Comparative analysis of differential algorithms used for generating delta update packages

Differential algorithms, frequently referred to as delta algorithms, represent key instruments in enhancing the efficiency of over-the-air (OTA) update procedure. One such tool that benefits from this approach is the Remote Device Fleet Manager (RDFM) - Antmicro’s open source framework that supports OTA delta updates among other features. This thesis involved a comparative analysis of delta algorithms, with a focus on finding a better alternative to rsync (up to that point used exclusively in RDFM) to minimize the size of delta packages generated by RDFM. The research evaluated the performance of various delta algorithms: rsync, fossil, xdelta, and go-delta, examining key metrics: delta size, delta encoding time, and delta decoding time. The algorithms were tested in realistic Linux update scenarios, generated by introducing changes to tailored Linux images using the Yocto Project. The results highlight the characteristics of each algorithm and the trade-offs between computational efficiency and delta compactness.

The analysis concluded with the recommendation of implementing the xdelta algorithm as an alternative to rsync to enhance the RDFM framework’s performance, which was done in post-thesis work and involved updates to the xdelta library, cgo integration with xdelta, and updates to RDFM itself.