define popularity of texture? 92
Freebies: number of visits to a cell 93
little temporal variations can use historical information to predict future 94
popularity of texture = popularity of cell 95
Per-byte Hit Rate 2Q 3Q OPT Ross 50 MB 0.58 0.62 0.70 Ross 25 MB 0.28 0.36 0.47 Freebies 50 MB 0.48 0.50 0.68 Freebies 25 MB 0.21 0.33 0.50 96
conclusion 97
understanding real avatar mobility is crucial to design good NVEs 98
謝謝 歡迎發問及指教 99 99
Recommend
More recommend