Re: [yangtools-dev] [controller-dev] Controller performance and plugin-to-plugin interactions

Robert Varga

On 03/20/2015 10:09 PM, Luis Gomez wrote:

The fix you are talking is in CBench itself? we know CBench has some limitations like single thread but that has never being a problem for other controllers to show much higher numbers than ours.
I seem to remember some issues with how it accounted for the various things we send out irrespective of what is happening, like statistics collection. I may wrong, though.

Again, from memory, I think it essentially round-robins bursts from individual simulated switches. That can cause the controller to fall in-sync with how those burst occur, which is not really representative of a real network, so some improvement in that area would be nice. But this is outside of my area of expertise, really :)


Join to automatically receive all group messages.