Author's demonstration of algorithm skills. Returns near-shortest circular path of 30 locations. This app mimicks the path finding service used by Amazon delivery vehicles and FedEx.
Traveling salesman problem related algorithm is computationally intensive. The least desired scenario is sending a request and wait for 5+ minutes. C++ is being used to boost the time taken to process the algorithm. The choice is deliberate to increase the number of requests handled per server.
Python and FastAPI are used for the backend. This choice is made to reduce development time, by working with already accumulated knowledge (please see udacity-cloud-devops-project-5 repository).
Apache Kafka is used to handle a large number of requests. Once a user (consumer) is subscribed to Apache Kafka (by requesting for shortest path of x points), producer (This application) solves the traveling salesman problem of x points, and it sends response back to consumer. All the details of Kafka events are stored in a hard drive. In sum, request bottleneck and the likelihood of server crashing (due to little RAM remaining) is reduced. Apache Kafka is a perfect choice to handle requests for this application where there are a large number of some time-consuming requests.
CircleCI is chosen as CICD solution (for now) to reduce development time. The integration and deployment templates are already prepared in previous project (please see udacity-cloud-devops-project-5 repository). In sum, working with already prepared CICD solution would free up time and resource to focus on area where needed.
-
250,000 vehicles are assumed based on this data here.
-
Including refresh, 2,500,000 requests (250,000 * 10 requests = 2,500,000 reuqests) are assumed.
- Marc Kuo. Traveling Salesman Problem for Deliveries. Routific. https://blog.routific.com/blog/travelling-salesman-problem