A Deep Dive on How To Leverage the NVIDIA GB200 for Ultra-Fast Training and Inference on Kubernetes
-
Kevin Klues
NVIDIA
Distinguished Engineer
Many organisations use gRPC for their service-to-service communications. At first glance, this is “just another protocol” running over TCP/HTTP. But working with gRPC load balancing can be a bit tricky as you have to use a different set of tools for testing (e.g. grpcurl instead of curl), the return codes are different and things like unary calls vs. streams require a good understanding of the protocol and its implications. We will give an overview of the protocol, explain what load balancing options we have using examples of AWS Load Balancers. We will explain how to perform functional and performance testing and finally go through some common pitfalls when Load Balancing gRPC.
World Congress 2025
Kevin Klues
NVIDIA
Distinguished Engineer
Maxim Fateev
Temporal Technologies
CTO and Co-Founder
Vadym Kazulkin
ip.labs
Head of Development
Paweł Siwek
Dynatrace
Principal Product Architect