Solution of some optimisation problems in the development of centralised and decentralised distributed systems

Abstract

This paper addresses optimisation problems that arise in the development of centralised and decentralised distributed systems. We unify the treatment of technology workflows and architectural patterns of corporate computer networks by modelling them as open queueing networks with product‑form stationary behaviour (Jackson networks). On that basis we formalise throughput ceilings, latency/availability trade‑offs, and actionable levers such as feedback reduction, proportional flow splitting, server pooling, and targeted capacity increments at bottlenecks. We further propose an extended architecture notation that separates protocol and infrastructure security, encodes independent failure domains and consensus type, and binds qualitative labels to measurable KPIs. The approach naturally supports both single‑criterion optimisation and multi‑objective (Pareto) analysis and aligns with current architecture‑description standards.