Project Teams for CS 5600
NOTE:
The Wiki page for the team projects has now been set up.
Any information on this page is obsolete. The current team members, etc.,
are to be found on the Wiki page.
These are the teams so far for the projects.
On Nov. 18, we will have permanently set the teams. Ideally, I'm
looking for teams of size three. I will also want the students
to come prepared with a tentative team plan. Part of the class
will be devoted to joint brainstorming.
If someone is looking for a more concrete project on which one can
make early progress, I can also suggest the newly added project:
"General API for DMTCP Coordinator" on the
projects page. A second, slightly harder project is:
"General API for DMTCP Plugin for Writing a Checkpoint"
I've also added some tentative remarks for each of the projects chosen.
All of these projects delve into the unknown. I provide a best
guess about the likely difficulties of the project. But there can
always be surprises. The evaluation will not be based
on my "Best guess" right now. It will be based on how the
project evolves.
The teams so far are:
- Checkpointing valgrind (valgrind attach)
-
Team members: Derick, Tristan, and Zack
Best guess: If one can diagnose the bug/issue for DMTCP and produce
a detailed plan for addressing it, that will be a success.
- Security: Multi-architecture Checkpoint-Restart
-
Team members: Ricky and Leo
Best guess: Very open-ended. If one can do small micro-experiments,
and then devise a coherent plan based on the knowledge obtained, that will
be a clear success.
- Checkpointing Hadoop (Big Data)
-
Team members: Haofan, Kan and Gabriel
Best guess: Relies on understanding the existing design of Hadoop
through reading its software. We can modify the design to create
a trivial back-end, so that we can concentrate on checkpointing
the front-end.
- Versioned symbols for ELF
-
Team members: Josh and William
Best guess: It relies on a careful understanding of the poorly
documented ELF standard. A lot of the difficulty will lie
in the details of the code (especially the elf.h and related files),
and using them correctly. I have some strong experience here, to help.
- Mesos: Fault-tolerant Resource Scheduling
-
Team members: Hriya and Jeet
Best guess: A lot of this relies on understanding the design
of Mesos. I am willing to create a hyper-linked version of
Mesos (similar to the xv6 and DMTCP source code for our course).
After understanding the design, the next question is how to do
a careful re-factoring of the existing software design.
(Anybody can write anything, given unlimited lines of code.
It takes skill to write compact, clear code.)
- Mesos: Load Balancing
-
Team A members: Devon, Rachit, Akshay, and Rahul
Team B members: Chris, Matt, and Mark
Best guess: Same remarks as for the other Mesos project.