Due to the lightweight nature of browser tabs, using them as containers provides a significant performance boost over traditional serverless architectures, such as those that use Docker or similar containers.

Below are the benchmark results of running a function on Browser Functions, vs running on Apache OpenWhisk, on the same hardware (MacBook Pro 2015). This is a baseline comparative test, running a trivial JavaScript function on the two platforms. Other tests with different workload or test profiles may reveal different performance results. Apache OpenWhisk was installed using this method, a production installation may perform differently. The blue line (braas) represents the performance of Browser Functions, vs the orange line for Apache OpenWhisk.

Response time

Response time in milliseconds - lower is better

Concurrency

Concurrent requests response time - lower is better

Throughput

Throughput - higher is better

Web Assembly native code execution

The following function for calculating the Fibonacci sequence was compiled to a native binary (using gcc with default settings), and to Web Assembly using emscripten, and compared against a JavaScript implementation:

int fib (int n) {
  if (n < 2) {
    return 1;
  }
  return fib(n - 2) + fib(n - 1);
}

Benchmark results for running fib(42) on a MacBook Pro 2015:

Implementation	Execution time
JavaScript	5.778 seconds
GCC binary	1.746 seconds
Web Assembly via Browser Functions	1.564 seconds

Note that the Web Assembly version running through Browser Functions is faster than running the natively compiled binary from the command line!

GPU acceleration with WebGL

It is possible to utilize GPU hardware acceleration by making use of WebGL. Some JavaScript libraries already do this, for example TensorFlow.js or GPU.js. Below is a comparison of multiplying two 512 x 512 matricies using CPU vs using GPU via WebGL: