The QoS requirements of game messages may differ because of the latter’s intrinsic characteristics. In this paper, we propose three content-based strategies for quantifying the effects of different QoS levels. The strategies assign appropriate QoS requirements for game messages based on our analysis of Angel’s Love action logs. We evaluate several transport protocols, including TCP, UDP, SCTP, DCCP, and our content-based transport protocol using the action logs of Angel’s Love. Through simulations, we quantify the performance of our content-based strategies. The results show that the strategies incur much lower end-to-end delay and end-to-end jitter than existing transport protocols.

1. Quantifying the Effect of Content‐based Transport
MNe
t
Strategies for Online Role Playing Games
Chih-Ming Chen1, Kuan-Ta Chen2, and Polly Huang1
M 1National Taiwan University 2Academia Sinica
Lab
Motivation Transport Strategies Performance Evaluation
Protocol MMORPGs Average end‐to‐end transmission latencies
Message Type In-Order Delivery Reliability
TCP World of Warcraft, Lineage I/II, Guild Wars, Ragnarok
Move Client Traffic Server Traffic
Online, Anarchy Online, Angel’s Love
600
UDP EverQuest, City of Heroes, Star Wars Galaxies, Ultima Attack √ TCP P_MRO TCP P_MRO
180 200 220 240 260 280 300
UDP P_MR UDP P_MR
Online, Final Fantasy XI Talk √ √
500
Mean delay (ms)
Mean delay (ms)
SCTP P_M SCTP P_M
TCP/UDP Dark Age of Camelot
Multi‐streaming
400
There is no consensus on protocols for MMORPGs • Put different message types into separate streams
300
• e.g., move message, attack message, talk message
MMORPGs requirements
200
Optional ordering
Low transmission latency • Certain message types do not require in‐order transport 20 40 60 80 100 120 140 20 40 60 80 100 120 140
No unexpected “lags” Client # Client #
• e.g., move message, attack message
Optional reliability Average end‐to‐end delay jitters (standard deviation of delays)
Objective • Certain message types do not require reliability
• e.g., move message
Client Traffic Server Traffic
250
600
TCP P_MRO TCP P_MRO
Quantify the effect of content‐based transport strategies Strategy Multi-Streaming Optional Optional
UDP P_MR UDP P_MR
200
500
SCTP P_M SCTP P_M
Mean jitter (ms)
Mean jitter (ms)
Evaluate existing transport protocols Ordering Reliability
400
150
MRO √
300
• TCP, UDP, DCCP, SCTP
100
MR √ √
200
Propose and evaluate three content‐based transport strategies
50
100
M √ √ √
usingnetwork simulations
0
0
20 40 60 80 100 120 140 20 40 60 80 100 120 140
Real-Life Game Traces
Real-Life
Client # Client #
Simulation Setup
Users’ action trace of Angel’s Love Trace‐driven network simulation using ns‐2 simulator Summary
Fishbone topology
• 1 game server Improvement of Delay Improvement of Delay Jitter
• 2 network routers Client Traffic Server Traffic Client Traffic Server Traffic
100
100
100
100
70 Clients 70 Clients 70 Clients 70 Clients
• n game clients 140 Clients
75
140 Clients 140 Clients 79 140 Clients
80
80
80
80
73
69 69
• m cross traffic pairs 55
60
60
60
60
Score
Score
Score
Score
45 46
Client Node (CN)
40
40
40
40
33
28
CNi+2 . . . CN CN2 . . . CN Traffic Node (TN)
20
20
20
20
8 10
n i 0
2 0 0
2
0 1 1 0 1
5 5
CNi+1 CN1
0
0
0
0
P_MRO P_MR P_M P_MRO P_MR P_M P_MRO P_MR P_M P_MRO P_MR P_M
Protocol Protocol Protocol Protocol
Client to Server Server to Client
TNj+1 TN1 Server Protocol Delay Jitter Delay Jitter
TNj+2 . . . TNm TN2
TN
. . . j TCP NA NA NA NA
SCTP ≈ ≈ ★ ★
DCCP (TCP-like) ★★★★ ★★★★ ★★★ ★★★
Bandwidth & propagation delay DCCP (TFRC) ☆ ☆ ☆ ☆
Link Bandwidth Propagation Delay UDP ★★★★ ★★★★ ★★★★ ★★★★
Server <--> Router 600 Kbps 70 ms PMRO ≈ ≈ ★ ★
Router <--> Router 600 Kbps 70 ms PMR ≈ ≈ ★★ ★★
Router <--> Client 64 ~ 128 Kbps 70 ms PM ★★★ ★★★ ★★★ ★★
User ID Actions (M: Move, A: Attack, T: Talk) Description: NA denotes incomparable, ≈ denotes similar, ☆ denotes worse, ★ denotes better, ★ ★ denotes
Router <--> Traffic Node 64 ~ 128 Kbps 70 ms
10159 MMMMMMMMMMMMMAMMAMMMMMMAMMMMMMMMMM much better, ★ ★ ★ denotes good, ★ ★ ★ ★ denotes very good, and ★ ★ ★ ★ denotes excellent.
11 pairs cross traffic with 500 Kbps sending rate
12454 MMAAMAMMMMMAAMAMAMAMAMAMAMAMAMAMAMAMA
16728 MMMMMMMMMMMMMMMMMMMMMMAMMTTTTTTTTTTTT Multimedia Networking and Systems Lab
Institute of Information Science, Academia Sinica
http://mmnet.iis.sinica.edu.tw