Why SkyWalking’s Architecture Powers Modern Cloud‑Native Observability

This article, excerpted from Wu Sheng’s "Apache SkyWalking实战", explains SkyWalking’s architecture and its advantages.

1. SkyWalking Architecture Design

SkyWalking consists of four core components:

Probes : language probes or other project protocols.

OAP Platform (Observability Analysis Platform) : a highly modular lightweight analysis engine composed 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 through GraphQL.

From a design perspective, SkyWalking follows three main principles:

Protocol‑oriented design

Modular design

Lightweight design

Protocol‑Oriented Design

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

1. Probe Protocols

Four categories of probe protocols are defined:

Language probe reporting protocol (registration, metrics, tracing, commands, service‑mesh telemetry) – all language probes (Java, .NET, Node.js, PHP, Go, etc.) must follow this gRPC‑based protocol.

Language probe interaction protocol – defines HTTP/MQ header interactions, using Base64 encoding from v2 and a simplified v3 from SkyWalking 8.

Service Mesh protocol – a proprietary protocol for Mesh services to upload telemetry data.

Third‑party protocol – adapts large open‑source projects such as Istio and Envoy.

2. Query Protocol

SkyWalking uses GraphQL for queries, offering flexible, composable queries and better extensibility compared to REST. The query protocol includes six categories:

Metadata queries

Topology queries

Metrics queries

Aggregation queries (range averages, Top‑N, etc.)

Trace queries

Alarm queries

Additional module‑specific protocols (data export, alarm, dynamic configuration) are also provided and can be replaced.

Modular Design

SkyWalking’s modular architecture enables easy customization and extension, allowing users to replace or add modules without breaking compatibility. Java probes use SPI for plugin replacement, while the OAP server defines modules and providers via YML, separating abstract services from concrete implementations.

Lightweight Design

SkyWalking is designed to be lightweight, avoiding heavy big‑data stacks while still delivering powerful analysis, scalability, and extensibility. It can run with simple JAR deployments and supports various storage backends (MySQL, TiDB, Elasticsearch) for high‑throughput monitoring.

2. SkyWalking Advantages

2.1 Consistent Support for Traditional Distributed and Cloud‑Native Architectures

SkyWalking supports traditional RPC/HTTP microservice architectures (e.g., Spring Cloud, Apache Dubbo) and seamlessly integrates with Service Mesh solutions (Istio, Envoy) on Kubernetes, providing unified monitoring across both environments.

2.2 Easy Maintenance

By limiting the technology stack and simplifying cluster deployment, SkyWalking reduces operational overhead while handling billions of monitoring data points daily.

2.3 High Performance

SkyWalking includes a specialized streaming computation framework optimized for monitoring data, leveraging bytecode techniques to balance extensibility and performance.

2.4 Facilitates Secondary Development and Integration

The protocol‑oriented and modular design allows easy integration of third‑party tools via HTTP or gRPC interfaces, supporting plug‑in development and stable ecosystem evolution.

For a comprehensive guide, refer to the book "Apache SkyWalking实战" authored by Wu Sheng and the core development team.