May 28, 2018 at 11:44 am #19632@mange-petterssonOnly available when logged in
I’m in a tough situation regarding performance testing.
I wrk with a system where several nodes are displayed on a map and objects moving on this map may or may not get priority to these nodes. We have recently discovered that we have performance issues where the objects positions tend to queue up for the evaluation to get priority or not.
I’m now trying to look at some way to make a performance test to try and measure where we stand and how we improve on our capacity.
Since all the evaluation is done in a back end component, all I’ve come up with so far is to set up a test environment where we get all moving objects from ourt live environment (we already have this) and then increase the number of nodes until you see that units start to queue up, there we have status quo. When performance is increased we run the same environment and se how many more nodes we can add for the same units before positions start to queue up again.
As I see it it’s quite hard to make a fair measurement of this backend component in any other way and that’s why I write here: Does anyone here have an idea how this can be measured in any better and perhaps entirely automated way?
I’m grateful for any tips and tricks.June 4, 2018 at 5:30 pm #19692@rpwheelerOnly available when logged in
What I can suggest:
- The communication between foreground and background have to be done with some messages system. If you trace what messages are and how they go through the system, you may find idea how to mock them.
- I wrote some click-acting scripts using SikuliX . One of them, back in 2012, was able to click buttons on slowly moving images to open articles in a news app and go back. I think something can be programmed to manipulate the nodes, and using techniques like pseudo-random numbers generation generate predictable chain of events to act on the system same way every next time.
You must be logged in to reply to this topic.