基于时空热度的分布式金字塔瓦片协同存储机制研究

Pyramid tiles have been widely used for online mapping, and the performance of the storage directly influences the quality of map service. Online map accessing has certain group-user access patterns, such as temporal periodic access rules and spatial hotspot area. In this study, tile storage is built based on SSD with RocksDB, and it serves as the cache of the Cassandra tile storage based on HDD. Through mining of tile request logs, maps’ search logs and GPS tracks, the temporal periodic access rules and the spatial hotspot areas in group-user access pattern are discovered. The discovered group-user access pattern is then used as the dispatch policy and replacement algorithm of the cache. The key research questions include:.a) The architecture for collaborative Storage of Pyramid Tiles based on spatiotemporal popularity;.b) Multivariate data mining for spatiotemporal hotspot areas discovery;.c) A cache replacement algorithm that integrates both storage capacity and hotspot value;.d) The adaptive policy of dispatching hotspot and cold-spot tiles..As spatiotemporal popularity is used to improve the performance of tile storage, the quality of map service will be greatly improved. The proposed work with a collaborative storage with dual-layer prototype system will be developed experimentally as the basis for future deployment in real map applications.

金字塔瓦片技术已广泛应用于在线地图中，存储的性能直接影响地图服务质量。在线地图访问存在明显的倾向性，具有明显的群体行为模式，体现为时间维度上的周期性访问规律和空间维度上的热点区域。本研究利用RocksDB建立基于闪存的瓦片存储，作为Cassandra瓦片硬盘存储的高性能缓存；通过对瓦片日志、地图检索历史和GPS轨迹数据的挖掘，发现群体行为的在线地图访问规律和热点区域；探索将在线地图访问规律和热点区域应用于瓦片缓存调度和置换方法。主要研究内容包括：①基于时空热度的分布式金字塔瓦片协同存储架构；②多元数据融合的时空热度挖掘方法；③容量与热度结合的冷热瓦片置换方法; ④自适应冷热瓦片调度策略。研究将时空热度应用于构建瓦片的缓存，有助于改善存储的性能和地图服务质量。研究构建了金字塔瓦片协同存储，并尝试部署于实际应用环境以验证研究成果的有效性，为进一步成果转化奠定基础。