Figure 1 Figure 2
During the weeks of March 4 and March 11, 1996, a group of 25 Information Systems staff conducted a total of 480 random interviews of users who were working in the general use Athena clusters (i.e. clusters open to all MIT students, staff, and faculty). The aim of the study was to find out what kind of use students were making of the clusters, in terms of type of work being done, number of hours of login time, and software applications used. Some of the questions referred to the user's current login session, while others asked for an estimation of the user's typical cluster use. We also asked a few additional questions about access to computers attached to MITnet in the dorms and ILGs. .
Method
All general use clusters were represented in the study, roughly in proportion to the number of workstations in each.
Cluster Interviews Seats % Interviewed % Seats 1-142 20 25 4% 6% Bldg 11 30 29 6% 7% 16-034 25 27 5% 7% Bldg 37 60 59 13% 14% 38-379 17 20 4% 5% 66-080 43 22 9% 5% 2-032 19 16 4% 4% 2-225 25 13 5% 3% 4-035 28 20 6% 5% 4-167 16 9 3% 2% Barker 23 16 5% 4% E51 16 12 3% 3% Hayden 20 8 4% 2% Rotch 14 8 3% 2% W20 124 128 26% 31% Total 480 412 100% 100%
Table 1. Distribution of interviews by cluster
Interviewers were assigned particular clusters, but were left free to choose the times of the interviews according to their convenience. Roughly 2/3 of the interviews were done during the working day (10 AM to 5 PM), and this may have biased some of the results, as we found that the pattern of use differs slightly between daytime and nighttime hours.
Demographics
Of the 480 people interviewed at random, 330 were undergraduates, 125 were graduate students, and 25 were a variety of others, including staff, faculty, researchers, alumni, and visitors. Approximately 7% of all undergraduates (4495 individuals) were interviewed, as were about 2% of all graduate students (5465 individuals). We know from independent analysis of login data that 84% of all undergraduates and 38% of all graduate students with Athena accounts logged in at a public cluster workstation during the period during which our survey was conducted. Distribution by departments is shown In Table 2.
Department Undergrad Grad Student Other Total 1 12 12 0 24 2 27 7 2 36 3 6 3 1 10 4 1 4 0 5 5 6 6 0 12 6 78 25 7 110 7 27 3 1 31 8 12 4 1 17 9 5 0 1 6 10 37 11 0 48 11 4 3 1 8 12 3 1 0 4 13 0 2 0 2 14 6 3 1 10 15 13 2 0 15 16 6 13 2 21 17 2 1 1 4 18 20 2 0 22 21 2 0 2 4 22 0 2 0 2 24 0 2 0 2 HST 0 2 0 2 STS 0 4 1 5 Unknown 63 13 4 80 Total 330 125 25 480
Table 2. Distribution of interviews by status and major
General activities in the clusters
Two questions were asked regarding what type of work is done in the clusters:
1. What activities are you doing during today's session?
Activity Number of people Email 379 Zephyr, News, Talk, etc 201 Document Preparation 179 Web browsing 179 Math/Data Analysis/Plotting 103 Programming 89 Library reference 43 Game playing 27 Other 38
Table 3. Activities reported during the current session (more than one answer allowed)
2. How do you allocate 100% of your time in a typical session? - average, all respondents
Because we realize that most Athena users do multi-tasking, they were told that the numbers need not add up exactly to 100%. Results in Table 4 and 5 are normalized so that the sum of a person's time is constrained to be 100%. Table 4 is the average for all users queried who answered this question. In Table 5, we selected only users whose estimated sum fell between 75% and 125%, giving a potentially more reliable sample.
Activity % of time Email 29% Document Preparation 19% Web browsing 13% Math/Data Analysis/Plotting 11% Zephyr, News, Talk, etc 11% Programming 10% Library reference 3% Game playing 2% Other 2%
Table 4. - How time is spent during a typical session - All users
Activity % of time Email 33% Document Preparation 20% Web browsing 13% Math/Data Analysis/Plotting 12% Zephyr, News, Talk, etc 9% Programming 9% Library reference 2% Game playing 1% Other 1%
Table 5. - How time is spent during a typical session - Only users whose total time estimates are between 75% and 125%
Users were also asked:
What is the primary purpose of your current login session.
Status Coursework Thesis UROP Other Total Undergrad 145 11 26 145 330 Graduate 39 44 0 42 125 Other 5 2 0 18 25 Total 189 57 26 208 480 Percent 40% 12% 5% 43%
Table 6. Primary purpose of the current session
Two thirds of the interviews were done during working hours (i.e. 10 AM to 5 PM) and one third during evening or night hours (i.e. 5 PM to 3 AM). Breaking down the results into daytime vs. evening hours, we find that users do more coursework and less "other" after 5 PM.
Time of day Coursework Thesis UROP Other Total 10 AM to 5 PM 113 (35%) 37 (12%) 20 (6%) 150 (47%) 320 5 PM to 3 AM 76 (48%) 20 (12%) 6 (4%) 58 (36%) 160
Table 7. Primary purpose of the current session, divided into daytime and evening interviews
Time logged in
Users were asked:
Think of a typical week during the semester. For approximately how many hours were you logged in in an Athena cluster?
Status Avg number of hrs/week Undergraduate 12.7 Graduate 14.1 Other 8.3 All 12.9
Table 8. Login time per week
The distribution of the number of hours per week is shown in Figure 1 above.
In an effort to determine how many logins were brief logins to check email between classes, we asked:
What percent of your logins are short sessions (10 minutes or under) that just involve reading and sending of email of using Zephyr?
Figure 2 above shows the distribution of replies. One fifth of the users surveyed are represented in the rightmost two bars, with more than 50% of their logins being short ones. These individuals might well benefit from an express service meant solely for those doing a quick check for email.
Use of Third Party Software
We asked users two questions about software use - one dealing with the software that they are using in the current session (Table 9), and a second asking them to name up to five software packages that they use most frequently, for each saying whether they use it occasionally, on most logins, or always (Table 10). We did not tabulate data on Web browsers, emacs, email, or things such as Zephyr or XV, concentrating mainly on our major third party packages, EZ, Latex, and courseware.
What software are you using during your current login session?
Software Application Users Matlab 82 FrameMaker 73 EZ 42 C or C++ 36 Maple 35 Latex 29 Xess 17 Library tools (Barton, OED, Medline, 16 etc) Miscellaneous Courseware 5 ProEngineer 4 Aspen 4 AutoCad 2 Fortran 2 Stata 2 Clu 2 Lightscape/Renderman/Radiance 2 Mathematica 1 Perl 1
Table 9. Software being used during the current session - More than one answer allowed
Name up to 5 software applications that you use most frequently, and for each, how often to you use it (weighted 1 = occasionally, 2 = at most sessions, 3 = always).
The column labeled Count in Table 10 signifies how many users mentioned this software, and the column labeled Weighted weights the count by 1, 2, or 3, depending upon whether the user uses it sometimes, frequently, or always. The column labled Percent gives the percent of users interviewed who mentioned this software package.
It should be noted that applications such as SAS or Stata tend to be heavily used in the Economics and Political Science clusters, and Aspen in the Chemical Engineering cluster. We surveyed only the general use clusters, and thus the numbers below do not necessarily reflect overall use of some department-specific software packages on Athena.
Software Application Count Weighted Percent FrameMaker 176 286 37% Matlab 161 265 34% EZ 102 149 21% Maple 75 110 16% Latex 66 110 14% Xess 59 84 12% C or C++ 50 80 10% Library tools 23 36 5% Miscellaneous Courseware 12 15 3% AutoCad 10 13 2% Fortran 8 16 2% ProEngineer 7 9 1% Stata 6 10 1% SAS 6 7 1% Aspen 6 7 1% Scheme 6 8 1% Clu 5 9 1% Mathematica 4 6 1% Lightscape/Renderman/Radiance 4 7 1% Perl 4 6 1% Other (<1 % each) 8 11 2%
Table 10. Frequently used software, both unweighted and weighted by frequency of use
Software use broken down by department
For the software packages listed in Table 10, we tabulate in Table 11 use by department or major (Course number). It is notable that 85% of students surveyed use at least one of the `academic' packages, 51% use at least two such packages, and 23% use at least three. The three departments that are the heaviest users of multiple packages are Mechanical Engineering, EECS, and Chemical Engineering, reflecting both their heavy enrollments and also their heavy dependence on Athena third party software.
Course Number Use at Use at Use at
surveyed least 1 pkg least 2 least 3 pkgs
pkgs
1 24 21 10 3
2 36 36 25 12
3 10 10 5 1
4 5 5 3 2
5 12 8 3 1
6 110 94 64 25
7 31 27 13 5
8 17 16 12 4
9 6 4 4 1
10 48 46 32 24
11 8 5 2 0
12 4 3 1 1
13 2 2 2 2
14 10 9 5 2
15 15 13 7 3
16 21 18 14 3
17 4 3 1 0
18 22 20 11 5
21 4 2 0 0
22 2 2 1 1
24 2 1 0 0
HST 2 2 1 0
STS 5 3 2 0
Unknown 80 56 27 15
Total 480 406 245 110
Percent 100% 85% 51% 23%
Table 11. Third party software use by department
Access to a computer on MITnet
Students were asked:
Do you have a computer directly on MITnet where you live?
Followup questions to those who answered in the affirmative asked about the type of platform, and whether the computer was owned by the person being interviewed, a roommate, or was available elsewhere in the dormitory or ILG. Note that we asked specifically about computers attached to MITnet. Also, since our interviews were done in the Athena clusters, results may not be representative of the entire student body, since students with their own computers on MITnet may not use the clusters as much as students without access to things such as electronic mail and Web browsers from their own dorm rooms or living groups.
Further analysis shows that there is no difference in computer ownership between people interviewed during class hours (i.e. between 10 AM and 5 PM) and those interviewed during evening and night time hours.
Status Own roommate's elsewhere in none Total
computer computer the dorm/ILG available or
no answer
`96 17 (19%) 8 (9%) 12 (13%) 52 (58%) 89
`97 28 (30%) 8 (9%) 16 (17%) 41 (44%) 93
`98 17 (26%) 5 (8%) 14 (21%) 30 (45%) 66
`99 18 (22%) 16 (20%) 15 (18%) 33 (40%) 82
Undergrad 80 (24%) 37 (11%) 57 (17%) 156 (47%) 330
Grad 20 (16%) 6 (5%) 4 (3%) 95 (76%) 125
Total 100 (22%) 43 (10%) 61 (13%) 252 (55%) 455
Table 12. Access to computers on MITnet in dormitories and ILG's
The number of students having access to a Macintosh in their dorm or ILG is approximately one half the number of those using some form of Microsoft operating system. If we assume that most of the Unix machines are running Linux (rather than being Athena workstations), then the number of Intel machines is close to 157, as compared to 44 Macintosh computers. This ratio of close to 4:1 holds whether we look only at those students who state that they own their computer, or at those having access to another computer where they live.
Status Macintosh Microsoft Unix (mostly Ambiguous
(Windows, Linux) answer
DOS, etc)
`96 9 17 9 2
`97 7 23 22 0
`98 8 14 13 1
`99 11 26 12 0
Grad 9 12 9 0
Total 44 92 65 3
Table 13. Type of operating system run on computers on MITnet in dorms and ILG's
General Observations
The following things were observed and reported back by interviewers:
- Certain clusters attract certain people due to their location and the time of day. Examples include Course 6 majors checking email between classes in 38-370. Many users of the machines in the Barker Library reported taking a break during a study session t o check mail. Users of the Hayden cluster reported that they go there because of the quiet.
- Many people use a near-by cluster as a place to kill some time between classes.
- What students are doing in the clusters depends on the time of day. "Real work" doesn't get started before 7 PM.
- Results may be biased, because students engrossed in a homework assignment may decline to be interviewed, or may put an interviewer off from approaching them.
- For many students, Athena is their only access to a computer.
- Zephyr is used by very many students, whether in its simply zwrite form or via xzewd.
- Almost everyone seemed to have a Web browser up on their screen.
- Although many people feel that express Kiosks for checking email would be a good things, others want to spend at least an hour on Athena once they log in.
- In a number of cases, a student was working on a non-computer assignment, but was logged in at a workstation to carry on a Zephyr conversation or take a Web- or game-break if he got tired of studying. One student mentioned that he has often been disturbed to see people log in, check email, and then remain at the workstation doing no further computer work, simply working on other homework.
Summary
Analysis of the data confirms many of the general observations that one gets by simply walking through an Athena cluster. The general use clusters are used for a wide range of activities, from quick checks of email between classes to long sessions doing solid course assignments. Use of the workstations for coursework increases from 35% before 5 PM to close to 50% during evening and nighttime hours.
Zephyr and Web browsing are two of the activities that most users mention doing, in addition to electronic mail. Among third party software packages, Framemaker and Matlab are the most heavily used, with EZ, Maple, Latex, Xess, and C following in popularity. 85% of those surveyed report using at least one third party software package, 51% report using two packages, and 23% use at least three.
During our interviews we encountered more undergraduates than graduate students in the clusters by a ratio of more than 2 to 1, but graduate students report a slightly higher number of hours per week in the clusters. Overall, the mean is 12.9 hours per week, while the median is 10 hours. One fifth of the users report that at least half their logins are short ones, just to check email. These individuals might benefit from an express service tailored to this purpose.
One quarter of the undergraduates surveyed who were using an Athena cluster own a computer which is attached to MITnet, while 35% report that a computer on MITnet belongs either themselves or to a roommate. This data does not necessarily imply any conclusions about computer ownership on the part of those students who do not use the Athena clusters, and additional data needs to be collected to answer this question. In addition, a student who owns a computer that he or she chose not to connect to MITnet would show up in our data as having answered No to this question about computer ownerhip.
It is clear that while a significant amount of use of Athena clusters is for educational purposes (coursework, theses, UROP projects), the clusters are also used for purposes such as email, Zephyr, general Web browsing, and as just a place to hang out between classes. What is not obvious is whether or how we should change the environment. Possibilities include:
- certain clusters earmarked for quick turnover and limited to email and Zephyr
- a few machines in each cluster for the above purpose
- express kiosks outside the clusters for quick email checks
- fewer general use clusters, and more specialized (e.g. high end visualization) clusters
- docking stations for students with PowerBooks
- leave things as they are for the time being
It should be noted that the reason for putting UNIX workstations in the clusters is no longer because they are necessarily more powerful than Macintosh or Intel computers of comparable cost. Rather, it is because we know how to run a large scale distributed enivonment using UNIX workstations that is easily maintainable by a relatively small staff. We do not yet know how to do this for machines running non-UNIX operating systems.