Service Mesh
Subject: Computer and Technology noun /ˈsɜːr.vɪs mɛʃ/ 1 definition
- distributed-systems
- cloud-computing
- microservices-architecture
- computer-networking
-
Definition (noun)
Entry #78 · UpdatedA Service Mesh is a dedicated infrastructure layer that manages, secures, and monitors internal service-to-service (east-west) network communication across distributed or microservice architectures. Separated into a lightweight data plane of network proxies and a centralized control plane that distributes routing rules and cryptographic identities, it transparently handles traffic shifting, load balancing, mutual TLS (mTLS) encryption, retries, and telemetry collection. By offloading cross-cutting networking concerns from individual workloads, a service mesh keeps business logic decoupled from network reliability and security enforcement without requiring code changes or language-specific libraries.
Common Forms
Sidecar-Based Mesh: Injects a dedicated user-space proxy container (such as Envoy) alongside every application instance to intercept traffic locally, offering strong workload isolation at the cost of per-instance memory and CPU overhead.
Sidecarless (Ambient) Mesh: Moves traffic interception out of application pods into shared node-level Layer 4 proxies and optional Layer 7 waypoint proxies, lowering resource usage and decoupling proxy upgrades from application restarts.
Kernel-Level (eBPF) Mesh: Processes packet forwarding, load balancing, and identity-based security policies directly inside the Linux kernel via Extended Berkeley Packet Filter (eBPF) hooks, reducing latency by avoiding repeated user-space proxy hops.
Managed Cloud Mesh: Offloads control-plane hosting, certificate authority rotation, and multi-cluster discovery to a cloud provider's platform instead of requiring teams to operate mesh infrastructure themselves.
Example
In an e-commerce Kubernetes cluster, when the
checkoutservice calls thepaymentservice, its local mesh proxy automatically upgrades the connection to mutual TLS using short-lived SPIFFE certificates, retries transient503responses up to three times with exponential backoff, and emits distributed trace spans. During a canary rollout ofpayment-v2, an engineer applies a mesh routing rule that shifts 10% of production checkout requests to the new version while keeping application code completely unchanged.