This thesis investigates the concept of sovereignty in satellite networks, starting with a review of its foundations, including how technological, operational, and regulatory sovereignty apply across the space, ground, and user segments. Building on this foundation, the thesis identifies and classifies common sovereignty problems in satellite networks, such as dependence on foreign-manufactured components, traffic routed over gateways and inter-satellite links in other jurisdictions, and limited control over network management and data flows. It then formulates the sovereign manufacturer placement problem: deciding where to deploy network elements from trusted manufacturers (e.g., satellites, gateways, ground stations) so that both regulatory constraints (such as spectrum licensing, landing rights, and data-localization requirements) and sovereignty constraints (such as a minimum share of trusted equipment on every end-to-end path) are met. The resulting optimization model will be evaluated on representative LEO constellation scenarios to quantify the trade-offs between sovereignty, cost, and network performance metrics such as latency and coverage.

Prerequisites

This study looks at how batch size affects the cost and reliability of a queue-driven batch consumer running on self-hosted campus servers, under a mix of per-message faults and invocation-level faults such as timeouts, crashes, and throttling. The testbed is built from open components: an ElasticMQ message queue with visibility timeouts and a dead-letter queue, a custom poller and worker pool that supports partial batch responses, and a PostgreSQL database behind a rate limiter. A simple analytical model predicts the optimal batch size from two dimensionless ratios: per-batch overhead to per-message work, and fault rate times per-message work. Because the platform is fully observable and has no usage costs, the model can be calibrated and validated over a dense experiment grid, with faults injected at the application, process, and network level.

This study looks at how batch size affects the cost and reliability of an SQS-triggered AWS Lambda consumer with partial batch response enabled. It uses a mix of per-message faults and invocation-level faults such as timeouts and throttling. A simple analytical model predicts the optimal batch size from two dimensionless ratios: invocation overhead to per-message work, and invocation fault rate times per-message work. The model is calibrated and validated with measurements on an SQS → Lambda → DynamoDB pipeline that runs entirely within the AWS always-free tier.

Prerequisites

Satellite-5G integrated networks are expected to deliver ubiquitous, resilient connectivity, but link failures, handovers, and outages on either the satellite or terrestrial segment can severely disrupt ongoing services. These disruptions make fast, seamless failure recovery essential. Multipath QUIC (MPQUIC) is an extension of the UDP-based QUIC transport protocol that lets a single connection use several network paths at once, such as a satellite link and a 5G link, so traffic can continue over the remaining path when one fails. However, how quickly and smoothly MPQUIC recovers depends heavily on two choices: the packet scheduler, which decides how data is split across paths, and the congestion control algorithm, which adapts the sending rate to each path's capacity and delay. This thesis, therefore, aims to systematically evaluate combinations of MPQUIC schedulers and congestion control algorithms under realistic failure scenarios, in order to identify the pairing that gives the best recovery performance in satellite-5G networks.

Ideally, the student must recommend the best CC and scheduler. If possible, propose a new CC or scheduler algorithm. 

Prerequisites