For most users, the file system is the most visible aspect of the operating system. By many it is perceived as the crucial component in the overall performance of a system. In this seminar we will study the evolution in the design of specific file systems and investigate the issues that are of importance for the performance of these file systems.

What are the key issues will we look at?

An important goal of the seminar is to gain understanding in the issues that are important in the performance of file systems. For this we will take time to understand how research into file system performance is done, how file system workload information is gathered and analyzed, and how we can use this type of information to reason about performance improvements to file systems.

Which File Systems will we study?

The systems that will be studied in detail are the Windows NT NTFS, the BSD UNIX fast file system,  the Sprite Log Structured file system, NFS, the Andrew file system, Coda, as well as a number of more experimental file system architectures.

What about caching?

Logging, caching and prefetching are all techniques used to improve the performance of file systems. We will spend considerable time in the seminar on the evolution of these techniques and on the current state-of-the-art.

Which subjects are not covered?

We prefer, with the limited time available for this seminar, an in-depth look at a small set of key performance issues and leave a number other, not less important subjects, for independent study or a follow-up seminar next semester.

The subjects not covered include: file systems disk layout, disks and disk arrays, replicated file systems, recovery,  parallel file systems,  NVRAM caching, tertiary storage, compression, encryption and security.

What about disks and RAID?

Unfortunately we have not enough time this seminar to also cover the exiting advances in disk technologies, sorry.

Is there also a practical side to all this?

Yes. One of our goals is to understand the virtues of augmenting the Windows NT NTFS with the performance improvements developed for the systems we will study. There are projects available for the analysis of trace data of the NT file system, for the development of a NT cache manager simulator and for implementing and testing new algorithms within Windows NT.