So we have data with over 10,000 features in a very small area of less than 50 acres or so and we are getting pretty slow performance of several seconds when the extent is changed via panning and zoom in/out on the client. We are publishing these as dynamic layers using ArcGIS Server 10.2 and displaying in a ArcGIS JS web app. Is there something we can do to improve performance and not show all features at once? Somehow intelligently dissolve the data so the UI isn't tied up when the extent changes on the client. Let me know if I need to explain this any better.


if possible, my recommendation would be to work with your FeatureLayer in selection only mode and display the ArcGISDynamicMapServiceLayer instead (as in this example)

  • Thanks John. So you feel that the Dynamic Map service should have no problem displaying over 10,000 features within an extent as the user pans the map and not tie up the browser UI?
    – wilbev
    Oct 27 '14 at 17:16
  • performance will be improved dramatically by asking the server to draw the features and return an image rather than passing raw geometry and attributes over the wire. Oct 28 '14 at 18:09
  • John, I should have been more clear but were not using a feature service at all here, only the dynamic map service. It's the dynamic map service that is taking several seconds to render when the extent is changed. I guess due to so much dense data it's causing the image from the server to take that long I guess? Does that sound right, more point data the dynamic map service is created from causes the render performance to decrease a great deal?
    – wilbev
    Oct 29 '14 at 19:35
  • 1
    thanks for clarifying. seems like a good reason to use a scale dependency and not draw the layer until people zoom in closer. if thats not a possibility and the data is in SDE, it might speed things up to use a SHP as the resource in the service itself. Oct 31 '14 at 17:55
  • The problem with scale dependency is we still want to give the user the symbology of data as they zoom in/out vs just turning off completely as they zoom out. It would definitely be fine if the symbology was greatly generalized as they zoom out but don't know a way of doing that. Yes, SDE is being used. So if we used SHP instead of SDE as the workspace type we may get better performance? File gdb likely better as well?
    – wilbev
    Nov 2 '14 at 17:28

I've never used this feature myself, so I can't give details, but you might want to try the point clustering feature:


This allows you to display several point features as a single feature (symbolized to denote the number of features that were grouped) when zoomed out, but switch back to displaying individual features when zoomed in closer.


Maybe you could try use sdegroup command i.e. merge features as described here. You could use merged features layer only at zoom levels that are slow to display i.e. huge number of features is fetched.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.