This week, I proposed a system architecture change to the group. Because we are ultimately only monitoring the performance of a load balancer and its effect on the overall system, we can reduce our minimum viable product to that of a single user-facing application load balancer that connects users’ browsers to a chosen server node from the front-end tier. This architecture can allow us to simplify the further back-end structure to one of two options. Either each front-end node has a dedicated video database to retrieve from or all front-end nodes have a shared scalable database that they can retrieve from. I am debugging a MongoDB implementation for the first of those options as a test.
As we are waiting on Jason’s group deployment automation for testing the system within an AWS environment, I also began researching our options for user simulation. Flood.io is a promising third party solution I found that I plan to bring up to the team soon. It is a scalable load testing platform that can be directly written to with user simulation scripts and provides its own automated application performance monitoring. Furthermore, flood can be directly integrated to an AWS deployment. A considerable amount of testing (500 user hours) can be done for free before we may need to decide to spend some of our budget on further tests.