Week 8 (18/10 - 22/10)

Research and experimentation

Determining what to do with feedback

The feedback session with Bartosz last week gave a new perspective to the project and made me reassess where to take it. The focus would now not be on the analysis, but on the pipeline leading up to it. Another factor here was creating a proposal as to what kind of Azure resources would be needed to facilitate such a pipeline and ensure it is as traceable as possible.

Tackling this problem required more research. First of all, I had to ensure my current choice of technology was up to the task. Luckily, it is! Django is used in many large applications around the world, and should work for the current use case. There are certain requirements to this, however. One of these is that the application should be entirely stateless. Cloud-based infrastructure is ideal for this. My database is entirely in the cloud (using Postgres), and Azure has queueing systems were that to be necessary.

The application will be run in a docker container, though what that container looks like will depend on the further results of my research. There are various options when it comes to communicating between Django and OpenFace. One option is to integrate the web-app and OpenFace in a single container which can then be deployed. This makes it more integrated and easy to deploy while still being functionally seperate within the container. It would then not be possible to scale the two independently of each other, however. Another option is to have two separate containers and have them deployed on the same VM, running docker commands instead of directly in the OpenFace terminal.

Azure has various services which can offer a more cloud-based approach. An interesting option is WebJobs: these allow small scripts or applications to be available at all times to perform small tasks for a larger application (or group of applications).

All in all, it is difficult to see what would be best. That is why there will be a meeting after the holidays to discuss all these options with an expert and create a final proposal.

Spoilers for next week

Next week is a holiday week. This is why I want to spend my time working on my portfolio (both on content and structure).