Embedded Linux Design and
Embedded Linux can have multiple definitions depending on the product. USA Firmware’s team of embedded system and Linux engineers understand the complexity of these definitions and protocols. For example, Linux can power a high-reliability embedded control module. In another product example, Linux can power a full-featured data-analytics engine with computational resources that rival some desktop systems.
The Many Facets of
Many Layers in a Product Solution
Embedded Linux solutions comprise several layers within a product. Solutions begin with the hardware itself, progress to the bootloader, advance to the Linux kernel, continue to Linux services and native utilities, and eventually write a product-specific application. All these layers must integrate correctly and reliably for the product to function correctly.
Every Layer a Moving Target
Open source embedded Linux is not the product of a single vendor. Each layer or subsystem is considered its own product with its own authors, to be released and updated on their own schedules. Each component’s authors decide how to maintain compatibility, or not, with previous releases. Incompatibilities often result from changes to APIs or functional behavior and from bugs introduced by a new release. Existential problems with a new release could include changes in licensing terms, which preclude further use of an application or library.
Integrating the Layers
The evolution of interface (API) dependencies between subsystems must be carefully monitored throughout the initial development of an embedded Linux product and well into its maintenance lifetime. This requires careful review of API and behavioral dependencies as they are introduced, review of the side effects of bug fix releases, and review of the effects that any revisions to underlying hardware might need.
Large-scale Linux installations have frequently chosen various forms of virtualization to avoid such dependencies. However, this practice frequently leads to much higher storage and computational resource requirements. Such solutions are not generally feasible for resource-constrained embedded systems.