We're in the midst of restructuring our SDE Schema and have been asked by our IT department to remove feature classes that do not participate in a Topology out of their current Feature Dataset, as it is affecting the performance of the database, especially when editing.
Can anyone help quantify what the perfomance increase we can expect to see once we move the dataclasses out, and they essentially become stand-alone feature classes?
The current Datasets not only contain Topology, but are also used to group similar feature classes together to make things easier to find in the Catalog.
Based on my experience, I've seen no noticeable difference in the performance and I've never really noticed if there was any difference. And although feature datasets are convenient to group similar featureclass (much like data folders) you actually lose a bit of functionality when it is used this way. Example: Editing a single featureclass in the dataset will lock everything in the dataset....
Thanks Mike. As we work in a versioned environment I'm not sure that 'locking' everything in the dataset is really an issue.
Good to hear you have not seen a performance impact.
there will be no meaningful impact on performance.
I too have been asked to use FDs in this way and it always takes a bit to talk folks back from the ledge of using them to organize features. One thing i have done to achieve the visual grouping of data themes in ArcCatalog is prepend the theme to the data set name.
Removing feature classes out of feature dataset would not help much to improve the performance. After all feature dataset is just a logical entity (unlike feature class which exists physically) intended for grouping of feature classes for various purposes. Following link has all the details:
As far as locking and feature dataset is concerned,
" If you acquire a lock on a feature class within a feature dataset, the lock applies to the entire feature dataset and its contents."
Details about locking can be found on :
Hope this would shed some more light on this....
This is the nth time I've heard this: Feature Classes in a Feature Dataset decrease performance. I'm wondering where (IT) folks are hearing that? However, I'm not debunking the notion that it happens...like others...I've never seen it. If you look at a SQL trace of a draw request of something that is in a FD, you do see some additional join statements. I've never detected any performance difference, even when looking for it with PerfQA.
I think I may post a poll on this topic....