Duke ITAC - February 3, 2005 Minutes

DUKE ITAC - February 3, 2005 Minutes

Minutes

February 3, 2005

Members present: Owen Astrachan, Mike Baptise, Pakis Bessias, John Board, Shailesh Chandrasekharan, Paul Conway, Dick Danner represented by Ken Hirsch, Nevin Fouts, Tracy Futhey, Christopher Gelpi, Michael Gettes, Daron Gunn, Guven Guzeldere, David Jamieson-Drake, Roger Loyd, Paul Harrod represented by Alfred Trozzo, Greg McCarthy, Melissa Mills, Kyle Johnson, Lynne O'Brien, George Oberlander, Mike Pickett, Rafael Rodriguez, Scott Smith, Dalene Stangl, Molly Tamarkin, Robert Wolpert, Steve Woody

Guests present: George Ward, OIT; Jim Rigney, Computer Store; Phil Lemmons, News & Communications; Ed Gomes, Perkins Library; Matthew Drummond, Duke Card; Sarah Roberts, OIT; Dan McCarriar, OIT; Jess Mitchell, OIT; Heather Flanagan, OIT; Chris Cramer, OIT; Bob Currier, OIT

Start Time: 4:04p.m.

I. Review of minutes and announcements

-Computing Facilities and Infrastructure committee

Mike Pickett says Peter Lange talked with Tracy to see if we should replace infrastructure we've got: is there anything, whether space or networking software, et cetera that is likely to wear out or run out in the next 10 years? This isn't new stuff; it's if we can keep the stuff we have running. I'm going to chair the committee, and after 6-8 weeks the committee will come back with a list of things we should be watching, maybe things like our cable TV infrastructure, wireless networking capacity and bandwidth.


-Calendar and Mobile Devices Pilot

Mike Pickett says we’ve been looking at group calendaring systems forever, and have narrowed it down from a group of 6 calendars to 2: Meeting Maker and Oracle Calendar. We do have an instance of Oracle up, but no one on campus has day-to-day experience with it. A small group of us got together to see how we could get more experience. We determined that some group needed to volunteer, and it turned out to be OIT. Starting sometime in march, we’re going to try it for a while on a pilot basis. Others (non-OIT) can participate in the pilot, but the group will be kept relatively small. We’re shooting for a recommendation sometime in the end of May.

John Board asks by switch over to you mean you are using Meeting Maker now?

Mike Pickett says OIT uses Lotus Notes.

Michael Gettes adds that by switch over we don’t mean that we’re getting rid of Lotus Notes, we’re just going to try out Oracle in addition to it.

David Jamieson-Drake asks is the idea that this would replace the calendaring system for those still using Lotus Notes as their primary mail system?

Michael Gettes says the testing won't just be OIT. There are others we will solicit to help us determine how it works in other environments. As to replacing Notes, Michael points out the Health System still plans to use Notes so it will remain at Duke regardless of whether Oracle calendar tests out successfully or not.

II. Microsoft Campus Agreement update -Tracy Futhey, Jim Rigney

Tracy says right now everyone who buys Microsoft software buys individual licenses purchased at a good campus discount rate. Still, there’s a purchase and associated paperwork for every acquisition. Microsoft has a campus agreement program where for a certain amount of money a university can get access to a set of software. It’s a sort of ‘rental’ software; you don’t keep it forever. The amount of the annual payment may not be different from what we pay anyway, and it would reduce the amount of paperwork going on. The challenge with exploring this is the change it causes in what money is being spent from which pocket. The savings go to the individual cost centers, even though the university has to pay a big bag of money. I want to introduce the notion to people to see if there is enthusiasm for this idea, or resistance, and get and idea of what things would look like if we had access to a suite of software packages.

John Board asks will this mean that people will get bare bones machines without an OS and basic applications?

Tracy says probably not, since most machines have OS bundles into purchase cost.

George Oberlander says most machines will need updates before they wear out.

Michael Gettes adds that many people don’t update, which causes problems.

Rafael says when buying computers, we get pre-loaded software from Microsoft. Would we then get a break on buying that?

Tracy says no.

Rafael says the strategy seems to order them with Linux installed, which is cheaper, and then install the operating systems.

Chris Gelpi says it strikes me that the system you’re describing is similar to what I’ve seem in political science. I gave my laptop to the support person, who re-imaged it and put on all the A&S stuff: Would this be like that?

Tracy says yes, we’re giving people the ability to update without having to go and buy all of the new versions again, though someone would still have to do the work of installation.

Robert Wolpert asks, would the software we would get would be different from what we would get otherwise? Are there expiration dates?

Tracy says no, it’s a self-policing system. Also, this software can be used on home computers as well.

Molly Tamarkin says, I’m not sure if this would be useful unless it would save us money.

Kyle Johnson says his concern is seeing Microsoft’s rotation schedule going from 1 to 2 to 3 to 6 years; I’m interested in seeing how this plays out in a 6 yr period rather than one. What if we don’t update for that time? It may not mean anything.

Michael Gettes says, what if we decided that we needed everyone on WinXP today? What are the costs to buy that for everyone versus using this program?

John Board asks if anyone has thought in terms of Microsoft’s announcement that it will only sell software to Microsoft registered machines?

Melissa Mills says we have a standard charge for new computer includes upgrading Microsoft products. It’s unusual for someone to request an upgrade outside of the cycle of getting the new hardware. I’m not sure that we want to commit ourselves to paying something every year that gets us locked into something when we might want to be looking at different products.

Greg McCarthy says he thinks it’s a great idea: if something comes in with software, the first thing we do is install all new versions.

Rafael Rodriguez asks, would we have to do it for the entire Duke community?

Tracy says no, we could do it for select parts of the community.

Robert Wolpert asks, is there a broader sweep of software this way than we would get on normal computers, beyond OS and Office?

Tracy says there are add-in options, but it raises the cost, and this is really only cost-effective with things broadly used.

John Board asks, if this is going to be ready for next year, when do we need to decide if we’re going to do it?

Tracy says probably sometime in the next two months.

Network bandwidth & "Traffic Ticket" Update Dan McCarriar, Chris Cramer, Bob Currier

Chris Cramer says this is the third semester we’ve done bandwidth ticketing, and I think it’s been fairly successful. Two years ago we had problem in ResNet. There was a lot more demand on ResNet than it provided capability for, so performance in general was terrible. Rather than limiting students up front, we said, lets take this as a chance to educate. Students are given a fairly large cap: if any student uses 5 Gbs of bandwidth in a given day, they’d receive a traffic ticket. If a student receives 5 tickets in a given semester, their uploading bandwidth would be limit rated to 64 Kb. Inbound is not restricted at all.

One of the decisions was that if a student came back and says they had an educational reason for serving up all this data, we would allow this. Two students initially claimed this, but it turned out they just misunderstood the different between upload and download.

Bob Currier says last semester had 88 ticket days (the number of days we actually issue tickets). There was a total of 790 tickets, fairly evenly divided between campuses, resulting in an average of 8.9 tickets issued per day. We’ve had 18 ticket days in the current semester and total of 187 tickets, with an average of 10.4 tickets per day. This is consistent with previous semesters, as tickets/day tend to be higher earlier in the term.

John asks if we have data on recidivism.

Chris Cramer says we maintain relatively little historical information.

Bob says as an unofficial, he believes several students have gotten 5 tickets in the past semester.

Daron asks if these are students who are knowingly transferring large amounts of data, or have their computers been hacked?

Chris says they do get some responses from students saying they don’t know how it happened, and we point them toward the Help Desk.

John Board says he takes those numbers as alarming, that we have that many students already this semester that were throttled, and are therefore ignoring our policy, so we are going to have to revisit this.

Tracy Futhey says the important thing will be to revisit this with the student body leadership, because they called for this, partly as an issue of equity of shared resources, and partly as an education campaign.

Robert Wolpert says we set the 5-ticket limit on an ad hoc basis. Would it be different if we set the ticket limit differently?

Chris Cramer says Bob, Dan, and I talked about this earlier, and the way we’re storing data we don’t know right now. We could as a project start keeping that information. There is an interesting question as to if one student or multiple students are repeat offenders.

Shailesh Chandrasekharan says from what I understand this is only for the students?

John Board says yes.

Chris Gelpi says, if we can limit their bandwidth, does it really matter that much?

Tracy Futhey says it has resolved the performance issue, but there is still the educational issue about appropriate use.

John Board says when this first came out one student tried to be clever and circumvent the process by setting up multiple ports. Do we still have problems with students trying to get around this?

Dan says there was one instance when a student tried to change their Mac address, but that was about it.

Michael Gettes asks if there is a cost to what we’re doing now that we should pay attention to?

Chris Cramer says there are some Help Desk costs in disabling peer-to-peer and dealing with hacked machines, which we’d have to deal with anyway.

Computing Labs - Solaris to Linux Stats and Update Mike Pickett, Sarah Roberts, George Ward, Bob Currier

Mike Pickett says with our charge of figuring out now much pain we would create for faculty and students if we got rid of Solaris labs, Sarah Roberts has done a lot with talking to faculty and other people in the community, while George ward is dealing with logistical side.

Sarah says we began a campaign of communication last semester in November and December by contacting primary users of Solaris clusters based on reservations of those clusters. We contacted these folks and gave them a general picture of what the state of this platform and the hardware was and solicited their input via a brief ViewsFlash survey, just a few questions. That went out to ten faculty members and we got six respondents. The responses were overwhelmingly positive; no one expressed any major concern with this at all. We had comments like “I think this is a great idea”, “Solaris is slow, Linux should be faster,” etc. We also noticed that these people had some experience with Linux already.

Then we talked to some campus administrators to present the survey and gauge their response, which was overwhelmingly positive. People were excited about the possibility of the move. The final move was to make an announcement in the Glove Box News mailing. So far, so good.

Since that time the steps we’ve taken are to move from the investigational mindset to putting it into a project plan and into a timeline format. We’ve come together as a team within OIT, reaching out to the tech groups. It looks like in mid-March we’ll start testing software and hardware including packages we need to have available on Linux machines, primarily commercial software. Another issue that has been an issue in the past is video cards, so we’re going to look at that early on as well. We’re going to go with Dell machines, essentially the same that serve as Windows machines in the labs. June will be the time for configuring new machines and beginning to decommission Solaris machines. The labs should be ready for second semester of summer school.

There are some technical issues already known that have to do with some incompatibilities. One problem has been solved, and a solution has been identified on another front.

Mike Pickett says when he heard this information that up until this point we’d talked about it as something tentative. Now I’d like to declare as a project.

Robert Wolpert asks if looking at dual or single processor machines.

George Ward says they are currently single processors. We didn’t see any gain from dual processors the last time we looked, but we can look into that.

[Bob Currier shows graphs on site license software usage.] Bob says there are two components, one if the log in utilization and the process utilization, currently being collected on the Solaris clusters.

Mike Pickett points out how much of the processes take place remotely.

Bob says what I’m doing is collected everything seen on Solaris machines out there on a machine-to-machine basis. Matlab is by far the most utilized application. I thought I’d see a lot of utilization of Mathematica, but we haven’t seen that much. My assumption is that most people using it are using laptops rather than cluster machines.

Tracy Futhey says presumably we can put a message on the remote log-in so people know we’ll be switching to Linux.

Sarah Roberts says that’s what they are planning to do.

Robert Wolpert asks if there are any data preservation issues?

Chris Cramer says yes there are, potentially.

Heather Flanagan says right now there is a remote login server, Godzilla, and we’re looking to keep that box around as something people can log into so people aren’t completely bereft of Solaris.

Michael Gettes says recognizing the issues with Solaris, it looks like more Solaris machines are being used now than Linux machines. I assume this is due to the cluster locations and availability?

Mike Pickett says yes, this is true.

Tracy Futhey asks if these graphs suggest that we have too many physical desktops running Solaris or Linux? If we have 96 laptops and average usage tends to be 30, peaks around 50, do we have more machines than we need?

ePrint Stats and Enterprise Planning Update Mike Pickett, George Ward, Matt Drummond

Mike Pickett says ePrint is the ability for students to print to common printers. It doesn’t actually print, it just goes into a queue onto the printer. Students swipe their cards, and see their queue of documents and decide which to print. It’s been extremely successful, very desirable service for a lot of people. Other people have interest in this: Molly Tamarkin has been vocal in getting this available in Nicholas. We’ve had a couple meeting to discuss technical architecture, and Pat Driver has been working with the tech team. It looks like everyone wants to make this an enterprise service, meaning there would be a way for people to bring up their own printers and print queues. The general thought is that sometime over the summer we will be rolling this out. We’re going to pilot some of this architecture. Molly will help us with this; if there are others with interest, let us know.

We’re currently running this with no policies; if you have a DukeCard, you can print, and they do: lots. We have nothing that says, “This is a reasonable amount of printing,” or something that says, “This is really for students, or for faculty and employees. “ This issue came up in our requirements discussion because some people said, “We don’t want to let everyone print on this, only our graduate students, and not everything they want to print.”

[Mike passes out data sheets.]

Mike says, there is one person printing about 150 pages per job, 8 jobs per day. Is that reasonable? I don’t think so, but we don’t have any policies.

Paul Conway asks if he would like this group to brainstorm policies?

Mike says if there were to be quotas the question is who would be appropriate to help think about this? As a service organization we need to think about to what extent we keep this wide open for employees, visitors, etc.

John Board asks do we have a longer view? Are we using less paper?

George Ward says were using more.

Matt Drummond says there was an 11% reduction in jobs that would have been printed without the queue mechanism in a given week.

Mike Pickett says growth in printing is still happening, but we’re saving about 5 reams of paper a week.

Melissa Mills asks if we want to create policies that have to do with the purpose of the printers?

George Oberlander says that if we took the cost discussion out of this discussion, it wouldn’t come in except as an environmental concern. We can’t decide if this works unless we take cost into it, to decide if OIT should be paying for it.

Michael Gettes says we also need to look at printing habits, like printing in smaller sizes to save paper.

Paul Conway says there is a trend to put course packs for free online that are now being printed out instead of photocopies. IT goes to the subsidy question for the learning environment we’re creating here. IT may not be OIT’s job to recommend that, but rather to suggest a process for that. The costs are significant.

John Board asks what are the costs for paper for a year?

George Ward says I doesn’t know off hand, in I’d guess the hundreds of thousands.

Robert Wolpert says he would like to suggest there is a benefit to not charging; it doesn’t amount to a lot of money, but it amounts to a lot of frustration and hostility that may outweigh the money we could save by cutting down on paper use.

Nevin Fouts says the ability to pull things out of a queue mitigates the concern of personal privacy; making it difficult to trust that would be a cost.

Michael Gettes says it may be reasonable to look at this as with bandwidth ticketing: just deal with the top users.

Other Business

None.

End Time: 5:30 p.m.