Over time, as applications scale up and down in a Kubernetes-managed environment, a cluster state of resource fragmentation may be reached. Such a state indicates that the hardware of a cluster is unable to be utilized to its fullest potential, as resources become locked by poor Pod to Node mappings. This thesis shows that such problems are not uncommon as workload requirements approach the resource availability of a cluster. Additionally, an experimental analysis of directed Pod eviction is carried out. Directed Pod eviction is a technique that attempts to combat the issue of resource fragmentation. It is found through this analysis that directed Pod eviction can reduce resource fragmentation and allow clusters to run the same workload using 16.7% fewer Nodes than what is consistently possible using standard Kubernetes scheduling.