12.07.2015 Views

Implementation of a Peer-to-Peer Multiplayer Game with ... - DVS

Implementation of a Peer-to-Peer Multiplayer Game with ... - DVS

Implementation of a Peer-to-Peer Multiplayer Game with ... - DVS

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

sensor node request messages <strong>to</strong> all nodes in the sensor node list, returning new sensor node suggestions.For each section, the nearest node outside the vision range is chosen as the new sensor node.When joining the network, the new node connects <strong>to</strong> some arbitrary node that is already in the network.That node provides information about nodes closer <strong>to</strong> the new node via a sensor node request.The new node then repeatedly sends sensor node requests, and its position updates are forwarded bythe sensor nodes, so that the node finally finds all <strong>of</strong> the neighbors in its vision range.EvaluationSystem quality is measured using the following pro<strong>to</strong>col quality model: PositionAge(p, q) is defined asthe time (in rounds) between player q sending his position update <strong>to</strong> p and p receiving the update.Within the interaction range (defined by the game application), the pro<strong>to</strong>col quality function PQ(p, q)between the players p and q equals the position age function. Outside the interaction range but <strong>with</strong>indist(p,q)−IR(1−vision range, it is defined as PositionAge(p, q) ) V R−IR , where dist(p, q) is the distance between thenodes. V R and IR are vision and interaction range. Outside the vision range it is set <strong>to</strong> 0. The pro<strong>to</strong>colquality PQ(p) for node p is the sum <strong>of</strong> PQ(p, q i ), and finally the <strong>to</strong>tal pro<strong>to</strong>col PQ quality is defined asthe average over all PQ(p i ).For evaluation, the system is run in a simulated network <strong>with</strong> up <strong>to</strong> 600 players. The default gameworld size is set <strong>to</strong> 1000x1000 <strong>with</strong> a relatively large vision and interaction range <strong>with</strong> radii <strong>of</strong> 200 and50 respectively. The node’s outbound band<strong>with</strong> is limited <strong>to</strong> 5KB per round. The experiments show thatthe pro<strong>to</strong>col quality does not depend on the <strong>to</strong>tal number <strong>of</strong> players but just on the density (<strong>to</strong> be moreprecise: the average numbers <strong>of</strong> players in the vision range). With 100 players, PQ is almost 1, which isthe optimal value. Client/server solutions have a PQ around 1.4 as all messages have <strong>to</strong> pass the serverand thus require at least two hops. Increasing numbers <strong>of</strong> players in the vision range require (linearly)increasing band<strong>with</strong>; a limitation <strong>of</strong> the latter reduces pro<strong>to</strong>col quality. But even <strong>with</strong> 50 players in thevision range, pSense is still slightly better than client/server.DiscussionpSense is a system based on locality (in the game world) that works completely decentralized. There isno global authority required and all peers have the same responsibilities. The evaluation results showan excellent performace as long as the game world does not get <strong>to</strong>o crowded. As many games have asmall number <strong>of</strong> important places <strong>of</strong> interest, this could though be a problem. One could think abouttechniques for dynamically decreasing the vision range (like VON) and/or trying <strong>to</strong> focus on importantneighbors (like Donnybrook).pSense only considers players and does not provide a solution for handling other game objects. Althoughin principle objects can be handled in a similar fashion as players, there are still synchronizationissues. The authors argue that synchronization and conflict resolution can be handled by a central serveras this is necessary rather infrequently. However, this is may not be a satisfying solution.2.4 Design Space for <strong>Peer</strong>-<strong>to</strong>-<strong>Peer</strong> <strong>Game</strong>sThis section concludes the related work section for peer-<strong>to</strong>-peer gaming by assigning the presentedsolutions <strong>to</strong> the different problem categories low-latency information dissemination, network maintenance,object management, and security. Clearly, not all presented approaches provide solutions for every <strong>to</strong>pic.Thus, a complete system needs <strong>to</strong> adopt parts <strong>of</strong> various <strong>of</strong> these approaches.2 Related Work 27

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!