Why SkyWalking’s Architecture Makes Modern Observability Seamless

This article is excerpted from "Apache SkyWalking实战" by SkyWalking founder Wu Sheng, detailing the system’s architecture and benefits.

SkyWalking Architecture Design

The official diagram shows SkyWalking composed of four core parts.

Probe : language probes or other project protocols (Tracing and Metrics).

OAP Platform (Observability Analysis Platform) : a lightweight, highly modular analysis engine consisting of a Receiver, streaming analysis kernel, and query kernel.

Storage Implementors : supports multiple storage backends via a standard interface.

UI Module (SkyWalking) : provides data query and visualization via GraphQL.

Design Principles

Protocol‑oriented design

Modular design

Lightweight design

Protocol‑oriented Design

Since version 5.x, SkyWalking strictly follows protocol‑oriented design, exposing several external protocols.

Probe Protocols

Language probe reporting protocol (registration, metrics, tracing, commands, service‑mesh telemetry). Implemented via gRPC since v6.

Language probe interaction protocol (HTTP/MQ headers, Base64 encoding, v3 simplifications).

Service Mesh protocol for telemetry collection from mesh platforms.

Third‑party protocols for projects such as Istio and Envoy.

Query Protocols

SkyWalking uses GraphQL for queries, offering flexible composition and better extensibility compared to REST.

Metadata query

Topology query

Metrics query

Aggregation query (average, Top‑N, etc.)

Trace query

Alarm query

Additional internal protocols exist for data export, alarm handling, and dynamic configuration.

Modular Design

The modular approach enables easy customization and plug‑in development. Java probes use SPI for plugin replacement, while the OAP server defines modules and providers via YML, allowing users to replace or add implementations without altering core code.

Lightweight Design

SkyWalking aims to provide powerful observability without requiring heavyweight big‑data stacks. It runs as lightweight JARs, supporting storage backends like MySQL, TiDB, or Elasticsearch, and can handle billions of data points daily with modest cluster setups.

Advantages of SkyWalking

Consistency Across Traditional and Cloud‑Native Environments

SkyWalking supports traditional RPC/HTTP microservice architectures (e.g., Spring Cloud, Dubbo) and modern Service Mesh solutions (Istio, Envoy), offering unified monitoring for both.

Ease of Maintenance

The system avoids complex technology stacks, reducing operational overhead while still delivering high‑value monitoring for large‑scale deployments.

High Performance

Its built‑in stream processing framework leverages bytecode techniques to balance extensibility and performance, handling terabytes of data in production cases.

Facilitates Extensibility and Integration

Protocol‑oriented and modular designs allow third‑party integrations via HTTP/gRPC, enabling seamless upgrades and custom extensions without code changes.

For deeper practical guidance, refer to the book "Apache SkyWalking实战".