Event-Driven Distributed Task Orchestration System with Applications to Automated PCB Design

Abstract

Printed circuit board (PCB) design is the process of taking a board schematic and design constraints and realizing a manufacturable design. Electronic Design Automation (EDA) software allows humans to manually design PCB’s by placing components and routing the electrical connections required. Allegro X AI by Cadence is a cloud-based tool that utilizes machine learning and optimization to automatically generate PCB designs.

Microservice-based architectures have proven to be popular due to their flexibility and scalability. X AI’s current process for generating a printed circuit board design is monolithic with logically separate stages, making it difficult to support flexible configuration of the ordering of downstream stages or branching off the current design and attempting different versions of a stage by varying input parameters and constraints.

In this thesis, we design a microservice-based architecture and orchestration system for automated PCB design. Our design structures the application as a directed acyclic graph (DAG) of microservices and achieves the following goals: decouples the stages of the design generation flow, supports flexible configuration and ordering of downstream stages, and brings the power of elastic compute from the cloud to the PCB design generation process.