Benchmarks for Bref running on AWS Lambda.

Scenarios

• PHP function: a simple PHP function, see php-function/index.php
• HTTP application: a simple PHP script that returns Hello world, see http-application/index.php

Average (warm) execution time

Average execution time for a lambda that doesn't do anything.

Number of samples: 900

	128	512	1024	2048
PHP function	140ms	33ms	16ms	15ms
HTTP application	1ms	1ms	1ms	1ms
Symfony	32ms	5ms	4ms	4ms

For comparison on a 512M Digital Ocean droplet we get 1ms for "HTTP application" and 6ms for Symfony.

CPU performance

The more RAM, the more CPU power is allocated to the lambda. This is clearly visible when running PHP's official bench.php script.

	128	512	1024	2048
bench.php	5.7s	1.4s	0.65s	0.33s

For comparison bench.php runs in 1.3s on a 512M Digital Ocean droplet, in 0.8s on a 2.8Ghz i7 and in 0.6s on a 3.2Ghz i5.

Cold starts

Number of samples: 20

	128	512	1024	2048
PHP function	420ms	250ms	230ms	220ms
HTTP application	370ms	310ms	300ms	300ms
Symfony	5.8s	1.6s	920ms	650ms

Notes:

• Symfony's cold start is high at the moment because we do not pre-generate the cache before deploying (it is generated on the first hit). This is an optimization that is planned.

Reproducing

You will need to install dependencies of Bref. Then:

• clone the repository
• create a S3 bucket in us-east-2 and change the bucket name in Makefile (replace it with yours)
• make deploy
• make bench-cold-starts
• make bench-function
• make bench-http