CS 599: Projects, Fall 2000
Overview of Projects
Students should use the class project to explore a topic that we have
covered in the course more deeply. There are two types of projects:
surveys of papers on particular research topics, and projects that
involve experimentation with grid computing tools such as Globus,
Condor, Legion, scheduling tools, network performance tools, etc.
Final project reports should be approximately 8 to 10 pages in
length.
Important Dates:
October 12: Half-page project proposal due
November 9: One-page project progress report due
December 7: Final project report due
Survey Papers
Students writing a survey paper on a particular topic should read and
summarize at least 8 technical papers. Besides discussing previous
work on the topic, the ideal survey paper should also include an
evaluation of different techniques to identify which algorithms are
preferred and a discussion of outstanding research issues.
Students who do survey papers should work invidually.
Students are encouraged to choose their own survey paper topics based
on their interests. Examples of topics include:
(1) Resource management in grids using advanced reservations:
This paper would survey the issues in providing advanced reservations,
touching on the needs of schedulers in grid environments. The paper
could also look at examples of advanced reservations in network
environments and comment on extending these techniques to storage
systems and computational resources.
(2) Scheduling strategies for grid computing:
Present a survey of traditional scheduling algorithms for parallel
computers. Next, explain how scheduling algorithms must change in
grid environments. Then, present a survey of existing grid scheduling
technologies for scheduling, including GARA, AppLes, Condor
matchmaking, Legion scheduling, etc. Compare the various approaches
and identify the most promising approaches.
(3) The use of objects in Legion:
Discuss the Legion object-oriented approach to managing computation,
network and data resources. Use examples from various papers to
describe the impact of the object model on the architecture, security
model, I/O performance, etc. Discuss the advantages and disadvantages
of using a model like Legion objects instead of the Globus
bag-of-services model. Present real-world examples of the use of Legion.
Experimental Projects
Students can choose to experiment with grid-related software. In some
cases, they may conduct projects under the guidance of a member of the
globus team. In other cases, the students may independently
experiment with software. Note that experimental projects are not
limited to Globus tools. Students are welcome to use other tools such
as Condor, Legion, etc.
Students are allowed to work in groups of two. With special
permission, students can work in groups of three. However, with more
students working on the project, the expectations of results will
increase.
Note that members of the Globus team will have limited time this
semester to support students. Students who choose to do experimental
projects should not expect a high level of user support!
At the end of the semester, the students should produce a project
report. It should include an overview discussion of the tools the
students investigated, a detailed description of the experimental
plan and methodology the students used, and a clear presentation of
whatever results were generated during the work.
Again, students are encouraged to define their own projects based on
their interests. Examples of possible projects include:
(1) Using globus tools to perform replica selection
This project would require the students first to learn the LDAP
directory access protocol. Next, the students would need to learn
about globus information services, including the MDS, the replica
catalog, and the Network Weather Service. The students could
experiment with existing catalogs to locate replicas of data and to
select among them based on current network performance.
(2) Experiments with Condor
The students should
first learn about Condor. Next, the students might volunteer to
participate in the SETI at home cluster, and observe how their
machines are used. They could do overall research on SETI, explaining
its approach and results. Finally, if the students have access to a
cluster of machines, they could create a Condor pool and experiments
with submitting jobs to the pool. This might involve having
applications and resources advertise their needs and capabilities
using classified ads, and matching these via the matchmaker protocol.