Multiplayer Online Digital Games (MODIGs) are gaining in popularity because of the strategic sophistication added when games are played against other humans, as opposed to computer artificial intelligence (AI) opponents. However, the actualisation of multiplayer games is not easy, due to their complexity. Multiplayer games are the combined applications of various areas, such as networking, graphics, AI, sound, and process optimisation. Among them, problems related to networking -- such as limitations in data transfer rate, latency, and jitter -- are the most difficult to resolve. Network latency cannot be avoided completely and introduces various problems such as inconsistency of player status, recognisable responsiveness, and irregular network lag.

Generally, the network architectures of MODIGs can be categorized into three groups: Client-Server (C/S), Peer-to-Peer (P2P), and hybrid. In general, MODIG designers prefer the C/S network architecture to the P2P system. The main reason for this is that the C/S model enjoys certain advantages, such as simplicity of consistency maintenance, improved security, efficient authentification, and ease of billing system management. However, the C/S architecture can cause network latency and often servers do become network bottlenecks. To solve this problem, server clustering methods are used, but these solutions may not be cost effective. This is why a number of games use the P2P network architecture, but in this structure the total number of players in any one game session is often limited, because of the network’s bandwidth constraints. In addition, the consistency maintenance issue becomes critical within this architecture.

There are two main approaches for maintaining consistency in MODIGs: conservative and optimistic. The former approach involves a send-and-wait philosophy, requiring acknowledgement frames and resulting in packet transfer delay, such that players may experience network latency. In the latter approach, the processes do not wait for other players' packets and advance to their own frames, thus no network latency occurs. However, when there is inconsistency between players, the processes must roll back to correct mis-ordered operations due to packet transfer delay. These can cause irritation and confusion to players, and thus the quality of game deteriorates. Overall, the optimistic approaches may not be suitable for network games.

To alleviate network latency and reduce bandwidth requirements in conservative consistency maintenance algorithms and the P2P-based approaches, a new system is proposed and designed. To reduce network latency, a conservative consistency maintenance algorithm named Locked Bucket Synchronisation (LBS) is proposed to mask latency and maintain perfect consistency among players. In addition, a distributed network architecture is adopted and a tree-based P2P system is proposed, to remove additional packet transfer delays between server and client. To reduce network latency caused by packet drop and delay, a smart transmission scheme (STS) is proposed. To alleviate the bandwidth problem, a packet aggregation method is introduced. These approaches are thoroughly examined and analysed.

To evaluate the efficiency of the proposed system, a network simulator, COMP2P, and a real network game, Duel-X, are implemented. The efficiency of the proposed consistency algorithm, LBS is compared with that of other approaches such as Lockstep (LS) and Frequent State Regeneration (FSR). The system architectures of COMP2P and Duel-X as well as the results of experiments are fully documented.

The experimental results from COMP2P show that the proposed LBS algorithm outperforms the LS algorithm under all tested circumstances, in terms of game execution speed. The LBS algorithm with the tree-based P2P system achieves almost optimal frame rate under the condition of a 10% packet drop rate, while the LS algorithm with the tree-base P2P system performs at approximately 40% of optimal frame rate.

The efficiency of the Smart Transmission Scheme (STS) with the LBS algorithm is compared with the Blind Transmission Scheme (BTS) and the experimental results indicate that the BTS increases the bandwidth requirement of players by 100% while the STS raises by only 10% the optimal value of the bandwidth requirement without degrading game execution speed significantly.

Finally, the proposed packet aggregation method together with the LBS algorithm reduce by 50% the bandwidth requirement of players, without lowering game execution speed, when compared with the case of using the LBS algorithm only.

To verify and validate the efficiency of the LBS algorithm, various experiments are performed on Duel-X. The experimental results show that the LBS algorithm masks network latency without harming consistency between players. When the LBS algorithm is adopted, the rate of frame interval approaches the optimal values that can be achieved under the FSR algorithm. This implies that the LBS algorithm improves the playability of MODIGs without degrading consistency between players.

The overall experimental results show that actualisation of reliable and robust MODIGs is achievable with a combination of P2P-based architecture and conservative consistency maintenance algorithms.