The advent of Satellite as a Service (SaaS) platforms has empowered satellite service providers (SPs) to rent portions of satellite capacity from infrastructure providers (IPs) to cater to the diverse demands of their users across multiple satellite services. To effectively manage costs and maintain a high Quality of Experience (QoE) for numerous concurrent connections, SPs should secure flexible capacity from IPs. However, the irregular and unpredictable nature of traffic demands from various applications complicates the capacity-renting framework. This study presents a dynamic capacity allocation framework that efficiently handles diverse traffic flows with varying arrival rates, aiming to minimize rental costs while meeting blocking probability and QoE requirements. Utilizing theMt/Mt/1 queuing model and a continuous-time Markov chain, the technical designs are framed as a statistical optimization problem. In this context, the system waiting-queue lengths are estimated using the transient probabilities of Kolmogorov equations. Subsequently, cumulative distribution functions are employed to re-formulate this stochastic optimization problem into a convex form, which can be tackled through the Lagrangian duality method.Through extensive simulations and numerical assessments, we illustrate our method’s efficacy, with the proposed algorithm outperforming benchmarks by reducing costs by up to 9.85% and 3.1%.

Efficiently utilizing the network resources to minimize the operation costs while satisfying customer’s Quality-of-Experience (QoE) related requirement as well as dynamic demands is a challenging task of all network operators. This paper aims to develop a stationary capacity allocation method that anticipates time-varying demand and keeps the network operating under constraints on a stochastic blocking probability. Queuing delay requirement is also regarded as an QoE-oriented practical design. Employing an approximation of time-varying queuing model and continuous time Markov chain (CTMC) for queue length, the technical designs are stated as a convex stochastic optimization based on which a dynamic capacity allocation is proposed by using Lagrangian and gradient descent searching method. Numerical studies confirm that our proposed framework can efficiently and dynamically allocate optimal capacity for a blocking probability of less than 1% and the probability of violating the queuing-delay requirement is less than 5%.