April 19, 2000

End of Year Report

UPSCALE: Undergraduate Physics Student's
Computing and Learning Environment

Department of Physics
University of Toronto

Prepared by David Harrison

As is customary, this report summarises the undergraduate computing facility in the past academic year.

WHAT'S NEW

On December 13 our nearly-eight year old Hewlett-Packard UNIX server was replaced by a new Dell 6300/400 Intel-based Linux server. At the same time, our ancient X-based curve fitter was replaced by a totally new linear fitter whose user interface runs inside any web-browser; the fitter is based on Mathematica and its locally written Experimental Data Analyst package plus Java and JavaScript.

Although the new systems were tested as carefully as possible, going into full production pinpointed problems with both the operating system and the new fitter due to the heavy demands put on them during the afternoons. Reasonable stability has now been achieved in both areas, with further improvements anticipated.

With the new stability, we gained the confidence to replace an ancient X-based data grapher with a new one based on the same technology as the linear fitter; this change occurred on March 19. By September we will use the same technology to offer new routines to:

• automatically fit reasonably simple spectra to Gaussians, Lorentzians, or Breit-Wigner shapes, including linear backgrounds.
• fit data to the form y = A Exp[B*x] without the biases introduced by common linearisation techniques.
• do robust line and curve fits.
• do general nonlinear fitting.

This we are now much closer to our goal of moving all of our services into web format, accessible with any browser.

USAGE

UPSCALE services can be divided into three main categories:

1. Interactive student services via traditional logins. This usage is summarised in the sub-section 1 below.
2. Delivery of information and analysis via the web. This usage is summarised in the sub-section 2.
3. File and print services for the Nortel and other PC's. This usage is summarised in sub-section 3.

Below I shall display the usual statistics and graphs. However, further insight into UPSCALE's operation may be gained from the fact that we sampled the number of computing jobs in the run queue (the "load average") every ten minutes between 3 PM and 4 PM on week days during the month of March. During this period we averaged 3.15 jobs in the queue, with a high of 12.36 and a low of 0.11.

Another number that is illustrative is that during the month of March our new linear fitter delivered 12,674 fits to our students. 13% of these were delivered to seats not in our building. We believe that this latter number is fairly small because prior to the introduction of the new fitter over Christmas, students were largely forced to do their laboratory data analysis in-house so became used to working in this way. We expect usage of the fitter by students not sitting at one of our seats to be much higher next year.

1 - Traditional Logins

The following figure and table shows the usage figures via traditional logins.

UPSCALE maintains two types of accounts for undergraduate students, offering somewhat different services. The "x" accounts are given to our first and second year students. The "special" accounts are for upper-year students. The above plot shows the logins per week for the "x" accounts (displayed as solid bars) and "special" accounts (displayed as white bars). The effects of the Christmas break, the TA strike, and the ReadingWeek break are clearly visible.

Some of the more interesting conclusions from the above numbers are:

• The number of accounts, which is essentially all students, is down somewhat from last year.
• The effects of the TA strike, particularly on our first year students whose laboratory was closed for four weeks, is less than might have been predicted.
• In previous years the number of students logging in remotely via telnet, rlogin, ssh, etc. were trivially small. The numbers have quantum jumped this year.

2 - Web Usage

In previous years one of the ways we summarised the delivery of information via the web was the total number of files delivered. I have realised that this is mis-leading, since it includes all image files that are included in web pages, no matter how small. A more meaningful number is the number of html, htm, and pdf files delivered. Below we present that number for this year and last year; in order compare this year to previous years we also include the total number of files delivered this year. Neither number includes the fits or graphs performed with our new web-based programs.

Also, in the following table, the phrase "top-level accesses" means access to the html "home page" for a particular sub-system. Certainly some students have bookmarked pages accessed by these "home pages" and this access is not "top-level".

This year all Guide Sheets in the I Year Laboratory for "non-core" experiments are distributed exclusively as pdf documents, which the students may then print. We also distribute a sample of the I Year Laboratory written test as a pdf document. We summarise these usages also.

Finally, the Physics Virtual Bookshelf is a collection of locally written documents that have been collected into a single place where they are accessible to a wider range of people than just the students in the courses/labs for which they were originally written. We regularly receive positive comments from around the world about this facility. Further, the bookshelf and its contents receive high rankings from http://www.google.com.

Once again, despite the TA strike, for the third year in a row usage of our web site nearly doubled.

The above plot shows the number of bytes per day delivered by our web site from Sept 1, 1999 through April 19, 2000. The ticks on the horizontal axis are the dates of the four tests plus the Final Exam for our largest course, PHY138Y. These students were accessing our bank of scans of problem set, test and examination solutions, and our on-line versions of overhead transparencies used in lectures.

3 - File and Print Services

The Nortel Applied Physics Laboratory is an integral part of our upper-year laboratories. It features 16 Windows-based PC's, for which our central server provides file and print services. We provide similar services to other PC's in the undergraduate wing.

For Mathematica applications, we have found that running the kernel on the server with the "front end" running on a PC in the Nortel laboratory is the preferred environment.

Previously, it was very difficult to track exactly how these services were being used. With our new machine, the latest version of the server software (Samba) allows us to monitor usage. Thus below we present some figures for the second term of this academic year.

If the proposed new PHY140Y laboratory becomes a reality, we intend to extend these services to that laboratory.