gRPC Conf 2025

Guest agents on containers and virtual machines handle communications between the host and guest machine and the same tooling can scale to distributed system use-cases as well. In this talk, we present an architecture that leverages gRPC for the communication interface in agents to enable building a container runtime, while also allowing us to evolve and simplify the language stack over time.

We will leverage the recently released Apple Containerization project as a case-study where this has worked well.
We will show how the agent has helped setup the environment for container workloads by configuring running process lifecycle, setting up the network configuration, and even configuring the emulator to run workloads of different architectures

All of this runtime configuration has been orchestrated initially in Swift and Go and is now a simplified stack with just Swift used across the board, both on macOS and also on the Linux bits.

As distributed systems grow in complexity, ensuring correctness under failure and race conditions becomes increasingly difficult. Traditional integration testing for gRPC based services often falls short: it’s flaky, non-deterministic, and hard to reproduce. What if we could simulate a full network stack deterministically from service-to-service communication down to dropped packets and reordered messages?

We present an innovative way to build and test gRPC-Rust (Tonic) services in Rust by leveraging Rust’s ecosystem. We will walk through how to build gRPC services that can be simulated with minimal changes to your code base. The result: rock-solid, reproducible test scenarios for everything from retries and backoffs to leader election and network partitions.

gRPC is a mature project used by countless companies across numerous industries. It's time for gRPC reach graduated status in the CNCF. In this talk, we'll cover some governance changes we're making to achieve that goal. We'll cover the new Contributor Ladder, Steering Committee, and everything in between.

Services often require contextual metadata that is ancillary to the active request including session information, routing strategies, failure testing, and many other use cases. Managing and transmitting this data across a wide ecosystem presents lots of difficulties coordinating the schema and responsibilities. Come learn how Netflix manages and shares contextual data across our deep and complex set of microservices to make various business decisions. These services include not only gRPC but also REST across various programming languages.

We will discuss first and foremost what contextual metadata is. This includes our utilization of protobuf to define and model strict relationships in a highly structured polyglot ecosystem. Then we will expand on how we propagate that data across different IPC transport frameworks, particularly using gRPC metadata headers. We will finish with lessons learned and how we effectively scale, manage and observe this system.

Are you passionate about Rust? Join fellow Rust enthusiasts in this interactive Birds of a Feather session! This isn't a typical presentation; instead, it's a unique opportunity to engage in open discussion, share your experiences, and get answers to your burning Rust questions from the collective wisdom of the group. Come prepared to connect, learn, and contribute!

io_uring is an open source asynchronous syscall API for linux which allows an user to submit a syscall and receive an asynchronous completion notification. It brings some advanced features such as (1) syscall batching and chaining - the ability to batch multiple syscalls into one submission into the kernel, (2) buffer pooling - automatic selection of receive buffers from a pre-registered set of buffers to hold data read over the socket. These features can help improve performance of the gRPC-C++ stack. This talk would cover our experiences with developing an io_uring based EventEngine for gRPC-C++ and benchmarking it against the default epoll implementation.