海量影像数据的发布集群系统与应用

海量影像数据的在线发布和共享是传统网络地理信息系统难以解决的难题,集群技术具有高度可伸缩、高可用、易管理和高性价比的特性,是解决海量影像数据发布难题的有效手段。本文在分析动态反馈负载平衡器模型的基础上,提出基于集群技术的在线影像发布服务及其体系结构,设计了区域分治法等四种不同复杂度的负载平衡算法,适于不同物理系统,最后给出影像发布系统的应用实例。

Disseminating Large-scale Geo-imagery Based on Cluster Technology

Conventional Web GIS has partially solved the problems of disseminating spatial data via Internet. However, few systems have the powerful capability to offer the high throughput and compute intensity of Largescale Geo-imagery Service. With the aid of geospatial information collecting technologies such as GPS and Remote Sensing, the real-time spatial data can be made available. Many organizations have built multi-scale heterogeneous massive spatial database of different levels containing high-resolution aerial, satellite, and topographic data. A cost-effective solution clustering multi low-cost commodity-class PCs is proposed, which has the characteristics of high-scalability, high-availability, easy-manageability and cost-effectiveness. This paper introduces the architecture of Large-scale Geo-imagery Service using cluster technology and describes the core component called load balancer, which chooses one of the four different load balancing algorithms to adapt to the physically deployed systems. Based on the self-adapted dynamic load balancing model, an agent-based feedback mechanism was developed to monitor service nodes' instant state and arbitrate which should serve the next coming request. We compared four inner approaches: round-robin scheduling, weighted performance scheduling, weighted least-load scheduling, and domain divided scheduling and found that domain divided scheduling method can improve performance dramatically for its high hit ratio cache.