Minimum-Cost Design of Auto-Scaling Server Farms Providing Reliability Guarantees

As next-generation mobile networks increasingly rely on virtualized infrastructure to deliver critical services, ensuring both the efficiency and reliability of server farms becomes essential. These infrastructures must meet stringent reliability guarantees to support time-sensitive applications in emerging 5G and beyond networks. In this paper, we address the design of auto-scaling server farms–specifically, selecting the most suitable server type and corresponding number of servers–by considering both service requirements and associated operational and infrastructure costs. To this end, we develop an optimization algorithm that combines (i) a queueing-theoretic model to estimate the resources needed to meet reliability constraints, and (ii) a general cost model that captures both capital and operational expenditures. We validate our approach through extensive simulations, comparing it against classical queueing-based methods and exhaustive numerical searches: our proposal reduces costs by 22% as compared against the benchmark, with solutions that are within 3% of numerical searches at 10% of the computational complexity, offering a new scalable and cost-effective methodology for designing reliable server farms.