Taurus - API Load Testing

Question

We are using the Taurus for API load testing and would like to know what is the possible configuration one should consider for increasing RPS limit?

As per Taurus documentation, To calculate the RPS, we should use the formula.

RPS = Concurrency / (Ramp-up + Hold-for)

But I couldn't get pass the threshold beyond 250 RPS for given configuration

execution:
  - concurrency: 600
    ramp-up: 40s
    hold-for: 10m
    scenario: api_scenari

Answer 1

I can think of several possible reasons:

1. You simply don't supply enough threads (virtual users), try using i.e. 1200 instead of 600 and see if the RPS gets twice as high.

2. Taurus is not sending requests fast enough due to lack of resources. Take a look at local monitoring widget and see how does CPU, RAM, Network, etc. usage look like. If it turns out that your machine cannot send requests fast enough you will need to switch to distributed testing

   

3. The server running API is overloaded and simply cannot respond faster. In this case you can either report it or investigate the root cause of the bottleneck.

More information: How do I Correlate the Number of (Concurrent) Users with Hits Per Second