Review: Cooperative Networking

V. N. Padmanabhan, H. J. Wang, P. A. Chou, and K. Sripanidkulchai, "Distributing Streaming Media Content Using Cooperative Networking," ACM NOSSDAV, May 2002.

There are two types of content distribution: (a) infrastructure based where the traditional client-server mechanism is followed. (b) P2P-based where there is not central server, just the originator. This paper introduces cooperative networking or CoopNet which combines the idea of these two content distribution mechanism. The originator of the content provides a center point which is contact for receiving the content. It keeps track of the peers downloading the content and when it is overloaded, it hands over some of the distribution work to the peers that have already connected or received the content.

CoopNet considers two types of streaming media content distribution: real time streaming and on-demand delivery of recorded media. For real-time streaming it constructs distribution trees made up of the participating peers so that not all the peers are connected to the central server. The central server is contacted only to find the join point. CoopNet implements multiple description coding (MDC) to split the stream for multiple distribution trees. The distribution trees are constructed based on the location of the peers and third party locationing service is used for this purpose. On-demand delivery is relatively less complicated and works much similar to the normal P2P file sharing. This research proposes of using distributed delivery using MDC.

Read more!

Review: Observations on Game Server Discovery Mechanisms

Tristan Henderson, "Observations on Game Server Discovery Mechanisms," In proceedings of the 1st workshop on Network and system support for games, pp 47-52, April 2002.

The author of this paper analyses the game traces to find the way how the clients chose a game server. In the current conditions, the server discovery is almost centralized, where the game server advertises themselves in some centralized directory service. The clients query this directory servers to get the list of servers, pings the servers, and chose a server that is nearby. The directory service is distributed over small number of servers positioned at different geographical locations to serve the queries originate from the area.

This centralized directory server-based implementation suffers from many drawbacks: single point failure, stale information, redundancy due to game servers advertising themselves in many directory servers, dependency on network conditions, and long server query time are some of the drawbacks.

motivated by the success of P2P file sharing system, a P2P server discovery mechanism is proposed in this paper. Each client maintains a partial list of existing servers it used in the past. These lists can be aggregated on-demand (when a new client requests) by a gossiping mechanism involving both clients and servers (servers act as the connecting points). This design alleviates or reduces the problems in the centralized implementation mentioned above. The paper does not provide any experimental/simulation results to evaluate the performance of the proposed system.

Read more!

Review: On the Geographic Distribution of on-line Game Servers and Players

Wu-chang Feng and Wu-chi Feng, "On the geographic distribution of on-line game servers and players," In proceedings of the 2nd workshop on Network and system support for games, pp 173-179, May 2003.

It is established that a large network delay between the on-line game servers and the players affects the playing experience. Therefore it is natural to assume that the distribution of the players is localized to the the nearest game server. This paper analyses the distribution of both the game servers and the players of three first first shooter (FPS) games. The IP addresses of the servers in the world is collected by initiating server queries from a game client. Player distribution is considered from the traces of only one game server. GeoByte geolocationing system is used to find the geographical location of the extracted IP addresses.

The distribution of the servers is concentrated in the northern hemisphere distributed in America, Europe, and east Asia (as anybody can guess). The distribution of the players accessing the one game server in CA, USA is bit surprising because 45% of the players are not from Americas. The authors of the paper discusses few reasons for this behavior: (a) disparity between geographical and network proximity; (b) application delay in servers; (c) broken server selection mechanism; (d) user preference; and (e) overloaded local servers. The paper further analyses the traces to provide the proof for the last reason. The traffic from overseas comes to the server when it is the peak time at places of the clients.

Read more!

Review: On the Impact of Delay on Real Time multi player Games

Lothar Pantel and Lars C. Wolf, "On the Impact of Delay on Real Time Multiplayer Games," In proceedings of the 12th international workshop on Network and operating systems support for digital audio and video, pp 23-29, 2002

The authors establish the growth of the multiplayer games from a splitted screen on a single computer to playing against humans over the Internet. The impact of delay imposed by the network in a distributed gaming scenario is first qualitatively presented. The delay can lead to a inconsistent presentation of the game state on the computers of the players leading to a confused or unfair results. Such inconsistencies are then studied using an experimental setup. Then the delay is handled by introducing a static location presentation delay. However, location presentation delay can affect the responsiveness of the game and, therefore, the responsiveness of the game at different level of presentation delay is tested using the experience of real players. It is shown that a presentation delay more than 100ms is not appreciated, which implicitly means that a network delay has to be less than 100ms. Introducing a dynamic presentation delay is noted as the improvement to produce better result.

Read more!

Got through the proposal exam

It's been a long time since I made an entry in this blog. It has been a frustrating period editing the proposal report over and over. Also it has been a useful period in which I had nice discussion with prof about my future research directions.

Today I faced my proposal/candidacy exam and got through it successfully. The comments from the examination committee are:

• to publish more papers.
• to find present day applications that can benefit from RAN.
• to give more exploration on optimization algorithms.
• to analyze the convergence properties of the network positioning algorithms.

Now I am to work on a magazine paper on LAP and look for the target applications in the process of writing the paper.

Read more!