networking-transport

To prevent load balancer idle timeout connection drops, set TCP keepalive intervals below your load balancer's idle timeout threshold. This maintains healthy long-lived connections by ensuring keepalive packets reset the idle timer before the load balancer terminates the session.

Use HTTP/3 QUIC instead of HTTP/2 to reduce tail latency when operating under packet loss or cross-region latency conditions. HTTP/3 eliminates head-of-line blocking at the transport layer, preventing delayed packets from stalling multiplexed streams like they do in HTTP/2.

Choose REST JSON for broad interoperability and simple request-response patterns, or gRPC Protobuf for low-latency service-to-service communication with strict schema evolution rules. Protobuf requires following wire-compatibility rules to prevent protocol mismatches during deployment.

Head-of-line blocking affects HTTP/2 streaming by causing all multiplexed streams to stall when a single TCP packet drops, since the protocol layer cannot reorder lost frames. This increases tail latency for production services relying on concurrent stream processing.

Yes, you can reuse gRPC channels and connection pools for long-lived streaming connections. Maximizing client reuse across websockets, SSE, and gRPC streaming reduces connection setup overhead while maintaining stable connections within idle-timeout operational limits.

Long-lived connection patterns like websockets and SSE face limitations from cloud load balancer idle timeouts, requiring keepalive tuning to prevent unexpected termination. Operational realities also include managing retry bounds and handling connection failures during cross-region streaming.