SSR Experiments

Experiments for SSR-ing off the main event loop.

Usage

$ yarn
$ yarn benchmark

Overview

We provide various "implementations" (e.g. a worker pool) that can invoke "scenarios" (e.g. a React SSR) with various parameters. Here's a quick outline:

• benchmark/scenarios:

  • noop: A simple setTimeout to just check that all the mechanics work. Concurrent workloads are all easily parallel.
  • react: Run an SSR render with random attributes. Input is repeat for how many elements to create. Does not currently peg all CPUs, but might be an aspect of the workload given all the string concatenation...
  • fib: Recursively calculate the Fibonacci sequence. Fully CPU bound and will peg all CPUs at max concurrency.

• benchmark/impl:

  • sync: Naive "run code in a for loop on main event loop" execution. The baseline.
  • jest-worker: Offload executions using child processes or worker threads using jest-worker
  • workerpool: Offload executions using child processes or worker threads using workerpool

Development

If you want to see the timing of individual implementation executions try this:

$ DEBUG=ssr:timer yarn benchmark

Example results

Key

• Demo: Scenario
• Conc: Level of concurrency and number of executions
• Args: Options passed to worker (e.g. "number of times to repeat")
• M loops: Main thread loops on 1ms timer (bigger is better)
• WI time: Single worker average "inner" time inside worker (smaller is better)
• WO time: Total worker "outer" time across worker (smaller is better)
• W result: Correctness vs synchronous baseline

Node 8

• os: darwin 18.7.0 x64
• cpus: 4
• node: v8.10.0
• execution: child process

Node 10

• os: darwin 18.7.0 x64
• cpus: 4
• node: v10.16.3
• execution: child process

Node 12

• os: darwin 18.7.0 x64
• cpus: 4
• node: v12.11.0
• execution: worker thread

Maintenance Status

Archived: This project is no longer maintained by Formidable. We are no longer responding to issues or pull requests unless they relate to security concerns. We encourage interested developers to fork this project and make it their own!