摘要
In recent years, several results have been introduced to enhance distributed GIS performance. While much more efforts have focused on tile map and simple symbologies on dynamic map, load balancing GIS servers have not been addressed by the GIS community so far. This paper, therefore, proposed dynamic distributed load balancing for D-GIS in order to quickly render information to client interface by involving a set of GIS servers which process clients’ requests depending of an algorithm. In the model, several concepts were introduced and defined: Virtual Server within physical machine which constitutes a setup environment for a single GIS server, Load Hash Table which contains information about virtual server’s capacity, real-time load and other mandatory elements, Request Split Table which splits requests depending of the input area’s Quantity of Information and stores request tasks composition for later reconstitution. At last we have Distributed Failover Callback Function Table level one (respectively level two) which determines whether or not the request had been successfully processed by the chosen virtual server (respectively physical machine). This table allows sending back the same request to another virtual server (respectively physical node). Two load handlers (primary and secondary) are defined in case of failure. Our Model achieves efficient load balancing by: providing efficient node selection;optimizing request routing;managing node failover;involving client’s request partitioning and introducing method type decomposition. A simulation of the algorithm shows a low response time when performing GIS operations.
In recent years, several results have been introduced to enhance distributed GIS performance. While much more efforts have focused on tile map and simple symbologies on dynamic map, load balancing GIS servers have not been addressed by the GIS community so far. This paper, therefore, proposed dynamic distributed load balancing for D-GIS in order to quickly render information to client interface by involving a set of GIS servers which process clients’ requests depending of an algorithm. In the model, several concepts were introduced and defined: Virtual Server within physical machine which constitutes a setup environment for a single GIS server, Load Hash Table which contains information about virtual server’s capacity, real-time load and other mandatory elements, Request Split Table which splits requests depending of the input area’s Quantity of Information and stores request tasks composition for later reconstitution. At last we have Distributed Failover Callback Function Table level one (respectively level two) which determines whether or not the request had been successfully processed by the chosen virtual server (respectively physical machine). This table allows sending back the same request to another virtual server (respectively physical node). Two load handlers (primary and secondary) are defined in case of failure. Our Model achieves efficient load balancing by: providing efficient node selection;optimizing request routing;managing node failover;involving client’s request partitioning and introducing method type decomposition. A simulation of the algorithm shows a low response time when performing GIS operations.
