PsyNet experiments currently support three main modes of deployment:
SSH to a custom server
SSH to a custom server#
This is the preferred option in most cases.
The experimenter acquires some kind of Linux server to which they have SSH access,
perhaps by purchasing a physical server
that can be connected to the internet, or by acquiring a cloud server via a service
such as Digital Ocean
or Amazon Web Services.
It’s recommended to find a server with at least 16 GB of RAM.
There is then a built-in Dallinger command that connects to this server and
makes it ready for hosting experiments:
dallinger docker-ssh servers add.
An alternative approach is to deploy experiments using Heroku. Heroku is a cloud computing services provider that helps to orchestrate deployments of web apps. The advantage of using Heroku is that it performs the provisioning for you, i.e. spinning up web-servers when you launch the experiment and taking them down once you’ve finished. However, it comes with a big pricing disadvantage: running a Heroku experiment for a month could cost hundreds or even thousands of dollars, depending on the size of server provisioned. In contrast, the custom server approach should only cost tens of dollars.
If you are collecting data in person it is also possible to run experiments on your local machine, e.g. a laptop you bring to participants. We have not tested this workflow so much and so would generally recommend the online server approach for now, though. Watch this space!