Today, the Autojoin system (a.k.a "BYOS" or "host-managed") only formally supports NDT. We should investigate how the Autojoin system could support other experiments such as Wehe and Revtr. In theory, and on the surface, it should be as simple as creating a new Docker Compose file for each experiment that runs the proper containers with the proper flags. In practice, I suspect this may also require adjustments to how Pusher is working, how GCS is structured, and how the Autoloader handles data from Autojoin nodes, since the whole system is likely to have been engineered with NDT specifically in mind.
Once we have Autojoin systems for other containers, then site hosts can choose to run any number of VMs with different M-Lab experiments.
FYI: @sermpezis
Today, the Autojoin system (a.k.a "BYOS" or "host-managed") only formally supports NDT. We should investigate how the Autojoin system could support other experiments such as Wehe and Revtr. In theory, and on the surface, it should be as simple as creating a new Docker Compose file for each experiment that runs the proper containers with the proper flags. In practice, I suspect this may also require adjustments to how Pusher is working, how GCS is structured, and how the Autoloader handles data from Autojoin nodes, since the whole system is likely to have been engineered with NDT specifically in mind.
Once we have Autojoin systems for other containers, then site hosts can choose to run any number of VMs with different M-Lab experiments.
FYI: @sermpezis