A service container for a service-oriented architecture approach to (neuroscience) experiments.
DuraPy is an open-source Python service container for neuroscience experiments utilizing a service-oriented architecture, aiming to provide the communication and process backends for each microservice. DuraPy, at its core, utilizes an event-driven paradigm: “commands” can be sent from any service and are broadcast to all services. Each command can have its own structure, and per-command handlers can be installed in each service that calls a specified function when receiving a command, enabling each service to perform custom functionality per command. As part of its service architecture, each DuraPy service runs in its own long-running process, with commands likely sent by the experimenter to initiate and terminate experiments.
In addition, a web-based graphical user interface (GUI), running itself as a DuraPy microservice, is provided to enable experimenter control of the experiment, a core feature often underestimated for modern experiments. In addition to supporting the sending and displaying of commands, managing service deployment, and viewing logging, the web GUI implementation, built on an open-source web framework Tornado, is explicitly designed to enable extensions, where experimenters can more easily build custom web pages for experiment monitoring or review.
Experimenters can configure custom database engines for storing/retrieving of commands. By default, the command database utilizes Redis, an open-source in-memory database, but future changes could enable DuraPy to utilize other existing networked communication frameworks.
In addition, logging infrastructure, Git-based deployment commands via the webserver and command line, and POSIX-compatible run scripts (via systemctl) are provided.
The current method of installation is to simply clone this repository and place it on your PYTHONPATH, and begin extending.
See the example and pingpong directories for examples on how to utilize DuraPy. The example implementation starts up, calls a few commands, and then terminates itself. This example only utilizes an in-memory database as a command database, and so is only suitable for showing how to configure a barebones process & webserver.
The pingpong example contains two different DuraPy processes, pinger and ponger, where once a session is initiated, the pinger sends "ping" commands, to which the ponger responds with derived "pong" commands. To run the pingpong example, in addition to running the pinger and ponger processes, run the webserver process via pingpong/webserver/webserver.py and issue an InitiateSessionCommand. To stop it, issue an EndSessionCommand. This pingpong example utilizes a Redis command database expected to be running at its default location (localhost at port 6379).