This webpage collects performance information about various hosted software
development services. The reports are updated sometimes, most recently on
February 2, 2022 at 12:25 UTC, and were generated with
Lighthouse
6.3.0.
Patches welcome!
Expand for details on methodology
We ran Lighthouse 6.3.0 on Ubuntu x86_64, with the
following command line:
lighthouse <url> \
--only-categories=performance,accessibility \
--profile=desktop \
--throttling.cpuSlowdownMultiplier=1
The desktop profile is used because it's rare that a developer is coding
on a mobile device — and some forges have separate mobile sites, so
a separate test suite would be warranted.
We left the throttling settings at the Lighthouse default, because it's
less rare that a developer is tethered to their phone or working from a
community with less privileged internet access. The simulated latency
also helps to reduce the variable impact of real-world latency.
Additionally, simulating a low-bandwidth connection helps to draw out the
differences between services more clearly. This controls for network
conditions; the same tests have been run from various datacenters with
various degrees of favorability to each hosting provider with no
discernible change in the results.
CPU throttling is disabled to offset the effects of not having a GPU
available in our testing environment. If you run these tests locally, it
is recommended that you remove this constraint.
The
source code
(upstream)
is available if you wish to run these benchmarks on your own system; it
takes about an hour.