1. Hybrid Architecture
for Networked Games
1

2. single
completely
centralized
decentralized
server
2

3. P2P with
Central Arbiter
3

4. A trusted central arbiter receives all game
traffic and resolve conflicts if necessary.
4

5. Benefits of
P2P + Arbiter
5

6. low latency among players
6

7. have trusted authority now
7

8. Drawbacks of
P2P + Arbiter
8

9. limited scalability
arbiter becomes bottleneck
9

10. Mirrored Servers
Architecture
10

11. Multiple servers, each replicating the complete game
states, serve clients from different geographical
regions. Each client connects to one server.
11

12. Servers may be connected with high speed,
provisioned networked, reducing synchronization
latency among the servers.
12

13. Updates from a client are sent to its server, which then
forwards it to other servers. The servers then forward
the updates to all relevant clients.
13

14. P2P-style state synchronization
among servers is easier, since
1. servers can be trusted
2. clocks can be synchronized
3. IP multicast may be used
4. higher availability
5. shorter delay
14

15. Benefits of
Mirrored Servers
15

16. no single point of failures
A client can use another replicated server
if their nearest server fails.
16

17. increase total network capacity
good if networking bandwidth is the
bottleneck
17

18. lower latency between
client and server
18

19. Drawbacks of
Mirrored Servers
19

20. latency between players increases
due to additional forwarding
20

21. cannot scale to large world
each server still keeps the complete states
21

22. how to assign
players to servers?
22

23. Assigning players to closest servers may lead to
overloaded servers.
23

24. Assigning players to remote servers may increase
end-to-end latency.
24

25. But we only care about latency between players
that are relevant to each other.
25

26. An Open Problem
26

27. let
AOI(p) be the set of players relevant to p
S(p) be the server of player p
d(i,j) be the delay between i and j
P(s) be the set of players assigned to server s
27

28. let
L(p) be the average latency experience by p,
considering only relevant players
28

29. want to find S() such that
maximum L(p) is minimized over all p
subjected to
maximum P(s) is constrained for all s
29

30. Zoned Servers
Architecture
30

31. Divide the game world into independent regions.
One server is in-charged of one region.
31

32. A client connects to the server serving its current
region, and is handed-off to another server when it
moves to another region.
32

33. Servers seldom communicate with each other.
33

34. Updates from a client are sent to its server, which then
forwards it to all relevant clients in the same region.
34

35. 35

36. Advantages of
Zoned Servers
36

37. Can scale to large world
Just add more servers
37

38. Drawbacks of
Zoned Servers
38

39. Single point of failures
(But can deploy mirrors for each region)
39

40. Constrained Game Designs
40

41. Supporting Seamless
Game Worlds
41

42. Divide the game world into regions.
One server is in-charged of one region.
42

43. Player may move around the world seamlessly
43

44. A client may see events and objects
from neighboring regions.
44

45. Events occurring in one region may affect
objects in another region.
45

46. Advantage: Partitioning is transparent
to the players. No artificial constraints in
game design.
46

47. Advantage: We can now resize the
regions to load-balance among the servers,
improving server utilization.
47

48. A client connects to the server serving its current
region, and is handed-off to another server when it
moves to another region.
48

49. Server communicates with each other to transfer
states, update states etc.
49

50. Updates from a client are sent to its server, which then
forwards it to all relevant clients in the same region or
neighboring regions.
50

51. or
51

52. A client connects to the servers serving its current
region and regions it subscribes to. Updates from
neighboring server are sent directly.
52

53. Q: How to partition
the game world?
53

54. Ideally: each server
should handle the same
number of players
54

55. Ideally: minimize
communications
between servers
55

56. Ideally: minimize the
number of hand offs
56

57. Idea: divide the game
world into small cells
and assign group of
cells to servers.
57

58. 58

59. I will simplify the representation to a grid of cells with
number indicating the number of players (or load
incurred on server) in this cell.
2 0 3
3 1 4
1 1 0
59

60. Centralized approach
to partitioning
60

61. Balanced Deployment
Algorithm
by Bart De Vleeschauwer et al
61

62. Idea: sort the cells and assign
them one-by-one, highest load
first, to the servers to minimize
the maximum server load
62

63. Suppose we have only two servers (blue and red).
2 0 3
3 1 4
1 1 0
63

64. Blue’s load = 4 Red’s load = 0
2 0 3
3 1 4
1 1 0
64

65. Blue’s load = 4 Red’s load = 3
2 0 3
3 1 4
1 1 0
65

66. Blue’s load = 4 Red’s load = 6
2 0 3
3 1 4
1 1 0
66

67. Blue’s load = 6 Red’s load = 6
2 0 3
3 1 4
1 1 0
67

68. Blue’s load = 8 Red’s load = 7
2 0 3
3 1 4
1 1 0
68

69. Problem: neighboring cells are not adjacent. There
are 8 “borders” in this solution. Communication
overhead is not considered.
2 0 3
3 1 4
1 1 0
69

70. Let’s define a cost function for
assigning a cell c to a server S:
cost(c,S) =
computation load increase in S +
communication load increase in S
70

71. For simplicity,
CPU load = players,
newtork load = borders
71

72. Clustered Deployment
Algorithm
by Bart De Vleeschauwer et al
72

73. Start with each cell as one
cluster
73

74. Repeatedly merge two adjacent
clusters with least overhead until
number of clusters equals to
number of servers.
74

75. 2 0 3
3 1 4
1 1 0
75

76. Least cost = 1 + 3 = 4
2 0 3
3 1 4
1 1 0
76

77. Least cost = 2 + 3 = 5
2 0 3
3 1 4
1 1 0
77

78. Least cost = 1 + 5 = 6
2 0 3
3 1 4
1 1 0
78

79. Least cost = 3 + 5 = 8
2 0 3
3 1 4
1 1 0
79

80. 2 0 3
3 1 4
1 1 0
80

81. Just a heuristic
the optimization problem
is NP-complete
81

82. Last few steps involve heavily
loaded cells and can create
uneven deployment.
82

83. We can tweak this heuristic in
many ways to improve it, but it’s
still a heuristic.
83

84. Simulated Annealing
by Bart De Vleeschauwer et al
84

85. A general method for finding
good solution in an optimization
problem (meta-heuristic)
85

86. At every step, move towards a nearby,
better configuration probabilistically
(allowing a move towards worse
solution, prevent stuck at local
minimum)
86

87. cost
config
87

88. Start with a solution from a
heuristic.
88

89. Move to a nearby config:
swap cells with an adjacent
server, or move cell from one
server to another
89

90. Is it a better solution?
If so, keep it as current solution
with some probability.
90

91. 2 0 3
3 1 4
1 1 0
91

92. Experimental results show that
Simulated Annealing works best
among the methods.
92

93. Disadvantages of
centralized approach
93

94. Expensive to compute a
good solution
94

95. How frequent should
the world be
repartitioned?
95

96. Localized World
Partitioning
96

97. Idea: Only repartition locally if a
server is overloaded. Look for a
lightly loaded server to shed
some load to.
97

98. overload
safe
light
98

99. Overloaded server sheds load until it’s load is below a
safe level. The new load of a previously lightly loaded
should remain below safe level.
overload
safe
light
99

100. l+o-s<s
l < 2s - o
100

101. Who to shed to?
101

102. Minimize overhead: try to shed
to as few neighboring server as
possible
102

103. Migrate cells to neighboring
servers that are not overloaded
103

104. Overload = 10, Safe = 7, Light = 4
0 1 1
2 0 3
1 1 4
5 3 0
104

105. Green server is overloaded.
Shed some load to neighbors.
0 1 1
2 0 3
1 1 4
5 3 0
105

106. If still not enough, look at the list of
lightly loaded servers maintained.
Pick the least load and shed to it.
106

107. Green server is still overloaded.
Neighbor (purple) server has hit the safe threshold.
0 1 1
2 0 3
1 1 4
5 3 0
107

108. Shed load to remote server.
But this could lead to high communication cost.
0 1 1
2 0 3
1 1 4
5 3 0
108

109. When communication overhead is too high,
isolated cells can be migrated to a nearby
server, “merging” with neighboring regions to
form contiguous regions, as long as the load
remains below the safe threshold.
0 1 1
2 0 3
1 1 0
1 3 0
109

110. Another Open Problem
110

111. No consideration on hand offs
111

112. Idea: If traffic is high between
point A and point B, then A and B
should be managed by the same
server.
112

113. 3 2
2 1 1
0 1
2 4 2
0 2
1 1 2
2 4
113

114. Minimizing hand-off means finding minimum cuts.
3 2
2 1 1
0 1
2 4 2
0 2
1 1 2
2 4
114

115. Minimizing hand-off means finding minimum cuts.
3 2
2
1
2 4 2
2
1 2
2 4
115

116. Minimizing hand-off means finding minimum cuts.
3 2
2
1
2 4 2
2
2 4
116

117. Question: how to combine the
three objectives:
1. limiting number of players
2. reduce communication cost
3. reduce number of hand offs
117