I've been putting 3.0 through a series of tests to see if we can use it as a replacement for MapObjects. Everything has gone well until I began comparing the performance of shapefiles to SQL Server. Below I'll briefly describe my test and hopefully someone can tell me if the performance can be improved.
I'm comparing the post-zoom redraw times to shapefiles, basically going from a scale of 1.6M to 800K, etc. and stopping at 700. I'm using four shapefiles of road features, and none of them is terribly large. The biggest layer contains local roads and has about 70,000 records. On the other extereme is the highway layer with about 700.
I have SQL Server 2008 Express running on my machine.
The initial load takes 3 seconds for shapefiles and 12 seconds for SQL Server. This is probably OK when connecting to the server, etc. is taken into consideration.
The range of redraw times for shapefiles is .003 to 3.2 seconds. All but two redraws are < 500 ms and all but 4 are < 75 ms.
Initially the redraw times for SQL Server were 9.9 to 11.7 seconds. I discovered "GeoCaching" and this dropped the times to 1.9 to 2.8 seconds. I tried "IsExtentCached" upon initial load with no considerable difference, then at each zoom, but there was no notable difference either.
I tried tweaking the spatial indexing in Management Studio. I found the program that uploaded the data had set the extent to -90, 90, -180, 180, so I adjusted that to the extent of my data, under the presumption that the grids were so large that my entire dataset was in one and all the data was being retrieved. No difference. I bumped up the all of the spatial index settings and saw no difference. Finally I DELETED the spatial indexes and there was no change in the performance either. It seems after observing this that the indexes aren't even used.
So my question is this: can SQL Server performance get any better? At certain zoom levels I'm facing 2 second redraws for SQL Server WITH caching vs. 3 millisecond draws for shapefiles, and these scales are where our application is used most of the time, so this is not acceptable. We would like to get away from shapefiles, so getting reasonably equivalent SQL Server is considerable.
I appreciate any comments...
Allen Huber