15-712 Semester Project

See the ideas list (viewable inside CMU)

Projects should be done in groups of two or three. If you're looking for partners or ideas, also try out the 712 Project Wiki Page.

Some prior 712 class projects are described in the following techreports: CS-CMU-05-138, CS-CMU-05-201. Another good resource to look at is An Evaluation of the Ninth SOSP Submissions, or How (and How Not) to Write a Good Systems Paper (by R. Levin and D. Redell, in Operating Systems Review, vol. 17, no. 3, July 1983, pp. 35-40) -- it is available on the Readings page.

Proposal

(3 pages; single spaced, one or two columns, 10 point font or larger)
Describe the project idea/application, how it relates to the course material, what work must be done (suggesting how it can be partitioned among you) and what resources you will need (including software systems you already have access to). Concentrate on convincing us that it will pertain to the course, that you will be able to complete it, and that we will be able to evaluate it. The third page should be dedicated to providing an outline of your intended final paper, identifying the specific experiments to be run and what questions they will answer.

In your proposal, provide three types of goals: 75% goals, 100% goals, and 125% goals. Think of these as the equivalent of a B grade, an A grade, and a "wow!" grade. The goals can be dependent or independent of the prior goals.

Think also about providing answers to the Heilmeier Questions in your proposal, as appropriate.

Interim Project Report and Literature Survey

Due 4/2. The interim project report should contain:

Your updated introduction and motivation based upon feedback from your proposal and any changes you've made as a result of your research so far.
A re-iteration of your proposed goals, with explicit discussion about what progress you have made to date on those goals and what your timeline is for accomplishing the rest of them by the end of the semester.
A well-filled-out background and related work section citing the appropriate work from the literature.
A sketch of your evaluation section outlining your evaluation plan and listing (as simple as an itemized list if you wish) the major components you plan for your evaluation.

Feel free to include additional material in your interim report: If you wish to include, e.g., your design section or preliminary results, we will give you feedback to help ensure that your final report is great.

Project Management

You should strongly consider using either Subversion or Git to perform source code control for your project and the paper you write describing it.

I also strongly suggest writing your course project report using LaTeX. It is the de-facto tool in which most CS research papers are written. While it has a bit of start up cost, it's much easier to collaboratively write complex research papers using LaTeX than Word.

Writing Papers

An Evaluation of the Ninth SOSP Submissions - or - How (and How Not) to Write a Good Systems Paper by Roy Levin and David D. Redell
CSP project startup documents by John Wilkes
Collected resources about writing and research

Simulation

The most popular network simulators are ns-2 and Opnet. ns-2 is free, and CSD has some licenses for Opnet.

Testbeds and Emulation

Emulab

Emulab is a network emulation environment at the University of Utah. It provides racks of machines and programmable switches that can be configured to form mostly-arbitrary network topologies, with controllable delay and loss between nodes. It's a great way to test real programs in repeatable conditions or at scales that you can't get on your own.

Emulab also provides a set of wireless nodes that you can control, located around their building. They also have an experimental mobile robot testbed that could make for fun projects (the robots are designed to move computers and radios around so that you can do repeatable experiments involving mobility).

If you wish to use Emulab for your project, you should join our class emulab project.

Planetlab

PlanetLab is a large-scale, distributed collection of machines that can be used for experiments and measurements. It has about 594 nodes scattered over 282 sites. The machines run something linux-ish that you can login to and run programs, and there exist a variety of utilities for automatically distributing software to the nodes, running programs, and so on.

You must sign up to use a PlanetLab account. Please only sign up if you're going to use the account, since it imposes some management overhead on people not involved with the course.

Wireless Nodes

Dave and some of the other faculty have a handful of various wireless nodes, cards, old laptops, etc., available that you may be able to use for your projects.

moteLab

moteLab is a wireless sensor network deployed in the CS building at Harvard. It has 30 sensor motes that users can upload system images to, collect data from, etc. To use moteLab, you'll need to send them an email (listed on their web page). Please CC: dga on the mail.

MistLab

Mistlab is a wireless testbed with 60 Mica2/Cricket nodes (sensor boards, much like the motes in moteLab) distributed over the 9th floor in the Stata center at MIT's Computer Science and AI Lab. For access, see their web page and send them an email, and please CC: dga on the mail.

Routing

The Click Modular Router
XORP the eXtensible Open Router Platform
Quagga - an open source routing suite

Measurement

Scriptroute (note: runs well on planetlab and the RON testbed)

Analysis

Books: Raj Jain, the Art of Computer Systems Performance Analysis - a very good overview of lots of mathetmatical techniques and queueing bits, aimed at a systems audience

Misc Tools

Profiling

perf allows to analyze performance by monitoring a set of events (e.g., hardware/software counters, cache misses, etc.). It has a simple command line interface.
gprof is a profiling tool. It allows to display the call graph and count how many times a function (or code line) is executed, by sampling the program at fixed intervals. Potential inaccuracies and I/O operations should be taken into account.
strace is a debugging tool that allows to trace the system calls (together with the arguments) made a program. It also allows to monitor the signals that a process receives. It a simple command line interface.
ioprofiler monitors a set of I/O syscalls and reports file access information, such as name, part, access latency.

Debug

ioreplay reproduces the recorded I/O workload of a process, without the need to run it again.
valgrind allows to detect memory and thread concurrency bugs through dynamic binary instrumentation. It integrates tools like Memcheck (to detect bad memory accesses, memory leaks, etc.), Cachegrind (to simulate first and last level caches and provide info such as cache reads/writes, cache misses, etc.), Callgrind (to record the function calls in the program), Helgrind (to detect synchronization errors like deadlocks, data races, etc.). It has a simple command line interface.
dbug uses an interposition layer, to control the execution of a distributed and multi-threaded system, and an arbiter to select the execution order of concurrent function calls. It implements a systematic testing method that allows to check for deadlocks, non-reentrant function call and violations such as segmentation faults.

Distributed Computing

zeromq is a messaging library that allows to build distributed or concurrent applications.
bftsmart is a fault tolerant state machine replication library. It allows to build dependable distributed applications, that can tolerate arbitrary failures (possibly malicious) of a predefined fraction of the system's components.
ganglia is a distributed monitoring system. It allows to collect performance information locally, and then to propagate and perform scalable aggregation for central monitoring.

Last updated: Thu Oct 10 21:54:23 -0400 2013 [validate xhtml]