Abstract #
The rapid development of software systems in modern avionics equipment poses serious challenges to existing mainstream avionics software architectures both in China and abroad. Thus, it is necessary to explore and design a next-generation architecture for avionics software. This paper analyzes the evolution of avionics software architectures, studies the latest software technologies in the computing industry, and proposes a technical approach and design scheme for the next-generation avionics software architecture. Performance comparisons with other architectures validate the effectiveness of the proposed solution.
Keywords
Avionics equipment; software architecture; embedded real-time operating system; virtualization; containers
Development and Current State of Avionics Software Architecture #
Evolution of Avionics Software Architecture #
Avionics software architecture has gone through three generations:
- First-generation: No OS, only a simple real-time monitor for booting, hardware initialization, and debugging. Applications were functionally modularized.
- Second-generation: Flat embedded OS (e.g., VxWorks 5.5, Tianmai 1). Provided hardware isolation and real-time multitasking, simplifying development.
- Third-generation: Partitioned OSs (e.g., VxWorks 653 2.4, DeltaSVM, Tianmai 2) and middleware. Multiple applications run on one device with functional integration.
Current Mainstream Architectures #
Most new avionics systems use third-generation architecture: Partitioned OS + Middleware + Partitioned Applications, leveraging high-performance processors, high-speed buses, and static configuration. In systems like Vehicle Management Systems (VMS), middleware integrates traditional flight control with electromechanical management functions.
In avionics systems using Integrated Modular Avionics (IMA), all avionics applications are consolidated on Integrated Center Processors (ICPs) using ARINC 653-compliant partitioned OSs and middleware.
Architectural Challenges in Future Avionics #
Modern warfare’s information-centric, autonomous, and networked nature demands smart, integrated, and networked electronics. Technologies like cloud computing, AI, and big data are being introduced into avionics. While multicore processors improve hardware performance, software functionality and complexity are exploding. Current architectures face several issues:
- Low hardware resource utilization: Partitioned OSs use static configurations. Unused resources are wasted.
- High configuration complexity: As software functions increase, defining and allocating fixed resources for each application becomes geometrically harder.
- Limited reconfigurability: Current architectures bind applications tightly to fixed partitions, hindering flexible reconfiguration.
Trends in International Avionics Architectures #
To address these challenges, major vendors have adopted virtualization:
- WindRiver’s VxWorks 653 3.0/3.1 uses virtualization to run multiple guest OSs (e.g., CretOS for ARINC653, POSIX-based Linux).
- DDCI DeOS and GreenHills Integrity-178 have also introduced virtualized, multicore-supporting versions for civil and military use.
Designing the Next-Generation Architecture #
Architecture Concept #
In commercial computing, similar problems are tackled using virtualization:
- Virtual Machines (hard virtualization): Simulate multiple OS environments on one physical machine.
- Containers (soft virtualization): Provide lightweight, isolated user-space environments that behave like separate systems but share the host OS kernel.
Compared to VMs, containers eliminate hypervisors and redundant OS layers, improving performance and reducing resource usage (2× better performance, ¼ the resource usage). Since avionics systems don’t require multiple OSs, containerization aligns better with avionics needs.
System Composition and Runtime #
The container-based avionics platform includes an embedded OS with container engine, middleware, and multiple application containers:
- Containers are type-specific: development/debugging containers for profiling, production containers with strict resource limits, or dynamic containers for low-security apps.
- The container engine allocates resources and instantiates container images based on configuration, then loads applications into them for execution.
Technical Feasibility #
Though mainstream container tools target Linux/Windows and adapting them for real-time embedded use is difficult, Chinese vendors have already developed soft-virtualization in avionics systems.
Given widespread adoption of domestic avionics OSs and deep expertise in the field, there’s strong potential for adapting containers into embedded real-time environments with supporting toolchains and middleware.
Implementation & Performance Comparison #
Based on DeltaSVM, a prototype implementation of the new architecture (DeltaSVM-C) was created. Tests were conducted using three architectures on the same hardware:
- Standard partitioned OS (DeltaSVM)
- Virtual Machine-based system using DaOS 8 with real-time VMs running DeltaSVM
- Container-based architecture with DeltaSVM-C
Test system: Intel i5-2500k (4-core, 3.3GHz, VT enabled, 4GB DDR3). Software was memory-resident.
Metrics tested:
- Interrupt response time
- Partition/container/VM switching time
- Memory usage
Results Summary:
| Metric | Partitioned OS | Virtual Machine | Container |
|---|---|---|---|
| Avg. Interrupt Time | Low | High | Moderate |
| Partition Switch Time | Low | High | Moderate |
| Memory Usage | Baseline | +300–600% | +50% |
Containers perform nearly as well as traditional RTOS partitions and far outperform virtualization. Their memory overhead is significantly lower than virtual machines while maintaining reasonable real-time performance.
Conclusion #
The proposed container-based avionics software architecture meets future system demands and offers several advantages over third-generation architectures:
- Higher resource utilization: Static app configurations within dynamically managed containers enable efficient resource sharing and instantiation.
- Simplified configuration and integration: Resource profiling containers and toolchain automation reduce the complexity of large system integration.
- Flexible reconfigurability: Containers support controlled resource allocation and dynamic replacement or restart—offering both determinism and adaptability.
*(This article is from Journal of Nanjing University of Aeronautics and Astronautics by Zhang Junhong and Tong Qiang, affiliated with AVIC First Aircraft Institute and Beijing Huazhixin Technology. Shared for educational purposes. Contact us for copyright.)