API Builder Performance Metrics

Here, we define a baseline performance metric for API Builder to help users with capacity planning or footprint estimations for deployments.

The figure we supply is an indicative baseline for the most basic sample flow: GreetFlow. We will use this figure as an upper bound for the performance you can expect from a running container.

The performance of a flow depends significantly on what the flow is doing. Connecting to external systems will be at the mercy of their response times and network latency, and complex data transformations (e.g. large XML transformations) will affect transaction processing. For this reason, we always recommend that you run your own performance benchmarking tests. You can find the tooling we used in the Apache Benchmark section at the end.

Test environment

  • MacBook Pro Mojave; 16 GB Memory; Intel Core i9 2.3 GHz
  • Running single instance of API Builder with logging disabled, and authentication set to "apikey"
  • Running apache benchmark (ab):
    • Keep-Alive enabled
    • gzip enabled

Performance metrics

Date of test November 2020 (Agra)
Number of requests 100,000
Concurrency 10
Total time 115.498 seconds
Bytes transferred 61,000,001
Time per request (mean) 11.550 ms
Time per request (mean, across all requests) 1.155 ms
Memory idle

12 MB

Memory max 12 MB
Requests per second 865.81

Apache benchmark (ab)

Apache benchmark (ab) was used for these performance tests.  This site explains how to install it.  To run:

$ ab -n 100000 -c 10 -k \
 -H "Accept-Encoding: gzip" \
 -H "Accept: application/json" \
 -H "APIKey: <key>"

Related Links