Concurrency Control for Intermittently Connected Client-Server Databases

• Main Authors: Lee Ying-Ching, 李英菁
• Other Authors: LihChyun Shu
• Format: Others
• Language: en_US
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/16256102224110493502

碩士 === 長榮管理學院 === 經營管理研究所 === 90 === As the price/performance ratio of mobile devices continues to improve steadily in recent years, it is now common for people to cache data on their mobile appliances and work on local data copy while they are away from home or o±ce. However, when people began to modify replicated data, maintaining consistency of data copies at various locations becomes an important issue. In this thesis, we address this problem from the perspective of the intermittently connected client-server database environment. In this environment, clients are disconnected most of the time. While disconnected, clients can update replicated data by way of tentative transactions. These updates will remain tentative until being validated when the clients reconnect. We propose two concurrency control protocols mainly for the purpose of update validation. While both protocols guarantee global serialization, the first protocol, called the graph-based scheduler, physically maintains the serialization graph and the second, called the match and go protocol, does not. We have designed the graph-based scheduler in such a way that the serialization graph can be e±ciently maintained (both in construction and reduction). In addition, minimal information will be required to be transmitted from a client to the server when the client reconnects. While the server assumes most workload in the graph-based scheduler, it is the reconnecting client that performs the update validation in the match and go protocol. Hence, the match and go protocol is more scalable than the graph-based scheduler. Another aspect that distinguishes these two protocols is that the match and go protocol requires that each mobile client always uses recently changed data (by other clients). The graph-based scheduler, on the other hand, allows a mobile client to compute a data based on a not-up-to-date value of another data. The set of histories that can be generated by the graph-based scheduler, denoted Gen(GBS), has a nonempty intersection with the set of histories that can be generated by the match and go protocol, denoted Gen(MAG). However,Gen(GBS) is incomparable with respect to set inclusion with Gen(MAG).