| Summary: | 碩士 === 國立臺灣科技大學 === 資訊管理系 === 88 === Data dissemination is an effective communication approach in an asymmetric communication environment because of its scalability in terms of the number of users it can serve. In most of the current researches, the use patterns of all the related users are collected as a clue for designing the broadcasting schedule of the disseminated data. It is shown that when the broadcasting frequency of each data item is in proportional to the square root of its reference frequency, the mean access time for the users is minimized.
While the optimal schedule for data dissemination is attained in the previous researches, the update of the disseminated data is not considered. In most of the practical applications, data are usually changed frequently, for example, the stock prices in the stock market are changed within minutes. In this thesis, we study the optimal broadcasting schedule in an asymmetric communication environment where update of the broadcasted data is allowed.
In this thesis, we propose an architecture in which updates on data are recorded in an invalidation report. Periodically, the invalidation report is broadcasted to the clients. When a client receives the broadcasted data, it will wait until receiving an invalidation report. Subsequently, the client use the invalidation report to exclude the data that is updated, leaving the up to date data for its latter use. According to an analysis, we show that the optimality of the mean access time is attained when the broadcasting frequency of a data item is in proportional to the square root of its reference frequency over the probability of successful validation (of this data item). This analytical result is verified with an extensive simulation.
|
|---|
