基于访问相关性的空间数据副本控制策略

Replication strategy for spatial data based on access correlation

为解决地理信息系统面对用户大规模密集访问时由于缓存空间有限且热点数据较多,简单的根据数据的流行度选择副本难以适应用户突发性访问行为等动态需求的问题,提出一种综合考虑数据流行度及其相互关系的副本控制算法.首先,根据数据流行度确定热点数据集以减少热点数据规模和降低计算开销;然后计算热点数据之间的相互关系以得到其访问相关性;最后利用数据之间的访问相关性动态选择与当前正被访问数据相关性高的数据作为副本,从而为用户的下次访问提前准备数据,实现降低用户访问平均响应时间的目的.算法实验表明:相比于传统方法,所提出的综合副本控制算法可有效提高5.9%~29.9%的平均访问响应速度.

Due to the limited high-speed caching space and the massive dataset, replication strategy based solely on datars popularities cannot work when usersr access behaviors change suddenly, a com- prehensive replication strategy used both datars popularities and datd relationships to select replicas was proposed. First, hotspot data were selected based on their popularities to reduce the size of data and then reduce computational overhead. Then, access correlations were computed based on their re- lationships. Finally, some data could be selected as replicas and stored into high-speed caching system according to their access correlations with the data being requested, so as to prepare the next data for users in advance and to reduce average request response time. Experimental results show that the pro- posed comprehensive replication strategy can achieve a lower average request response time than some other algorithms by about 5.9% ~ 29.9%.