ENSC 427: COMMUNICATION NETWORKS SPRING 2018 FINAL PROJECT PRESENTATIONS Comparison of gaming Client/Server paradigms: Peer hosting vs Dedicated Server https://ensc427team2.weebly.com/ Randeep Shahi (301224617) rsshahi@sfu.ca Nathan Zavaglia (301260106) nzavagli@sfu.ca Team 2

Outline Introduction Overview of Related work Problem Description Implementation Results Discussion References

Outline Introduction Overview of Related work Problem Description Implementation Results Discussion References

Introduction Networking in video games has been a huge influence in the way that video game connectivity has evolved Since the introduction of early games Most games were client-server based Recently they have become more peer-to-peer based. We wanted to investigate the differences between a client-server based network paradigm vs a peer-to-peer network paradigm

Overview of Related work Client-Server model most prevalent (Yahyavi and Kemme) [1] Peer-to-Peer can be scalable to large numbers (Knutsson, Honghui Wei and Hopkins) [3] Acceptable delay for MMO but not for fast paced games No need to maintain dedicated servers – P2P is lower cost Not much new research Few simulations Most research dated Using non-standard/custom tools

Problem Description Our goal was to simulate 2 different paradigms of network connectivity Standalone/dedicated servers Peer-to-peer networks Metrics are global ping, local ping, jitter, and packet loss Logical topologies

Outline Introduction Overview of Related work Problem Description Implementation Results Discussion References

Implementation We implemented our simulation using Riverbed Modeler 17.5 Originally proposed to use ns3 Found that riverbed was easier to use for our purposes Our simulations were a variation of our 2 main paradigms(client- server or peer-to-peer) These variations were Normal (no adding or dropping) Client joins  And Client Leaves

Implementation…continued For each simulation Each client connected to the server about 15 to 115 seconds into the simulation Each simulation lasted 30+ minutes of simulated time (simulation time was 35 seconds) Each scenario had 8 clients Simulated on a LAN for simplicity (links are 100baseT duplex ethernet links)

Implementation…continued UDP packets were sent between the clients (and server) Packets sent from clients were 100 bytes each. Increased from 80 bytes as referenced in [4] in order to better saturate network Packets received by clients were 100 bytes for simplicity Clients connected to routers with DS3_PPP links Router to Router/server are connected by 100baseT links

Implementation con’t… Diagram of client server model

Implementation con’t… Diagram of peer-to-peer network

Outline Introduction Overview of Related work Problem Description Implementation Results Discussion References

Results (delay) This was the baseline that we will compare against Ping ranges between 0.095 seconds to 0.090 seconds Curve started to level off as the simulation ended

Results (delay) 3 clients drop out of the sim at 10 minutes, 20 minutes, and 30 minutes As the clients drop out, the ping for the first 3 clients drops a marginal amount ranging from 0.22 seconds to 0.19 seconds Unknown as to where the simulation would have leveled out; hit academic license limit of 50 million events per simulation.

Results (delay) 3 clients are added to the simulation at 10 minutes, 20 minutes, and 30 minutes As the clients are removed , the average number of dropped packets though the internet begins to increase as more clients are added Unknown as to where the simulation would have leveled out; hit academic license limit of 50 million events per simulation.

Results (jitter) This was the baseline that we will compare against Jitter for all 3 clients was about 0.025 seconds Curves level off before simulation ends

Results (jitter) 3 clients drop out of the sim at 10 minutes, 20 minutes, and 30 minutes No noticeable changes were observed from baseline Curves level off before simulation ends

Results (jitter) 3 clients join the sim at 10 minutes, 20 minutes, and 30 minutes No noticeable changes were observed from baseline Curves level off before simulation ends

Outline Introduction Overview of Related work Problem Description Implementation Results Discussion References

Discussion According to our simulation, Clients joining a p2p network has a definite impact on the ping but has a negligible impact on jitter Future works: P2p and client-server delay analysis P2p and client-server in-depth packet loss Modifications to p2p physical topology

References [1 ] A. Yahyavi and B. Kemme, “Peer-to-peer architectures for massively multiplayer online games,” ACM Computing Surveys, vol. 46, no. 1 , pp. 1 –51 , Jan. 201 3. [2] C. Neumann, N. Prigent, M. Varvello, and K. Suh, “Challenges in peer-to-peer gaming,” SIGCOMM Comput. Commun. Rev., vol. 37, no. 1 , pp. 79–82, Jan. 2007. [Online]. Available: http://doi.acm.org/1 0.1 1 45/1 1 98255.1 1 98269 [3] B. Knutsson, Honghui Lu, Wei Xu and B. Hopkins, "Peer-to-peer support for massively multiplayer games," IEEE INFOCOM 2004, 2004, pp. 1 07. doi: 1 0.1 1 09/INFCOM.2004.1 354485 [4] Johannes Färber. 2002. Network game traffic modelling. In Proceedings of the 1 st workshop on Network and system support for games (NetGames '02). ACM, New York, NY, USA, 53-57. doi: http://dx.doi.org/1 0.1 1 45/566500.566508 [5] Y. W. Bernier, “Latency Compensating Methods in Client/Server In-game Protocol Design and Optimization,” Valve Developer Community. [Online]. Available: https://developer.valvesoftware.com/wiki/Latency_Compensating_Methods_in_Client/Server_Ingam e_Protocol_Design_and_Optimization. [Accessed: 09-Feb-201 8].