How To Build A Performance Testing Stack From Scratch


(Heather) #1

@mfleming recently wrote a great article series on the Dojo called How To Build A Performance Testing Stack From Scratch. The first article in the series is available to all Club users. The articles that follow are available to Pro subscribers. I think it ties in really well with Performance Testing 101 for me. I jumped straight into learning how to use the tool rather than why I would use it.

Some questions in the first article really got me thinking about how other people go about such things.

Who would want to take action if a performance issue was found?

When you’re young and naive (as I once was :stuck_out_tongue: ) you automatically think the answer to this is “everyone surely”. When you get a bit further in your career you realise that’s really not the case.

My first question:
In your experience, who would want to take action if a performance issue was found?

My second question relates to the second section of the article:
In your own experience, what have you found to be useful metrics and what have you found to be less useful metrics to measure?

My final question:
Would you change or like to change how you share the results of your performance analysis?


Performance Testing Training Material
(Andrew) #2

Great, thanks Heather!

Looking to join the Dojo but this may have to wait until after christmas! I can see the courses you sent links to look really good, though.

Andy


(Heather) #3

You can access the first one with a Club account to get an idea of the course :smiley: There’s so much coming to the Dojo in the new year!


(Mohamed) #4

Hi Heather ,

here is my thoughts about your questions :slight_smile:

In your experience, who would want to take action if a performance issue was found?
A : Performance results and issues should be shared with the stack-holders but the stack-holders themselves are variant according to the performance testing position during the dev cycle and how critical the issue is.So the ownership could be a developer , team lead , product owner or CEO.

In your own experience, what have you found to be useful metrics and what have you found to be less useful metrics to measure?
A : Response time , Transaction response time and error % are always useful and important.We can add other metrics like “Standard Deviation” , throughput and avg.buytes as indicators to system behavior.

Would you change or like to change how you share the results of your performance analysis?
A: I think that the test results content should be different according to whom going to read it , for technical people all possible results , metrics and reading should be exist.But for less technical people graphs and conclusion should be more readable for them.

Mohamed


(Chris) #5

I agree with Mohamed’s comments here, though for Q1, I’d add that it depends on the root cause. The overarching system I’m working on has applications integrating with service layers and databases. If I simply said there was a performance issue, then owners of other parts of the stack wouldn’t take notice. I would need to demonstrate which part of the stack I believe has the bottleneck. Once I can prove that, then I would go to the responsible group. To prove which part of the stack is the bottleneck is the tough part. I’d need to think about how the end consumer users the customer facing system and time how long certain operations take. Then I would try to mimic those key operations at each layer in the stack. E.g. API Calls and db calls. I could then use the original timing with each layers timing to show the overhead of each layer and show which layers takes a disproportionate time.