I need to know how would I see trend of a service? How would I know if a service has improved in a month? If I run the script every Sunday of a month then how can I use weekly measurements to say service is faster or slower? For example, I am interested in waiting time and blocking time of a request against a service. How can I calculate a number from a timeseries measurements? Maybe the bar chart of this calculated number can give me trend of performance of a service.
I use Datadog for logging metrics
I would guess you will just make a dashboard(s) to compare them, the same way you will do it for any other metric.
Is there something specific that you think won’t work? (obviously, if it’s datadog specific, it’s better to ask their support as they will likely be better equipped to help you).
I guess the biggest problem will be the fact that unlike most other measurements you will have huge(week-long) holes in your data, but you can just zero this out and have 2 dashboards with different time intervals shown. So it will look something like the cloud test comparision.
If you want just a single number … I guess datadog has away to do as well, I have seen dashboards doing it, but I am not versed in datadog - again their support will likely be able to help you a lot more on datadog specific questions.
On the other hand just because (for example) the p(95) of something is lower this week, doesn’t mean that it’s all better, maybe the median has raised 10x and the p(99) is also huge just the p(95) is down, so you will need to figure out what is important to your case and monitor it.
@mstoykov
I am new to performance testing so there are unknows in my understanding. Maybe K6 should write a blog on how to get maximum out of K6 after getting the measurement.
Half of my problem statement is Datadog specific and half lack of understanding… I figured out Datadog side of my problem. I am using change graphs so I can map last week run data with current week run data. Also, in the dashboard I am averaging out the measurements to a single number.
Thank you for mentioning about this:
On the other hand just because (for example) the p(95) of something is lower this week, doesn’t mean that it’s all better, maybe the median has raised 10x and the p(99) is also huge just the p(95) is down, so you will need to figure out what is important to your case and monitor it.
I will look into it.
