This project will not likely be of much use generally but the code is available here for you to study or borrow from if you find value. This project implements several components that take the number crunching from a private project and divides the work so that it can scale to cloud workloads.

It looks something like this:

The client is developed in Unity (private repo). It uses the GPU and compute shaders to do a lot of number crunching in parallel and allows you to view and explore the results. The goal of this project is to dramatically increase the speed and surface area of the scan space by moving the compute operations into the cloud.

The Unity client sends a JSON message to the Gateway. The Gateway processes the message and potentially generates many work request messages, dropping them onto the message bus. Any number of large compute + GPU instances pick up a request and crunch on it for the amount of work specified in the message.

Each message includes a minimum bar that a result must meet to continue. Results that qualify are dropped onto the message bus and further processed. A Persistence service picks up a copy of these results and serializes them to a database. Separately, one or more QoS services (quality of service) pick up the results, buffer them, sort them ordered by the "best" results and enqueues them into a "best effort" sorted queue back to the Unity client to visualize the progress.

The Unity client receives the results over a persistent websocket connection, currently with the Gateway but it could be separate.

TODO

TODO