PhD Oral Exam - Azadeh Azhdari, Information and Systems Engineering

When studying for a doctoral degree (PhD), candidates submit a thesis that provides a critical review of the current state of knowledge of the thesis subject as well as the student’s own contributions to the subject. The distinguishing criterion of doctoral graduate research is a significant and original contribution to knowledge.

Once accepted, the candidate presents the thesis orally. This oral exam is open to the public.

Abstract

The advent of cloud computing has revolutionized application development, deployment, and operation, leading to cloud-native computing. By leveraging containers and microservices architectures, cloud-native systems decompose monolithic applications into modular, fine-grained components, enabling scalability and agility. In parallel, the telecommunications industry has undergone a major transformation with Network Function Virtualization (NFV), shifting network functions from specialized hardware to software components deployed on virtualized infrastructures.

NFV has further evolved by adopting cloud-native principles, resulting in Cloud-Native Network Functions (CNFs), which adopt microservices-based designs for flexible service deployment. Within network function lifecycle management, decomposing Virtual Network Functions (VNFs) into smaller sub-functions enables efficient deployment and resource allocation. However, it introduces challenges in selecting optimal decomposition structures. This fine-grained decomposition leads to modular and independently deployable components that require continuous updates to introduce new features and fixes. In the operation phase, this leads to frequent software upgrades, where managing version compatibility and dependency constraints among interdependent components becomes challenging and may risk service disruption. To support this evolution, Continuous Integration and Continuous Deployment (CI/CD) pipelines automate system updates. However, their complex execution flows make tracing failures and analyzing causality across pipeline stages challenging.

This thesis addresses these challenges through three main contributions. First, we propose a cost-aware VNF decomposition framework that explores decomposition structures and identifies best configurations with reduced resource consumption. Second, we develop a topology-aware graph-based software upgrade framework that models version compatibility, dependency constraints, and system topology to generate optimized upgrade plans while ensuring global consistency. Third, we introduce a causality analysis framework for CI/CD pipelines that models execution workflows as a knowledge graph and employs supervised learning to identify relationships between execution steps and generated artifacts helping automatic failure tracing and troubleshooting.