Shards: a system for systems

Operating system construction is often focused on the internal operation and architecture of a general purpose system. This thesis instead focuses on systems built in response to a specific purpose, design intent, application load and platform. These are referred to as custom systems in the thesis. These focused systems have known demands, constraints and requirements that provide a target for system design and optimisation. These systems can perform valuable and demanding tasks which may encourage optimisation effort. The first challenge is discovering and capturing these attributes in an encoding that can be machine manipulated.<br><br>The second challenge was to use this information in a way which makes custom system construction economical, thereby widening the range of systems for which such efforts are appropriate. A bespoke and manual system construction is too expensive for the more narrowly deployed systems being considered. The operating systems field generally assumes a long lived and widely deployed general system which can afford significant design effort up-front which is not applicable in this case. The proposed solution was to balance the advantages of modular functionality with automated configuration, construction and tailoring based on the captured demands of the proposed system. Effectively the operating system is compiled as an integrated part of the system. <br><br>In such an approach new inputs not relevant to general systems, such as application code and design intent, are known in advance and can inform the system generation process. This leads to an operating system structure that is determined by and optimised to the needs of the proposed system. A clean architecture is often a design goal for system construction. In this case the ideal is an operating system so integrated into the overall system there is no clearly identifiable run time structure. The Operating System could become part of the hardware, system operation or applications of the system.<br><br>The final goal was to build a foundation in which construction work or advances can be captured and reused. Building a complete 'system of systems' in a single project would be an impractical undertaking. The effort was to build an approach and framework which could grow as a side effect of its use and application. This allowed the lessons learnt and work done in one project to potentially enrich both this approach and the domain of operating systems.<br><br>