CODEWITHSUNDEEP
performancetiminglatency
Protector one
Protector one

Reputation: 7261

Performance timing: average, minimum or something else?

When timing my (web) application for performance/latency, should I use the minimum time measured in n runs, or the average? Or something else?
If it's the latter case, when to use what?

If your answer is going to be something along the lines of, "research it, dude", could you point me to a good resource?

Upvotes: 1

Views: 117

Answers (4)

Protector one
Protector one

Reputation: 7261

http://www.webperformancetoday.com/2012/02/13/non-geeky-guide-to-performance-measurement/ This is a good article on "response time" timing. (Websites.)

Upvotes: 0

user425367
user425367

Reputation:

I would say you have to figure that one out your self. Are you benchmarking for

  • Request throughput
  • User experience
  • Checking timeout thresholds

Depending on what you want your analysis to answer different metrics should be used.

Upvotes: 0

Dan D.
Dan D.

Reputation: 74655

Use median rather than mean (or average).

For reasoning, see the page Mean Delay Considered Harmful by Stanislav Shalunov (the author of thrulay).

Upvotes: 1

unludo
unludo

Reputation: 5010

You could retrieve the values, store them by time, with couple (time_of_the_call, response_time). Then you may treat these data with tools, draw graphs, make statistics. I think an average, minimum, etc, is not sufficient, you need a set of measures.

For example you may put you data in a csv file and import in excel, or even use google graph api to draw real time graphs.

Upvotes: 1

Related Questions