I was wondering if there are baseline characteristics based on the number of nodes/links in a diagram. In my testing so far I’ve found some wildly different numbers based on a few things.
For instance, when I create a new diagram with 243 nodes and 292 links (layered digraph) it takes about 2.5 minutes once the data is returned to the client. When I change it from a layered digraph to a force directed diagram it takes about 20 seconds.
Now, when I show a minimap and change it from force directed to digraph it takes about 5 minutes.
When I had a diagram with 458 nodes and 507 links the amount of time to manipulate the graph locally was considerably greater.
I’m still using 1.2.8 and I’ll upgrade the version shortly but is there anything I can use to compare against? Or what the dependencies may be (for example in most of these there is a template map with four different node templates - what’s the performance impact, etc.)?
Thanks.
LayeredDigraphLayout is significantly slower than the other Layouts. That’s fundamental to the algorithms – it’s true on other platforms too.
Unfortunately it’s particularly slow in JavaScript, so where your size diagrams shouldn’t be a problem in .NET (GoXam or GoDiagram), it sounds too big for JavaScript.
You could try fiddling with the LayeredDigraphLayout properties. I don’t know what kinds of graphs you have, but you could start experimenting with the packOption property to see if any combination produces acceptable results faster.
You could also try TreeLayout, which lays out in layers, but is intentionally not too smart about multiple parents. It is very fast and might produce OK results for your kinds of graphs.
Thanks Walter. I was able to make the tree view work for me and the re-draw performance is almost an order of magnitude better. I’ve got a layout now that isn’t perfect but the performance is great.
Much appreciated.