Alauda Service Mesh

Architecture

This architecture is built on the Istio, OpenTelemetry and Jaeger technology stacks, achieving the full lifecycle management of distributed tracing. The system comprises five core modules: data collection, transmission, storage, querying, and visualization.

TOC

Core Components

  1. Istio System

    • istio-proxy
      A sidecar proxy that automatically captures and generates trace data for service-to-service communication without code modification.
    • istiod
      Controls the configuration and lifecycle of istio-proxy, including the tracing configuration for service mesh.

  2. OpenTelemetry System

    • opentelemetry-operator
      A cluster-level Operator responsible for deploying and managing the otel-collector component, providing OTel automatic injection capability.
    • otel-collector
      Receives tracing data from applications, filters and batches it, and then forwards it to the jaeger-collector.
    • Tracing UI
      A self-developed visualization interface that integrates with the jaeger-query API, supporting multi-dimensional query conditions.

  3. Jaeger System

    • jaeger-operator
      Deploys and manages the jaeger-collector and jaeger-query components.
    • jaeger-collector
      Receives trace data forwarded and processed by the otel-collector, performs format conversion, and writes it to Elasticsearch.
    • jaeger-query
      Provides a trace query API, supporting multi-condition retrieval including TraceID and labels.

  4. Storage Layer

    • Elasticsearch
      A distributed storage engine that supports efficient writing and retrieval of massive Span data.

Data Flow

  • Istio Writing Process
    Application -> jaeger-collector -> Elasticsearch

    The istio-proxy automatically captures trace data from service-to-service communication, which is then sent directly to the jaeger-collector for persistence in Elasticsearch.

  • OpenTelemetry Writing Process
    Application -> otel-collector -> jaeger-collector -> Elasticsearch

    The application generates Span data via SDK or automatic injection, which is standardized by the otel-collector and subsequently persisted to Elasticsearch by the jaeger-collector.

  • Query Process
    User -> Tracing UI -> jaeger-query -> Elasticsearch

    The user submits query conditions through the UI, and jaeger-query retrieves data from Elasticsearch; the UI visualizes the results based on the return data.

ON THIS PAGE