An Experimental Evaluation of the Coda Distributed File System Using the Loki State-Driven Fault Injector
Ryan M. Lefever · 2003
Experimental evaluation is an important way to assess distributed systems, and fault injection is the dominant technique in this area for the evaluation of a system's dependability. For distributed systems, network failure is an important fault model. Physical network failures often have far-reaching e#ects, giving rise to multiple correlated failures as seen by higher-level protocols. This thesis presents an experimental evaluation, using the Loki fault injector, that provides insight into the impact that correlated network partitions have on the Coda distributed file system. In the evaluation, Loki creates a network partition between two Coda file servers, during which updates are made at each server to the same replicated data volume. Upon repair of the partition, a client requests directory resolution to converge the diverging replicas. At various stages of the resolution, Loki causes a second correlated network partition, thus allowing the correlated partition's impact on the system's performance and availability to be evaluated. This thesis begins with the details of how Loki is used to evaluate a system. Then it presents a lightweight fault injection proxy as an approach to using Loki that is employed in the Coda evaluation. An overview of the Coda evaluation case study is provided. Finally, experimental results are presented.