seeking entity (quantity) limit guidelines?

Are there any guidelines (published or working) for the population limits or entities? We are working to establish some boundaries that will influence design choices for the distillation and presentation of data within Cesium.

I have experimentally allocated entity points until [browser] failure [using chrome]. Ditto polyline segments, and polygon rectangles. These experimental results help, but we would welcome further feedback for these limits from Cesium’s design perspective…

Thank you,


Hi Kirk,
We’ve managed a few thousands (up to around 20,000) - depending on the machine running cesium on.

According to Patrick here:!topic/cesium-dev/PHUVh0p3NlU, 5000 entities is a lot.

I’m surprised myself I got to 20,000 with minor optimization, but then again, that was on a very strong machine. Older machines (i.e. i5 with only 2GB of vCard memory) had a real hard time handling 5000 and more. Most older machines crashed before that.

Would be happy to know your numbers as well (as well as machine spec) so we could gather community experience on that issue.

First, the answers. Then, the rationale.

Using a healthy desktop with 16GB memory, and an nVidia GT720 card, I was able to map over 40,000 point entities. As expected, they drew as they were created within a tick callback. My experience with more complicated structures was quite the opposite. Polyline segments and polygon rectangles both failed between 2000 and 3000 entities.

Why is this a concern? “Big data”. Some folks want to plot oodles of stuff. But it appears that a distillation process may be required in order to keep the entity quantity manageable.

Yea, my entities are a bit complex. Each has a label, billboard and some
geometry (or geometries).
We're showing big data as well. I do have a clustering algorithm that might
help you. It's hard for me to get things out of the work computer since it
is not connected to the internet, but I could try and enhance it when I
have the time (or if you want, you can share your ideas).
Here's the explanation, code and examples for the clustering algorithm.
With it, you should be able to show more data.
The next step would actually be to use web workers to do the clustering.
Shouldn't be much of an issue, since the clustering algorithm already
exists, just need to the time to actually do it.