Ethics/Interview Notes

John Schade
1. What do you know about computer clusters (more specifically, Beowulf clusters)?

They allow the potential for significant computing power for those who know how to use it.

2. Describe your own knowledge of computer programming.

I have none at all. I don't have the time to learn a language.

3. How do you think you would use the cluster (personal projects, class assignments, student research)?

More for personal projects and student research (If the students are willing to learn how to use the cluster). As an ecologist, I would use the computing power to help answer evolutionary questions.

'''4. Do you envision a team of Computer Science student "consultants" helping you with a project, or do you envision you or a group of students from your own department taking full control of the programming? If a team of consultants was not provided, would you be less likely to use the cluster?'''

Yes, a consultant would be required and the lack of one would be a disincentive to use the cluster. We would need programmers to help us.

5. Place the concerns of "Reliability", "Speed", "Being up-to-date", "Ease of Use", and "Security" into the categories of "Most Valued", "Moderately Valued", and "Less Valued".

Most Valued: Speed, Ease of Use (Tie) Less Valued: Being Up-to-Date

6. Assuming a solid documentation base existed for the cluster, would the use of command-line interfaces instead of intuitive graphical user interfaces deter you from promoting use of the cluster to your students?

Although the command-line is a deterrent, if a consultant were available to help enter commands, it'd be ok. A willing student could also learn how to use the command-line.

'''7. What kind of uptime do you expect the cluster to have? Would periodic/weekly downtimes be reasonable?'''

Some computing projects we would have could take weeks to complete. Therefore it'd be desirable to have the cluster running for weeks at a time. Period downtimes are ok if nothing is scheduled to be running for a long time.

'''8. Do you have expectations about maintenance or turnaround time for fixes? If so, what? How, if at all, would these expectations differ during summer and breaks?'''

Problems pop up spontaneously for me, so I'm the type of guy who would want it fixed as soon as possible. I'd prefer less than an hour, but more than that would be reasonable.

9. Do you have questions for us about the cluster?

When you finish your work with the cluster this term, will the cluster ever "get better"?

Olaf Hall-Holt
1. What do you know about computer clusters (more specifically, Beowulf clusters)?

There are 2 Clusters: A)Castaway--set of P3 CPU's--designed for prototyping B)Helios--mainly used for production purposes

2. Describe your own knowledge of computer programming.

Not a whole lot of parallel programming. I am more suited for the algorithmic side of programming.

3. How do you think you would use the cluster (personal projects, class assignments, student research)?

I'm not sure I want to invest the time to figure out how the cluster works. Therefore, I probably wouldn't use it for personal use or research unless there was someone else (probably student) who was interested in it and could handle the bulk of the technical work.

'''4. Do you envision a team of Computer Science student "consultants" helping you with a project, or do you envision you or a group of students from your own department taking full control of the programming? If a team of consultants was not provided, would you be less likely to use the cluster?'''

Yes, a consultant would be required and the lack of one would be a disincentive to use the cluster. I would be "much less likely" to use the cluster without a consultant.

5. Place the concerns of "Reliability", "Speed", "Being up-to-date", "Ease of Use", and "Security" into the categories of "Most Valued", "Moderately Valued", and "Less Valued".

Most Valued: Ease of Use Moderately Valued: Speed Less Valued: Security

6. Assuming a solid documentation base existed for the cluster, would the use of command-line interfaces instead of intuitive graphical user interfaces deter you from promoting use of the cluster to your students?

Whether command-line or GUI, I would be equally likely or equally unlikely to use the cluster

'''7. What kind of uptime do you expect the cluster to have? Would periodic/weekly downtimes be reasonable?'''

Periodic downtimes are acceptable. 50% uptime (he laughs)

'''8. Do you have expectations about maintenance or turnaround time for fixes? If so, what? How, if at all, would these expectations differ during summer and breaks?'''

My expectation would be, if I use the cluster, to have a maintenance/consultant that could get through the hurdles of parallel programming. Of course, it would have to be someone who WANTS to use the cluster and has extensive knowledge of the cluster. I will be largely devoted to my research, but if a student were able to build up a potentially useful application for the cluster, I would of course be inclined to use it.

9. Do you have questions for us about the cluster?

Have you thought about a "SchemeMPI", or similar where all functional programming would be used for the cluster. THis may speed things up considerably. Check up on "Termite" or "Erlang". I think there is a lot of potential application for functional programming in the cluster.

Steve McKelvey
1. What do you know about computer clusters (more specifically, Beowulf clusters)?

Poor man's supercomputing tool, linux boxes connected on a ethernet, allows for parallel processing.

2. Describe your own knowledge of computer programming.

"A good deal", know Java, Scheme, Fortran, C, C++, Basic, Pascal, APL, bash, emacs lisp.

3. How do you think you would use the cluster (personal projects, class assignments, student research)?

Personal research as well as with students in operations research, currently use a student version of commerical software and a proprietrary program providing a scalable version of the same functionality.

'''4. Do you envision a team of Computer Science student "consultants" helping you with a project, or do you envision you or a group of students from your own department taking full control of the programming? If a team of consultants was not provided, would you be less likely to use the cluster?'''

"Pictures consultants", would be just as likely to use it, but would use it less often, students would be less likely to use it.

5. Place the concerns of "Reliability", "Speed", "Being up-to-date", "Ease of Use", and "Security" into the categories of "Most Valued", "Moderately Valued", and "Less Valued".

Reliability and Speed are paramount, followed by being up-to-date and ease of use, security is not important (all based upon current project).

6. Assuming a solid documentation base existed for the cluster, would the use of command-line interfaces instead of intuitive graphical user interfaces deter you from promoting use of the cluster to your students?

Command line interface is not a deterrent.

'''7. What kind of uptime do you expect the cluster to have? Would periodic/weekly downtimes be reasonable?'''

Week of uptime is acceptable, published schedule and 3-5AM ish hours much preferred, don't expect any 6 day computations, not effective pedagogically.

'''8. Do you have expectations about maintenance or turnaround time for fixes? If so, what? How, if at all, would these expectations differ during summer and breaks?'''

"Expectations". During research, flexible, but for students, there would be projects with deadlines, need a near-gaurantee of uptime during the critical times.

9. Do you have questions for us about the cluster?

What software exists to take advantage of parallel processing?

If/when the cpu time begins to become a commodity, what plans are there to schedule?

Desires: Sparce matrix numerical packages, software libraries

Doug Beussman


1. What do you know about computer clusters (more specifically, Beowulf clusters)?

Yes - regular computers linked together to give more power than regular multiple processing.

2. Describe your own knowledge of computer programming.

Basic, not adept at C, c++, Java. Lab based view.

3. How do you think you would use the cluster (personal projects, class assignments, student research)?

Protein database searching 17K-20K proteins, search should take 5-20 seconds. No real use today, perhaps down the road.

'''4. Do you envision a team of Computer Science student "consultants" helping you with a project, or do you envision you or a group of students from your own department taking full control of the programming? If a team of consultants was not provided, would you be less likely to use the cluster?'''

CS students have potential overlap between chemistry and CS majors.

5. Place the concerns of "Reliability", "Speed", "Being up-to-date", "Ease of Use", and "Security" into the categories of "Most Valued", "Moderately Valued", and "Less Valued".

If being up-to-date refers to protein database data and not applications, then the most valued are Reliability, Being Up-to-Date, and Ease of Use; Speed is moderately valued, and security is less valued, as he is not doing anything that needs security.

6. Assuming a solid documentation base existed for the cluster, would the use of command-line interfaces instead of intuitive graphical user interfaces deter you from promoting use of the cluster to your students?

A GUI is preferred for classroom use, although a command line is acceptable for smaller groups (i.e.: research students).

'''7. What kind of uptime do you expect the cluster to have? Would periodic/weekly downtimes be reasonable?'''

Periodic downtimes are acceptable, but daily/weekly schedule - published ahead of time is useful.

'''8. Do you have expectations about maintenance or turnaround time for fixes? If so, what? How, if at all, would these expectations differ during summer and breaks?'''

Standard turnaround time of a Day, a week or longer is detrimental to research, breaks are least important, while summer is most important. Break turnaround could extend to Several Days or a Week, but Summer turnaround should be Less than a Day. Train others on routine stuff - power outage protocol others can follow if trained.

9. Do you have questions for us about the cluster?

What limitations exist for software? Windows programs acceptable? (No - linux only). Looking at alternatives to Protein Prospector in parallelized software.

Anne Walter


1. What do you know about computer clusters (more specifically, Beowulf clusters)?

They can be made from old or new components, are used for parallel processing, can be used in various configurations (use all, half, a quarter). The people who run them are usually of good humor.

2. Describe your own knowledge of computer programming.

Used to be good at Fortran 4, some BASIC, mostly on cards. Then started hiring programmers.

3. How do you think you would use the cluster (personal projects, class assignments, student research)?

Interested finding out how she can integrate it into classes. Looking at it from an administrative (department chair) perspective.

'''4. Do you envision a team of Computer Science student "consultants" helping you with a project, or do you envision you or a group of students from your own department taking full control of the programming? If a team of consultants was not provided, would you be less likely to use the cluster?'''

Envisions a CIR-like structure for users of the cluster. Would probably not recommend using the cluster without help from CS students, unless student working on project knows enough about programming. Looking more for cooperation between her students and the CS students.

5. Place the concerns of "Reliability", "Speed", "Being up-to-date", "Ease of Use", and "Security" into the categories of "Most Valued", "Moderately Valued", and "Less Valued".

Reliability and ease of use most important, especially for areas in which lots of users will be using the cluster (BLAST). Security is least important except in preventing her from killing the cluster by accident. Moderate value placed on speed and being up-to-date.

6. Assuming a solid documentation base existed for the cluster, would the use of command-line interfaces instead of intuitive graphical user interfaces deter you from promoting use of the cluster to your students?

Deterrent, especially for BLAST users.

'''7. What kind of uptime do you expect the cluster to have? Would periodic/weekly downtimes be reasonable?'''

Most important that classes planning to use cluster set up schedule ahead of time and that the schedule is honored (no downtime during those periods). Other times more flexible.

'''8. Do you have expectations about maintenance or turnaround time for fixes? If so, what? How, if at all, would these expectations differ during summer and breaks?'''

Summer, if there are projects, turn around time should be really fast. For normal failures, should be relatively fast any time, worse failures, may take longer. Not expecting quick turn around during midterms or finals, or over breaks.

9. Do you have questions for us about the cluster?

How will we screen projects and schedule runs? Are there names other than production and development for the clusters? Are there any other developments at the moment?