Design and implementation of wavescope storage manager and access scheduler

[摘要] In this thesis, I designed, implemented, and analyzed the performance of an optimized storage manager for the Wavescope project. In doing this, I implemented an importation system that converts CENSAM data into a format specific to the processing system and cleans that data from measurement errors and irregularities; designed and implemented a highly efficient bulk-data processing system that is further optimized with a parallel-processor and disk access reorderer; carefully analyzed various methods for accessing the disk and our processing system, resulting in an accurate and predictive system model; and carefully ran a set of different applications to analyze the performance of our processing system. The project involves low-level optimization of Linux disk I/O and high-level optimizations such as parallel-processing. In the end, I created a system that is highly optimized and actually usable by CENSAM and other researchers.

[发布机构] Massachusetts Institute of Technology