This presentation celebrates a decade of gRPC, highlighting its widespread adoption and crucial role in modern distributed systems, particularly within cloud-native environments and the burgeoning AI/ML landscape. The speaker details gRPC's success drivers, including its high performance, language agnosticism, reliance on Protocol Buffers and HTTP/2, and its ability to facilitate microservices architectures with built-in reliability features. Key developments such as proxyless gRPC service mesh and the upcoming gRPC Rust preview are discussed, alongside its growing significance in AI/ML pipelines for efficient data ingestion, rapid inference serving, and enabling new agent-to-agent communication protocols like MCP and A2A at Google.
gRPC: A Decade of Innovation and Future Directions
0:03 Introduction & Milestone: Rajie, gRPC and gRPC Conf exec sponsor, reflects on 10+ years of gRPC usage and announces the 10th anniversary of its open-source release. The talk will cover gRPC's adoption, current status, and future.
0:33 Rapid Adoption & Industry Impact: In its decade since release, gRPC has seen rapid adoption across various industries due to its high performance and multi-language support, becoming a go-to for efficient and scalable communication in modern distributed systems.
0:51 Notable Adopters: Cisco was an early adopter, followed by Netflix (a significant contributor), Spotify, Reddit, and LinkedIn, which migrated 50,000 endpoints to gRPC.
1:31 Core Success Factors for Cloud-Native: gRPC's success stems from its efficient communication using Protocol Buffers (reduced message size) and HTTP/2 (network efficiency, real-time), making it ideal for microservices exchanging large data.
1:52 Language Agnosticism & Portability: Its language-agnostic nature allows developers to build services in their preferred languages, crucial for polyglot microservices in cloud-native environments.
2:09 Enabling Microservices & Reliability: gRPC supports microservices architectural patterns, fostering developer velocity and parallel team work, and includes features like load balancing and retries for improved reliability in distributed systems.
2:36 Internal Google Usage: Google heavily utilizes gRPC internally across projects like Ccd, ContainerD, and Kubernetes, solidifying its status as a popular choice for scalable, efficient cloud-native applications.
3:00 Future: Cloud-Native & Proxyless Service Mesh: gRPC is committed to becoming even more cloud-native friendly, introducing proxyless gRPC service mesh to streamline deployment, reduce operational overhead from sidecar proxies, and improve resource efficiency.
3:42 gRPC Rust Preview: Significant progress has been made on gRPC Rust, in collaboration with Tonic maintainers, with a preview launching soon, and dedicated talks and code labs available for deep dives and hands-on experience.
4:19 Revolutionizing AI/ML: gRPC is revolutionizing AI/ML models by providing speed, efficiency, and data streaming support, making it a key building block in AI/ML pipelines for training large models and serving inference requests quickly.
4:49 Efficient Data Handling in AI/ML: Its data serialization (Protocol Buffers) combined with the ability to handle large payloads ensures reliable data piping for training pipelines and low-latency, high-throughput inference requests.
5:16 Industry AI/ML Adoption: Prominent frameworks like TensorFlow utilize gRPC for managing inference requests, and NVIDIA's Dynamo Triton framework uses gRPC as its transport to ensure real-time predictions.
5:39 New AI/ML Protocols from Google: Google is developing new AI/ML protocols using gRPC, including the Model Context Protocol (MCP) for AI agents to connect to external tools, and Agent-to-Agent (A2A) protocol for AI agents to collaborate and delegate tasks.