next up previous contents
Next: 5.2 Experiments Up: 5. Evaluation Plan Previous: 5. Evaluation Plan

   
5.1 Criteria for Success

Given the above, these are the criteria I have set for testing the success of my work:

1.
I should be able to generate at least one working, useful, and non-trivial file system in each of the categories of stateless, in-core, and persistent. I intend to generate the following FiST file systems: Crossfs  (Appendix sec-appendix-typical-stateless-crossfs), Cryptfs (Appendix sec-appendix-typical-in-core-cryptfs), and Cachefs (Appendix sec-appendix-typical-persistent-cachefs).

2.
For each such kernel level file system generated from a FiST description, I should be able to generate a user-level file system that runs in Amd.

3.
The same FiST inputs should generate working file systems on at least three different Unix operating systems. I intend to produce code for Solaris (SVR4 based), FreeBSD (BSD-4.4-Lite based), and another operating system that has an established vnode interface, but is sufficiently different from ``pure'' SVR4 or BSD (for example HP-UX, AIX, or Digital Unix).

4.
The overhead of interposition should be comparable to that of previous work on stackable file systems, and should not exceed 10% for Wrapfs. See Section 3.6.2 for details of current performance.

5.
I should be able to show how to write FiST descriptions for a variety of other file systems.


next up previous contents
Next: 5.2 Experiments Up: 5. Evaluation Plan Previous: 5. Evaluation Plan
Erez Zadok
1999-12-07