We are using Piazzza. Make
sure get enrolled in this class on it; it will be the primary way of
making announcements, advertising review sessions, and answering
questions. Just follow the link, create an account, and add CS244B. If
you need to contact the staff directly, our addresses are above.
Many systems and applications beyond a relatively modest scale end up being
distributed for a variety of reasons, including fault-tolerance, performance,
security and the geographical nature of the requirements.
Although the common case of a distributed system is separate computers
interconnected by the Internet or some networking technology,
many of the challenges and solutions come into play as soon as the application
consists of two or more processes that do not share a common address space.
This course explores the issues to address in designing and implementing
distributed systems with a particular emphasis on how to deal with the
shared state between separate processes within such a system.
Topics include: distributed shared memory, object-oriented structuring of
distributed systems, real/logical/virtual time,
directories and naming, distributed file systems,
authentication and security, distributed process scheduling,
presentation protocols and communication support for distributed systems.
The material complements network-layer courses by building on the transport
layer to provide higher-level applications and services.
The lectures are based on notes by the instructor,
typically one chapter per 2 lectures.
The chapter for the next week will be posted on-line in postscript and pdf
by Sunday night of the week of the lectures, if not before,
noted as updated to this year.
The last year's version is available now.
There may be a guest lecture or two that depart from this pattern
and will be noted as they arise.
Distributed Shared Memory 2013
Structuring of Distributed Systems 2013
Clocks: Real, Virtual and Logical 2013
Directories and Naming 2013
Accounts and Authentication 2013
Transactions, Agreement and Reconiliation 2013
Distributed Storage Systems 2013
Future Directions and Issues 2013
Current Events (still from previous years)
Below are a list of current events/readings related to the course. We expect you to have read the articles and
understand the basic problems raised, but not specific details.
Academic Paper in China Sets Off Alarms in U.S. Posted 4/1/2010
FAA exec offers blunt, scary assessment of its network security Posted 5/1/09
GhostNet is watching Posted 5/1/09
As augmentation to this material,
there may be one or two research papers that are listed as readings at the
end of each chapter and/or listed together with the chapter material
that you are also responsible for, unless otherwise stated in class.
The chapters also list research papers as references,
meaning they are optional reading and not required in any way
to do well in the course.
Students should be able to access these papers online using Stanford's
site license for the ACM digital library from any computer with a
Stanford IP address. If you have difficulties accessing the papers
online, please contact the CA.
B. Stone-Gross et al., Your Botnet is My Botnet: Analysis of a Botnet Takeover UCSB Technical Report Note: not covered on final
P. Helland, Life beyond Distributed Transactions: an Apostate’s Opinion
G. DeCandia et al. Dynamo: Amazons Highly Available Key-value Store SOSP 2007
Smart Cheaters Do Prosper: Defeating Trust and Reputation Systems AAMAS 2009
Public Key Certificates Considered Harmful
This morning's MD5 attack - resolved Article related to Public Key Certificates Considered Harmful from student, Max Martynov, and his comments
D.Cheriton and D. Skeen,
Understanding the Limitations of Causally and Totally-Ordered
SOSP'93, Dec. 1993.
Internet Time Synchronization: the Network Time Protocol,
IEEE Transactions on Communication.
D. Jefferson et al,
Distributed Simulation and the Time Warp Operating System,
S. Baset and H. Schulzrinne
An Analysis of the Skype Peer-to-Peer Internet Telephony Protocol
Designing a Global Name Service,
5th PODC, August, 1986.
Demers et al.,
Epidemic Algorithms for Replicated Database Maintenance,
There is no text that really corresponds to the material and focus of this
However, if you really want a textbook treatment,
you can refer to: Distributed Systems: Concepts and Design
by George Coulouris, Jean Dollimore, and Tim Kindberg, Addison Wesley,
There is expected to be 2-3 programming assignments with the 3rd possibly
morphing into a homework assignment because of time constraints.
The first programming assignment is planned to be posted by the first lecture.
Details to follow.
- Assignment 1 review session: Friday, April 5. 4.15pm to 5.05pm @ Huang 18
Assignment 1: Mazewar
Mazewar Review Session Slides are here.
Assignment 2: Replicated Filesystem
Replicated Filesystem Review Session Slides are here
Working on assignments
You may work on the assignments on any Unix machine with a modern C/C++ compiler and
debugger. But since we will evaluate your assignment on one of the Linux
the myth cluster in the Gates building, we
strongly recommend that you develop and test your code on one of these machines.
You may find the Unix command
Exams are closed book, closed notes.
The final is normally 45 percent of the grade, midterm 15 percent
and assignments the remaining 40 percent.
ssh -X myth.stanford.edu to be useful at your Linux or OS X prompt.
The midterm is scheduled for May 2, 7:30-8:45pm. Location Braunlec
Off-campus students are encouraged to come onto campus for the exams
if at all feasible.
If necessary, remote SCPD students may
take the exams at their company sites with a proctor;
Stanford SCPD is responsible for distributing exams to proctors as well as returning the completed exams.
These sample exams should aid you in your studying:
Late assignment policy
You are allowed 3 late day credits for the quarter which you may
use as needed to accommodate your schedule.
Otherwise, late assignments may be penalized 10% for each 24-hour period or
fraction thereof (including weekends) that they are late.
Beyond 3 days, they are graded/credited only at the sole discretion of the TA.
Please try to make prior arrangements with the TA or instructor
if you anticipate any problem with this policy.
Since lectures are made available via Stanford Online
on the same day they are taped,
SCPD students are expected to follow the same schedule
as the rest of the class.
This means that the above late assignment policy applies to
SCPD students as well; exceptions are not granted due to students' work obligations.
It should go without saying, but because computer science project courses have a long
and ugly history of honor code violations, we will say it anyway:
The Stanford honor code
applies to all work done in this course. All work you submit must
be your own. Suspected violations of the honor code will be investigated and referred to the
Office of Judicial Affairs.
Honor code violations are a serious matter, and being found guilty of one can ruin your
academic career. Review the honor code.
If you ever find yourself uncertain about how it applies to your situation, ask.
Asking what you might think is a silly question is
better than risking your career.
The instructor and CAs try to help students who get behind or encounter
There is more flexibility here than in dealing with violations of the
No incompletes are given in this course, so make sure you determine before
the drop deadline whether you can complete it satisfactorily.