Extending the Snarl File Repair Component for Distributed Storage Systems
Abstract
This thesis extends the Snarl file repair component for distributed storage systems, and evaluates extensions. Snarl is an application using alpha entanglement codes to improve recovery rates of content stored in distributed storage systems. This work extends Snarl by adapting it such that it can be used for other systems than the Swarm network, which it was limited to in the original implementation. The extensions include an abstraction layer making it simple to extend Snarl to be used with other systems, and changes to Snarl’score algorithms to separate tasks into distinct processes. Additionally, an extension for the InterPlanetary File System (IPFS) is added. Finally, the newly added IPFS extension is evaluated. File recovery rates and network overheads are measured for several different percentages of data loss and peer loss in the network, demonstrating that Snarl performs well in these scenarios, and provides much better file recovery rates than simple data replication with the same storage overhead.