Memphis.dev recently organized its first hackathon; participants were tasked with saving the island of Zakar by predicting wildfires from temperature sensor data and social media posts. To set up the hackathon, I simulated seven years of data, generating more than four million messages.
Each participant needed to load that data into their Memphis Cloud account, so I wrote an upload script using the Python SDK. When I tested the script on my home network, I was easily seeing ingestion rates of 11,000 messages per second. But when I tested it using my Memphis Cloud account, messages were only being produced at a rate of 30 messages per second. Oh no! What gives?! And how did I end up fixing it?