A Synchronization Method for Distributed Interactive Applications

Project BackgroundDistributed Interactive Applications (DIAs) refer to applications that involve human and machine interactions over a distributed environment. In a typical DIA, where machines are connected by networks, latency usually exists in transmitting messages from one machine to another through networks. In some less demanding DIAs, such as the ATM machines used in banks, a user is typically given a rather long time to react to the information presented to him/her. While waiting for the user’s next instruction, the information presented does not change. However, in some more demanding DIAs, such as online games, the information presented to the user changes continuously and the user needs to react to it at the very moment when the information is presented. Unfortunately, if the information comes from a remote machine, because of network latency in delivering the information, the user may be reacting to an outdated piece of information. If the information is generated from multiple remote machines, some part of the information can be more outdated than the other due to the differences in network latency among the machines. In a multi-user environment, different remote users may also be seeing different views of the information. These problems are generally referred to as the synchronization problem. A typical solution to this problem is called “rollback”. When the system detects that a user has reacted to an outdated piece of information, it undoes the operation. However, when the application involves more than one user, this rollback action may cause disputes among the users. There are many methods developed to address the synchronization problems. However, most of them are for the less demanding type of DIAs. Very few methods are developed for the more demanding ones. Some of them either expect the users to adapt to the latency or simply delay the local actions to minimize the discrepancy.Project Objective and SummaryIn this project, the researchers propose to develop a distributed synchronization method to address the synchronization problem of the most demand type of DIAs, such as online games. As the information in this type of DIAs is changing continuously, it is not possible to ask the users to wait for the synchronization of each information update before reacting to it. Our distributed synchronization method is based on a relax consistency control model, which does not impose strict consistency. Instead, it allows the trajectory changes of each object to deviate from the correct ones by an acceptable amount. The new method is composed of a number of techniques, including thelocal-time adjustment scheme, thetrajectory adjustment scheme,and theself-tuning dead reckoning scheme. The local-time adjustment scheme modifies the local time in order to align object movement. The trajectory adjustment scheme modifies the local object moving trajectory in order to minimize the object trajectory discrepancy. Finally, the self-tuning dead reckoning scheme dynamically adjusts the dead reckoning frequency to maintain an alignment of the object among remote machines.