I have a cached map service that I am using in a Silverlight based application. For this cached service, I have cached a core area and checked the option for On Demand Cache. I find that the areas within On Demand Cache are very slow. I am trying to understand how this process works. If I am zoomed in at the scale of 4000, it should be generating cache for a very small area, not sure why it takes much longer. I am using ArcGIS online tiling scheme. Does it generate the cache for all scale levels or just the one level.

I am looking for some feedback.

Thanks Jay

  • Hmmm... It's about time you accepted some of the great answers you've gotten to your previous questions, like this one. All you have to do is tick the checkmark icon next to the best answer: that makes it more prominent to future readers.
    – whuber
    Aug 10, 2011 at 22:34

2 Answers 2


It only generates the cache for one scale level, but as described here, the caching engine only works on a minimum "supertile" of 2048 x 2048 pixels, which if you're using the Google/Bing Maps tiling scheme is 16 x 16 tiles. If you're at the intersection of several supertiles, the server could potentially be generating 1000 tiles from one request, even though you're only seeing 10-20 of them in your application.

  • What I do most often to combat the slowness issue is to pre-generate areas most frequently visited by users. This may not always be possible however, but it is extremely effective when it is possible. If for example you're displaying a state, and you are fairly sure your users will most often be viewing the metropolitan areas, you can pre-generate the tiles for those areas at the required scales. Aug 11, 2011 at 14:51

Bringing back an old question for a quick update; cache on demand performs very poorly.

This is still an issue in ArcGIS Server 10.3.1 and the problem extends back to at least version 10.1. There is a bug report about this issue here: http://support.esri.com/en/bugs/nimbus/QlVHLTAwMDA4Njg1Mw==

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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