Dynamic load balancing for the Web Map Services through "Query partitioning for the variable sized and un-evenly distributed data".
In my previous post I had basically investigated the problems of distributed map processing and rendering of scientific data and information, and the integration framework of Map servers with the Geo-Science Grids from the interoperability, scalability and performance points of view.
Load balancing and the caching are the first methods coming to mind in order to enable scalability and good performance results. However, the implementation is not easy for the interactive geo-science applications due to the nature of the spatial data. Partitioning of feature data (represented in sets of polygons, line strings and points) is hard due to the spatial nature, varying sizes of the feature collections, and uncertainty about the query location. We also do not know the workload previously. The work is partitioned into independent work pieces, and the work pieces are of highly variable sizes. It is not possible to estimate the size of total work at a given server.
Here in my *draft* document I tried to solve these issues and give a summary of the proposed load balancing algorithm composed of query partitioning through caching and critical regions.
http://complexity.ucs.indiana.edu/~asayar/proposal/loadBalancing.pdf
No comments:
Post a Comment