Performance of multiple featureclasses in 1 sdf vs multiple sdf's is more than 10x faster

[ronnielouie]

Create feature source where there are multiple feature classes in one SDF. Also create feature source of the same data where there are multiple SDFs, (one for each feature class).

(A package containing data set to reproduce this is available, but cannot be attached to ticket at this time because it exceeds the attachment size limit.)

The slow folder contains each layer published as an individual sdf whereas the fast folder is all layers published to a combined SDF. In both cases there are 28 feature classes.

Preview both maps.

Refresh both maps and notice that the fast one based on a single sdf loads in about 3 to 5 seconds whereas the one based on the 28 sdf's takes over 50 seconds.

[jbirch]

I wonder if this is related to #175 ... some kind of systemic slowness in reading schemea.

As far as priority goes, these are both important performance issues, but are at different priorities. Can someone rationalize them?

[ronnielouie]

Fixed by ​http://trac.osgeo.org/mapguide/changeset/2111

[jbirch]

The comment on that changeset appears to describe the inverse situation of this ticket. Was the description of one of them incorrect?

[ronnielouie]

The original defect reported that single file was faster than multiple files. That is incorrect, as further investigation saw that multiple files was faster then single. The comment in the changeset is correct and accurately describes the results after implementing the fix.