gRPC is easy to get started with, and greatly simplifies distributing systems across the network. The backend that powers the Datadog product makes heavy use of it. During Datadog’s hyper growth phase, as this system became larger, problems related to gRPC usage emerged, such as load imbalance, unexplained request failures, corrupted data, or servers running out of memory. Our backend is now composed of millions of containers running in Kubernetes across many regions. The backend is written in different programming languages, such as Java, Go, and Python. Most of our workloads are data intensive and we need service to service communication to be as efficient and reliable as possible. In this session, we will tell our journey to standardizing gRPC usage and the benefits we obtained. By the end of the talk, attendees will have a better understanding of issues that can arise while scaling gRPC usage, and how to apply some of this knowledge to their own organization.
Wednesday September 20, 2023 1:20pm - 1:40pm PDT
Mariposa Grove