CS 395T, Spring 2009: Topics in Secure and Distributed Storage Systems

Latest Announcements

3/20: Office hours schedule for rest of term

Mike's office hours are going to be on an odd schedule for the rest of the term. The schedule is below. As always, if you would like to meet at another time, please email to set up a time.

Mike's office hours will be held at

  • 2:00-3:00 PM on 3/26 (Thu)
  • 5:30-6:30 PM on 4/6 (Mon)
  • 2:00-3:00 PM on 4/16 (Thu)
  • 2:00-3:00 PM on 4/20 (Mon)
  • 2:00-3:00 PM on 5/1 (Fri)
  • 2:00-3:00 PM on 5/8 (Fri)

Mike will not hold office hours at

  • 2:00-3:00 PM 3/27 (Fri)
  • 2:00-3:00 PM 4/3 (Fri)
  • 2:00-3:00 PM 4/10 (Fri)
  • 2:00-3:00 PM 4/17 (Fri)
  • 2:00-3:00 PM 4/24 (Fri)

3/20: Happy spring break

See all announcements RSS 2.0 feed

Course Info


Storage is a classic subject in computer systems, and current trends are raising a fresh set of questions about this time-honored topic. One trend is the prevalence and professionalism of cyber-attacks on storage, including theft, mutilation, and denial-of-access. Another trend is the move by small and medium-sized organizations to "outsource" computation and storage by renting CPUs and disks from a better-provisioned provider. A third is people sharing data across organizational boundaries.

This seminar will cover questions raised by these trends. The weekly work will be to read, and be prepared to discuss, 2-4 papers, most of which will be recent research papers. Mike and Don will lead the first few paper discussions; after that, students will lead the discussions. Depending on interest, there may be a term project in which students are encouraged to undertake original research.

Topics to be covered will depend in part on students' interests. Possibilities include research questions such as the following: Should we replace the abstraction of a file, which has been ascendent for several decades, with something that explicitly incorporates a notion of provenance and revision history? How should we build a "utility storage" system that can serve tens of millions of users yet not require those users to make assumptions about the failure model or trustworthiness of the storage system? (Current solutions assume either limited scale or fail-stop failures.) Right now, the default configuration of a large-scale computing and storage infrastructure is "arrays of cheap PCs"; will upcoming changes in economics and technology affect that default and, if so, how?


Code of conduct

Please read the UTCS Code of Conduct; it applies to this course.


This is a seminar, and one of the purposes of a seminar is for students to think about, and discuss, research papers. My (Mike's) experience has been that such discussions work best when everyone in the room is engaged. As a result, we will have the following participation policy and implementation:


Last updated: Thu Apr 16 16:09:00 -0500 2009 [validate xhtml]