Back to glossary

Service Mesh

A dedicated infrastructure layer that handles service-to-service communication in a microservices architecture, providing traffic management, security, and observability without changing application code.

A service mesh deploys a lightweight proxy (sidecar) alongside each service instance. All inter-service traffic flows through these proxies, which handle cross-cutting concerns transparently: mutual TLS encryption, load balancing, circuit breaking, retries, timeouts, and distributed tracing. The application code makes simple HTTP or gRPC calls; the mesh handles the operational complexity.

Popular service mesh implementations include Istio (feature-rich, Kubernetes-native), Linkerd (lightweight, simpler), and Consul Connect (HashiCorp ecosystem). They provide a control plane for configuring traffic policies and a data plane of sidecar proxies that enforce those policies at runtime.

For AI microservices architectures, a service mesh simplifies operations significantly. Traffic splitting for canary deployments of new models, mutual TLS between services without certificate management code, automatic retries with circuit breaking for flaky model APIs, and detailed latency metrics between services are all handled transparently. The trade-off is added resource overhead and operational complexity from managing the mesh itself.

Related Terms