The History of Computing at Chemistry

Tom Atkinson

---

[This article first appeared in the Fall 1995 issue of the Department's Alumni Magazine.]

---

Introduction

Atkinson's 1st Law of Computing: All computers and all computing are simply a combination of a very, very large number of very, very trivial things.

The involvement of the Chemistry Department in computing over the last 4 decades parallels the evolution of the vibrant, volatile, exciting, explosive computing and communication industries, as the department made use of successive waves of new technology. Figure 1 illustrates this evolution with a few examples of the technology over the years. Table 1 shows how the price - performance has evolved. This evolution of computing can be viewed as a multi-dimension space (enabling technology, hardware architecture, operating system (OS) architecture, application software, user interface, memory technology, storage technology, communication technology, locus of user access points...) with concurrent and correlated development proceeding along the many different dimensions.

The underlying technologies of computing and communication evolved during this period of time from vacuum tubes, to transistors, to every higher density integrated circuits (SSI, MSI, LSI, VLSI, ...). The distribution of computing technology went from individual (paper, pencil, slide rule, mechanical calculator), to centralized, to departmental, to group, to individual as the price performance ratio continually fell over this period of time. Before the 1950's computation was typically manual for most scientists and was limited to relatively simple theoretical calculations and data analysis. The advent of computing facilities allowed the expansion of the computation tasks in variety and complexity. Thus, computational utilization in the department has expanded to also include data acquisition, experimental control, simulation, manuscript preparation, visualization, communication, presentation, peer collaboration, grant management, instruction, instruction management, and on-line information delivery.

Our involvement in computing beyond the use of the existing hardware/software facilities has also changed over the years. When I arrived in September 1972, many members of the department were actively programming and building interfaces (from components). The experimentalists were programming in assembly language and those doing theoretical calculations and data analysis at the central site were using Fortran. As the seventies progressed, these activities expanded and with the advent of the microprocessor, people were even building computers from components. At this time, computer science was a young discipline and the physical sciences and math often provided software and hardware people to the computer industry. By the mid-1980's, interfacing and other hardware activities in the department were beginning to wane. Most of what you needed was now commercially available and it is very difficult to compete with a volume manufacturer on cost, documentation, and support. By the early 1990's even the programming activities had diminished. Again, many codes are now available as commercial or freeware products, especially utility and productivity software, e.g. manuscript production, plotting.

Members of the Department have played a role in shaping the course of computing at MSU. Chemistry, along with Engineering, Physics/Astronomy, and the National Superconducting Cyclotron Laboratory have traditionally been heavy users of computing technology, both within the departments and at the central site. These units along with the MSU Computer Laboratory (MSUCL) and the Administrative Information Services (AIS) have been the focus of much of the computer developments on campus. Many members of the department have served in various capacities in the advising structures guiding the university in the acquisition, implementation, and operation of computing and communication technology. Harry Eick and Jim Harrison served in the earlier years. Dick Schwendeman served on various committees within the Computer and Communication Advisory Committee (CCSAC) an advisory committee to the central administration. He served as chairperson of CCSAC. I have served on the Network and Communications Committee (NCC), a subcommittee of CCSAC since 1981 and have been the chairperson since 1984. I was chairperson of CCSAC in 1994-95.

Harry Eick served as Acting Director of the MSU Computing Laboratory for several years in the 1970's. He then served as the Assistant Director of the Computer Laboratory for MERIT activities. This was in the early years of MERIT. Paul Hunt was appointed as Director of Academic Computing in 1985. The MSU Computer Laboratory reported to Paul and he administered central funds for computing at the central site and provided grants to the units. In 1989, Paul's position was elevated to Assistant Provost. In 1991, Paul's position was further elevated to Vice Provost for Computing and Technology which now encompass AIS, Broadcast Services, and Instructional Media Center (IMC) as well. Paul is currently the Chairperson of the Board of Merit Inc., the operator of Michnet, the state educational network, and until recently of NSFnet the national Internet backbone. He has also served as Chairperson of the Board for CICnet, a regional network.

Figure 1 - Evolution of Computational Tools: Paper and Pencil, CRC "Rubber" Handbook, Slide Rule, HP-35, SGI Onyx computer and Gateway Notebook computer.

Table 1 provides some insight into the nature of this industry over the last 2 decades. This table contains price and performance data on a number of the systems used in the Chemistry Department. Also included are a number of machines operated by the central site (MSUCL). The CRAY 1S, the first super computer and the CRAY T916, current top of the LINPACK list, are included for comparison. The table contains the node name or the faculty owner of the first example in the department of that particular hardware, the year of purchase, the cost of purchase and the cost adjusted to 1995 dollars for systems purchased by the department, the system type, and the LINPACK, measures of performance: Mflops (N=100) and Mflops (N=1000). The right column is the ratio of Mflops( N=100) to the cost in thousands of 1995 dollars. The Mflops columns contain the number of millions of floating point instructions the system is capable of performing in one second. For the right three columns the larger the number the better.

Table 1 - Price Performance History of Chemistry Department Computers

The measure of performance used is Jack Dongarra's LINPACK benchmark. As with any benchmark, the LINPACK benchmark is a specific and constant program and an associated fixed set of data that can be executed on different machines. The time required for the benchmark to run on a given system can then be used to compare the expected performance of the system with others. This particular program/data set solves a problem in linear (matrix) algebra and is computationally intensive, i.e. measures CPU and memory performance, not performance of peripheral equipment such as disk drives. No one benchmark program can completely evaluate the performance of a given computer system. And certainly, benchmarking and other performance measures can not fully assess the value of a given system to its owners. Still the LINPACK numbers have been collected on a large set of systems over the last 10 to 15 years, more than any other single benchmark, and do provide some indication of the relative merits of the systems.

The Early Years

Figure 2 - The K&E Teaching Slide Rule: TVA, Patty McCarthy, and Tibor Naby.

The earliest computational aides utilized by departmental members were the paper, pencil, slide rules, log and other mathematical tables (See Figure 1). Figure 2 shows the large K & E slide rule used in instruction.

Figure 3 - MSU Computational Facility, ca 1955.

In the 1940's and 50's, mechanical and electro-mechanical calculators were being used. Figure 3 illustrates a Frieden facility located in Kedzie circa 1955.

Figure 4 - The Wang Calculator.

In 1969, electronic calculators had arrived in the department. Figure 4 shows a picture of the Wang "shared logic" calculators being used in Chemistry. A central electronics package (5" x 8" x 24") was connected to up to 8 user stations by a multiple conductor cable. These stations were distributed through out the building. The Wangs were still in use in 1972. The arrival of the HP-45 electronic calculator in the early 1970's brought significant and affordable computational power to the individual.

Around 1970, the department launched into our first experiment in local general computing with the installation of an IBM 1130 in Room 337. This entailed the conversion of Room 337 from the original stockroom into a computer facility complete with raised floors and the glass viewing wall. The "Glass Room" housed the IBM 1130. One purpose for the facility was to serve as a remote job entry terminal for the MSU Computer Lab's CDC 6500. For this to occur, IBM had to write the "terminal emulation" software that allowed the IBM 1130 to communicate with the remote CDC 6500. The communication link was a phone line. The IBM 1130 was also used for stand alone computation by Jim Harrison and Vince Nicely.

This facility was viable for only a few years, its fate sealed when the MSU administration adopted the practice of subsidizing computing at the MSU CL for campus users. At that time the IBM 1130 was replaced with a CDC EI 200 remote job entry station that was connected via a MSU owned and installed dedicated cable to the computationally superior CDC 6500.

The Central Site

MSU entered the computer age during the mid 1950's when the MSU Computer Laboratory constructed the Michigan State Integral Computer (MISTIC),,. This machine had 1024 40 bit storage locations and was constructed with vacuum tubes and used Hollerith cards and punched tape for input. This machine went into production in 1957. During 1962 - 63, this was augmented with a Control Data Corporation 160-A. This was about the time that Stanley Kubrick brought us Dave and HAL having their conversation in "2001." In turn these machines were replaced with CDC 3600 (1963), CDC 6500 (1968), CDC 750 (1979). In 1987, the CDC 750 was replaced with an IBM 3090-180 VF and a DEC VAX 8650. In 1989 a Convex C-220 and a BBN GP-1000 (Butterfly) 96 node parallel processor were added. In 1993, the IBM 3090 was upgraded to a 3090-200J.

The Mini Computers Arrive

In 1967, Al Tulinsky brought the first minicomputer into the department, a Digital Equipment Corporation PDP 8, a 12 bit computer with a 4Kx12 bit core memory and a 32Kx12 bit disk drive that was interfaced to a Picker x-ray diffractometer. Dick Vandlen developed much of the software for this system. This facility automated the collection of crystallographic data. A second Picker with an PDP8I was added a few years later and became the departmental small molecule crystallographic facility. Chris Enke obtained a DEC PDP 8I that was housed in Room 408 for the development of computer interfacing to experiments. This computer also had a number of analog lines running to Jim Dye's lab in the basement allowing the sharing of the facility for acquiring voltage signals from the Dye group instrumentation.

By the time I arrived in September 1972, Stan Crouch had obtained a DEC PDP 8e. Shortly there after, the department implemented a shared DEC PDP 8e in order to expand the availability of computers in the laboratory. Unlike other shared approaches, the "rolling 8e" went to the experiment rather than bringing the experiment to the computer. The "Rolling 8e" was built into a wooden cabinet about 2' deep x 6' high x 8' long that also contained a Teletype, storage drawers, etc. It was rather cumbersome, and usually required a number of people to push it up and down the halls and elevators. Still the machine was in high demand for a number of years. In fact the sign-up procedure was rather elaborate and I often though we should hire a lawyer to arbitrate the sign-up rules. Within a few years however, the Leroi, Schwendeman, Eick, ... groups had obtained their own PDP 8's and the desire for use of the "Rolling 8e." had diminished. Our last PDP 8 acquisition occurred when we updated Al Tulinsky's PDP 8 to a PDP 8a in the late 1970's.

Figure 5 - CEMCOMGRAF (Room 337): Cal Data 135, Lear Siegler ADM-3 terminal, Sykes 8" floppy Disk, Vector General Display.

In 1974, Chris Enke purchased a DEC PDP 11/40, a 16 bit computer with 256 Kbytes of memory and a 1.4 Mbyte removable RK05 hard disk drive. This marked the beginning of the PDP 11 era. A CalData 135 (See Figure 5) was purchased in 1975 and was an exact emulation of the PDP 11/40. This was the first emulation of one manufacturers CPU by another manufacturer. The product was very good and considerably cheaper than the original. The CalData 135 became the heart of the CEMCOMGRAF (ChEMistry COMmunications and GRAphics) facility in room 337. Unfortunately for us, DEC was able to legally restrict the manufacture of the product. During the next 10 years, the department acquired large numbers of LSI 11s, PDP 11/23s, and PDP 11/73s. In this time period we were also using Data General 16 bit Nova minicomputers on some of the Xray diffractometers and Nicolett 1070s and 1080s and Bruker Aspects on the various NMR instruments.

The mini-computers were used not only for data acquisition/experimental control but were also used for local computation in general. Before this time, the more serious data analysis, manuscript preparation, and data presentation had been done only on mainframes at the central facility with the exception of the IBM 1730. As the power of these computers grew more and more complex computation was done.

The Midi's

Atkinson's 8th Law of Computing: There is never enough memory.

Atkinson's 12th Law of Computing: Theoretical Chemical and Physical calculations will consume as many CPU cycles as are available.

Figure 6 - FPS 164 ("Glass Room")

In 1982, Kathy Hunt, Paul Hunt and Al Tulinsky purchased a DEC VAX 11/750 (CEMVAX, the first) for their research efforts and housed it in Room 213. This was the first "midi-computer," the first 32 bit computer, and the first virtual memory machine in the Department. This facility provided a significant jump in computing power and was competitive with central site facilities. Two years later, a FPS-164 attached array processor (See Figure 6) was installed in the "glass room" in Room 337 and the VAX 11/750 moved there to be the "front end." This facility was implemented with departmental, college and university funds and became a university facility.

As with many acquisitions of major instrumentation, once the decision was made to implement this facility, we were in a great hurry to have it up and running. As it turned out, the FPS-164, which consumed 100 amps of 120 VAC and dissipated the corresponding large amount of heat, arrived before the installation of the EDPAC air conditioner. The FPS-164 was installed anyway . So, throughout January, February, and March of 1983, Paul Hunt or I would make several trips a day to the "glass room" and opened and closed the windows and turned several large floor fans on and off to regulate the temperature which was measured with a couple of Hg thermometers from the stock room. For those of you that took CEM 838, this "Michigan Air Conditioning" is an example of a closed loop control system. Paul and I were the feedback loop.

Individual faculty began acquiring this class of computer over the next years as the price performance ratio for computers continued to fall rapidly. A succession of MicroVAX, VAXStation 2000, VAXStation 3200, VAXStation 3100, and VAXStation 4000 (CEMVAX II) computers appeared over the years. The VAX stream was joined by a series of SUN work stations beginning with a SUN 3/160 (horus, the first) followed by a number of Egyptian gods (toth, aten, bastet, ..., Kermit Johnson has always liked Egyptology.) in the forms of SPARC 1, SPARC II, SPARC 10, ...

In 1985 AT&T gave the department several 3B2-300 minicomputers. Over the next two years additional 3B2-400 minicomputers were donated to the department. These UNIX machines were utilized by a number of the research groups and the Business Office. The last one was retired this year.

By 1989, the FPS/VAX facility had been overshadowed by industry developments and a large maintenance budget ($30K/year). At that time, four faculty (Kathy Hunt, Paul Hunt, Jim Harrison, and Al Tulinsky) with significant departmental support purchased an Ardent Titan 3000, eventually with three parallel processors, at that time. In short succession, this was replaced with a Stardent Titan 3000 and then a Kubota Titan 3000. Actually, the box did not change, merely the nameplate, as the manufacturer went through several mergers and acquisitions over the course of a few years.

In 1992, the department acquired three Silicon Graphics R3000 Indigo workstations. These workstations were not only superior machines for graphics, but were computationally competitive with the Titan. As with other technological threads, faculty members began acquiring a series of SGI Indigos, Indigo²s, Indigo Extremes, Challenges, ... In 1994, a Silicon Graphics 4 processor, 256 MByte Onyx became the departmental computational work horse. In addition, faculty have HP 9000 and IBM RS 6000 systems.

By the 1990s, the traditional (Well, they were good terms for 5-10 years at least.) terms micro-computer, mini-computer, midi-computer, and mainframe had lost their meanings as a midi or mainframe launched one year was often eclipsed by a "micro" chip the next year. Indeed, most computer systems now use a "micro computer" i.e. integrated circuit, CPU. A related phenomena: SGI recently announced that the next version of their operating system, to be released in 1996, will no longer support the SGI R3000 Indigos that were new in 1992.

From the mid 1980's to the present, the majority of the computing was done within the department. Computing at the central site being perceived as relatively more expensive.

The Micro Age

The Micro Age began with Intel launching processors built on a single or a few integrated circuits in 1974. The first examples were 4 and 8 bit CPUs. Probably the first microprocessors to appear in Chemistry were the 8 bit Intel 8080 and the Rockwell 6502. The single board AIM-65, based on the 6502, was used in CEM 838. The 6502 was also the engine for the Apple II and the Commodore Pets. Chris Chang and Peter Wagner were advocates for the Apple II. John Allison used a number of Commodore Pets to interface to experiments. Paul Hunter and Tom Clarke were boosters of the Commodore 64 based on the Motorola 6510.

IBM launched the IBM PC and XT in 1981 both based on the Intel 8088. Over the next 15 years, new models arrived based in turn on the Intel 80286 (AT), 80386 (32 bit), 80486, and the P5 (Pentium). The first PC arrived in Chemistry in 1983. The first large scale use of the IBM PC in the department started with 8 secretarial machines purchased in 1984. I have always maintained that the heavy involvement with the mini-computers lead to the department not entering the PC world for several years.

On a Sunday night in January 1984, a George Orwellian advertisement aired during the Super Bowl launched the Apple Macintosh computers based on the 32 bit Motorola 68000. The commercial has intentionally never been aired since except on a very few special programs, but the Macintosh is still with us. Within the next year John Allison brought the first MAC into the Department. Over the 1980's further acceptance of the MAC by chemists in general and our department in particular was spirited by the CHEMDraw program for drawing 2D chemical structures. Of course, the "user friendliness" of the MAC graphical user interface (GUI) was another attractive aspect of the MAC. Today, one third to one half of the department have chosen the MAC for their desktop. Over the years, new models of the MAC arrived as new microprocessor chips were developed by Motorola ( 68010, 6820, 6830, and 68040). In 1993, the Macintosh OS was ported to a new platform, the 32 bit Power PC RISC microprocessor.

Parallel Processing

In the never ending search for more computing power, improvement of the individual computer typically receives the first efforts. Improvements in the hardware architecture, the device technology used to implement the hardware, and the software are always candidates for improvement. For a long time, however, the industry has looked to parallel processing, the simultaneous use of more than one computer, as a way to get more power for a given computational task. The essence of parallel processing is to have a group of computers, a communication framework to allow the individual machines to communicate with each other, and software or other mechanisms to distribute the work among the various individual machines and otherwise synchronize the operation.

Parallel processing, in a simplistic sense, began for Chemistry as soon as there were more than one machine available to an individual. The person would merely submit jobs to more than one computer at a time. The individual is the controlling mechanism and "footnet" provides the communication links.

Another example of parallel processing, in the more general sense, came to the department in the late 1970's when Chris Enke began using multiple microprocessors to comtrol instrumentation. In the mid 1980's when we implemented the first VAX Cluster which uses Ethernet for the communication mechanism. In the same time frame, Kermit Johnson was implementing the first UNIX cluster on the NMR SUNs using the Network File System (NFS), and the BSD "r" commands. In both cases, the cluster appears to the user as "one machine", allowing files to be located on any disk on any machine in the group, use of any printer, jobs to be submitted to any processor in the group, etc. Similar technologies include Appletalk and Work Groups for Windows, both providing file and printer sharing and other distributed facilities.

However, when a computer scientist talks about parallel processing, they almost always mean a set of tightly bound CPUs, with associated memory systems, communication channels, and associated architecture. The Department has been involved in parallel processing in this formal sense since the late 1970's, when Chris Enke evolved the multiple microprocessors control of instrumentation into more tightly bound systems with communication mechanisms centered around a multiprocessor I/O bus (Affectionately called the "Bruce Bus" after Bruce Newcome, one of the principle developers.).

The VAX 750/FPS 164 was an example of asymmetric parallel processing, i.e. the various processors are not equivalent. The VAXstation 3520 (xray02) is the first example of symmetric multiple processing (SMP) in the department. The two processors of the VAXStation 3520 are scheduled for different tasks. The Ardent Titan was the first SMP system where parallel jobs were run with more than one processor allotted to parts of the same job. The SGI Challenge and Onyx also use multiple processors in parallel for a given job.

On the Periphery

Atkinson's 4th Law of Computing: All disks eventually crash.

Atkinson's 5th Law of Computing: From time to time, files are inadvertently deleted from any given volume of media.

Atkinson's 6th Law of Computing: When a person starts out computing, he or she is very careful about copying any files created or modified to separate media and/or periodically exiting from editors so that the results of the editing session will be written to disk. These good practices continue for a while and nothing bad happens. The person becomes a little lax about saving and backing up information. Still nothing bad happens. The person becomes a little more lax... However, after some time has passed, the person gets burned badly, e. g. a disk containing the only copies of important files crashes or the power goes off after 3 hours of editing and nothing has been saved to disk, or a file is inadvertently deleted. After such an unfortunate occurrence the person is once again very careful about backing up their files and periodically saving the results of editing sessions. These good practices continue for a while and nothing bad happens... For me, these spikes come about three years apart.

Atkinson's 7th Law of Computing: All disk drives fill up.

A very important dimension of a computer is how information is input, output, and stored, both during and after processing. The technology and functionality of these peripheral aspects of computer systems have evolved with the CPU hardware and the software.

During the 50's and 60's punched paper tape and punched paper Hollerith cards were the standard form of input and program and data storage. Department members were using Hollerith cards until about 1980 for input to the central site. Paper tape was a very common media on the minicomputers until the arrival of first DECtape and then floppy disks.

The first disk drive in the department was the DF-32 on Al Tulinsky's PDP 8. The IBM 1130 also had disk drives. During the mid-1970's a number of DEC and Diablo drives utilizing the removable IBM disk cartridges were used. Tables 2 and 3 contains a brief synopsis of disk drives used in the department. At this time, the price of disk storage is about $200/GIGAbyte.

Up until the mid 1970's, memory was constructed using collections of small ferrite torodial beads with wires running through the center of the torus to change and sense the magnetic state of the ferrite. Each bit of storage made use of one of these beads. As a result memory systems were small and expensive. A "core" plane with 4K x 12 bits for the PDP 8e was $4000 in the early 1970's. A major break through came with the advent of solid state memory. In 1974, we bought our first solid state memory in the form of a 4K x 12bit memory from Signal Galaxy for the PDP 8e. This module was based on 1K bit semiconductor memory ICs. Since then, the number of bits on one memory chip has gone from 1K bits to 4K to 16K to 256K to 1M to 4M to 16M bits. This trend has allowed the typical memory size of our computers to go from the 4K and 8K words of the PDP 8s to the 256Mbytes of the SGI Onyx.

Another storage technology worth mentioning is magnetic tape. The industry standard 1/2" tape was in evidence in the 1960's. The tapes began with 7 tracks and stored 556 or 800 bytes per inch (BPI). Next came the 9 track 1/2" drives which stored 800 BPI at the beginning of the technology and went on to be capable of storing 1600 and 6250 BPI. In the 1970's the small reels of DECtape were popular. In the 80's and 90's, DEC's TK-50s, quarter inch tape cartridges (QIC) and DAT tapes were used.

Table 2 - Floppy Disk Drives

Table 3 - Hard Disk Drives

Legend for Table 2 and 3

The User Interface

Atkinson's 13th Law - You aren't a real computer person until you have toggled in 100 bootstraps.

In the earlier days of computing the user interface to the computer consisted of the front panel, a card or paper tape punch, a card or paper tape reader, and a printer. The front panel was a collection of switches and lights that allowed the operator to start and stop execution, deposit information into memory, examine the contents of the registers, and generally monitor the operation of the computer. Before the self operated mini-computers of the early seventies, front panels were important only to the operators and system people of the central site mainframes. The front panel provided the means to deposit, "toggle in," the initial program, the "bootstrap", required to load and execute the user's program or even the early operating systems. The early PDP8's, PDP11's, and the Nova's all required one to toggle in the bootstrap. The front panel switches and indicator lights were also used for communication with the user's programs.

By the mid-seventies, the development of Read Only Memory (ROM) allowed bootstraps to be permanently stored within the computer. As the size of ROM available grew, the complexity of the programs permanently stored also grew. Eventually, a program sophisticated enough to have a dialog with the operator at a connected terminal was possible. This allowed the replacement of the expensive and bulky front panels with a "front panel" emulator, a program stored in ROM. This continues today, with computers such as the personal computers having the "front panel" and OS loading capabilities included in the ROM along with other functionality, e.g. hardware configuration settings, and components of the operating system (the bios).

The early mini-computers often made use of the Teletype model ASR 33 or 35 as the user interface. These devices provided a keyboard and a paper tape reader for input, a printer and a paper tape punch for output. These devices were mechanical containing hundreds of levers, springs, wheels, cams, and solenoids. They were noisy, messy, i. e. oil and punched "bits" of paper everywhere, uppercase only, and connected to the computer or modem via a 110 baud serial link. We loved them.

By 1973, the department was using the Tectronix 1410 storage CRT terminal which provided graphics as well as text. By 1975, the first low cost CRT terminals, ADDS, Lear Siegler ADM-3 (See Figure 5) and Heath H19, an emulation of the DEC VT-52, arrived in the department. These electronic terminals allowed faster communication, 300, 1200, 2400, and even 9600 baud, with the computer and provided both upper and lower case.

In 1975, the department purchased a Vector General, a very high end ($53K) 21" monochrome graphics display (See Figure 5). The VG was capable of real time rotation and translation of the three dimensional object being displayed. This was achieved with a hybrid collection of analog and digital techniques. The VG was driven by the CalData 135. Unfortunately, the display came with no software. Steve Ernst and I spent a year or more writing FORTRAN, assembly, and display code for the facility. The VG was displaced by the arrival in 1981 of an Evans and Sutherland PS 300 monochrome display in the Tulinsky lab. This was upgraded to a color display within a year. Later the PS 300 was upgraded to a PS 350 and then to a PS 390. The PS 3xx displays coupled with the program named "frodo" allowed the protein crystollagrophers to view, rotate, and translate multiple 3D objects.

By the early 1980's we were beginning to get "workstations," i.e. computers with more tightly integrated displays allowing faster output of information to the user. On the low end were the Macintoshes and the IBM PCs. On the high end were the VMS and UNIX workstations. Graphics capability was now the norm. Color was also becoming common. With the workstations came windows and the graphical user interfaces (GUI). Mice also began to appear.

The OS Wars

An important component of any computer system is the operating system (OS). The operating system is a program that runs on the computer and does the many common things necessary for the user that the person would have to include in their programs otherwise. For instance, the OS loads programs into memory so they can be executed, schedules when a particular program will actually execute, keeps one user or program from treading on another, provides for directories and files on the disks. All user programs and other applications execute under the control of the operating system.

The development of the OS has been a crucial component of computer evolution. Over the time span of interest here, the OS has gone from nonexistent to very primitive to ever more robust, providing more and more functionality (and consuming more and more memory and disk space). Early examples were PTOS and OS8 for the PDP 8s. DOS/BATCH, RT-11 and RSX-11M for the PDP11s.

In the mid-seventies, as we were migrating to RSX-11M from DOS/Batch, we had the opportunity of going to UNIX, the new operating system that had been developed by Bell Labs. UNIX was novel in many ways but the most important was that it was written in a high level language, "c", and was thus portable, i.e. with a relatively minor amount of work, it could be moved to a new platform. Bell Labs had developed it on various DEC PDP architectures and it was becoming popular for the PDP 11 family. I decided to stay with the DEC operating systems because of two facts. First, at that time the license holders couldn't even talk about the UNIX except with other license holders. When I talked to people on campus that were running UNIX, they said "We are not allowed tell you about it but you really ought to run it." Second, there was no charge to use the system, but you had to contact the lawyers at Western Electric and negotiate a contract for using the system.

The early 1980's saw the arrival of the DEC VAX and DEC's operating system VMS. During the same time, Bell Labs had modified UNIX to utilize virtual memory and many installations, including Bell Labs, ran UNIX on the VAX. As the 80s progressed, groups strongly advocating each of the two operating systems developed. Still the popularity of UNIX was touchy. I have an article from the early 1980s predicting the disappearance of UNIX.

A very important development, the development of Reduced Instruction Set Computer (RISC) architectures, turned the future of UNIX around. Prior to this time all processors built were Complex Instruction Set Computers (CISC) partly in order to minimize the amount of memory required for a usable system. With the advent of solid state memory, the cost of memory began falling, allowing the reexamination of the principles of computer architecture, especially as applied to integrated circuit processors. Studies showed that much simpler instructions sets would suffice. The general idea was that the simpler instructions sets would be easier to implement allowing the succeeding generations of processors to be brought to market more rapidly. This philosophy was embraced by the computer industry, bringing the SUN SPARC as the first example. A number of companies designed and built a number of different RISC processors. However, these new architectures needed software. A modern operating system requires many man years of development. The portability of UNIX allows the manufacturer of a new architecture to invest one or two man years to porting the system to the new architecture and the product is ready for market.

Another fact that lead to the saving of UNIX was that a whole generation of computer scientists grew up with UNIX at the universities during the 1970's and 80's. In addition to being essentially free to the universities, UNIX came with the sources, unlike most other operating systems, a natural for the study and development of operating system software. Still, if VMS had been made portable earlier, the scientific community would have very likely stayed with VMS. As it turned out the price performance of the RISC machines running UNIX was too good for the scientific community to ignore. The computer scientists would have gone with UNIX in any case.

While the VMS, UNIX battle was going on at the high performance end of computing, the Macintosh and the IBM PC were battling for the desktop at the lower end. Early on this was a contest between the command line driven MS-DOS and the graphical user interfaced MACs. With the advent of Microsoft Windows, followed by Windows NT and Windows 95, the operating environments of the two lines have become very close. In addition, the power of the processors utilized in the MACintosh and IBM PC lines is now so great, that the distinction between the high and low end workstations is no longer drawn. One chooses among UNIX, Macintosh, Microsoft Windows, and VMS for the desktop. In fact, VMS, Windows NT, and MAC OS are all portable now and are running on a number of platforms. By the way, the principle architect of Windows NT was Dave Cutler, who happened to be the principle architect of VMS.

The Local Network

Atkinson's 11th Law - Sooner or later, you will need to move data from any one computer to another.

Atkinson's 9th law - Never move binary files among heterogeneous systems.

The evolution of communication technology has paralleled and contributed to the evolution of computing. Internally, many communication techniques, e.g. parity in memory, error correcting memory, bus technology, are used within the hardware architecture of computing systems. In addition, the user interface is often physically separated from the computer and requires a communication link between the two. And today, of course, the "information highway" is a penultimate communication application with millions of computers connected via a multitude of communication links of various types.

In the early days of computing, information was input to the computer via Hollerith punched cards or punched tape, both paper. Information was output via line printer listings. The individual prepared his or her sets of input information and "carried" them to the central site where they were handed to an operator for reading. The individual would return in hours, days, or weeks to pick up the output produced from that run and retrieve the input. When Al Tulinsky was a postdoc at Brooklyn Polytechnic, this communication link consisted of a bus rides across Manhattan to IBM World headquarters and the Watson Lab at Columbia University to submit his jobs to IBM 701, 704, and 7090 systems. Jim Harrison, Dick Schwendeman, Harry Eick and many others remember walking over to the MSU Computer Laboratory to submit decks of Hollerith cards.

Early in the evolution of computing, magnetic media: tapes, disks, and even cards, were developed and allowed larger amounts of information to be carried from place to place. As networking surfaced in the early 1980's, the use of these portable paper and magnetic media to carry information from one place to another became known as "footnet" and "sneakernet." A variant of this is when you take the computer to the experiment as in the "Rolling 8e" and the current notebook computers.

By the early 1970's there was a dedicated set of cables running from the Chemistry building to the MSU Computer Lab providing the communication link between a remote job entry station in Room 337 and the CDC computer in the MSU Computer Laboratory. In the same time period, 110 baud modems appeared on the scene, allowing a remote connection between a teletype and the MSUCL.

In 1979, MSU undertook the deployment of three parallel broad band cable (BBC) TV systems on campus. One of these cable systems delivered instructional and other TV signals. The second system was dedicated to data applications. The third is a spare. Almost immediately, the data cable provided point to point 9600 baud links between remote energy utilization sensing and control stations in buildings around campus. Physical Plant uses these facilities to remotely control various Heating Ventilation and Air Conditioning (HVAC) systems around campus in order to conserve consumption of energy. Other point to point connections connected various administrative offices with the AIS mainframe. By 1990, more than 110 buildings on the main part of the MSU campus had been connected to the BBC system.

In 1983, MSU launched the first "network" on the Data BBC using Continental Telephone's Contelnet technology. This consisted of a large number of Buffered Interface Units (BIU), i.e. terminal servers. The BIU provided eight asynchronous serial connections for links to terminals, printers, and computers. In addition, the BIUs were linked together with a 2 Mbaud communication channel occupying one of the TV channels on the DATA BBC. The BIUs provided a virtual 9600 Baud connection between any two ports on any two BIUs on the network.

In 1985, an Ethernet on Broadband (EOB) network was put in place on the Data BBC, along side of the point to point and Contelnet networks. Initially this network connected MSUCL, Engineering, Physics, NSCL, and Chemistry at 10 Mbps. This connected the Ethernets that had been appearing within the buildings over the previous year or two. Additional buildings were connected to the EOB at the rate of about 10 per year. At this time, over 80 buildings are connected to the campus network.

In 1991, an optical fiber communication facility were installed connecting 10 campus buildings. The first application of this cable plant was to implement a 100 Mbps Fiber Distributed Data Interconnect (FDDI) backbone network connecting the Administration Building (AIS), Chemistry, Communication Arts, Engineering, MSUCL, NSCL, and Physics/Astronomy. At this time, Phase II of the optical fiber project is extending the fiber to 6 more buildings. Phase III of the project will be bringing the total number of buildings on the fiber to 47 by early 1997. At that time, 27 buildings will be connected to the FDDI backbone network.

The Wide Area Network

On the wide area scene, the mid 1970's saw the beginnings of the MERIT network which originally connected the computing centers of MSU, UM, and Wayne State together. Harry Eick was involved in development of MERIT during this stage. MERIT allowed an individual to run jobs on the remote systems and to transfer files among the machines. One of the first applications was to print the General Chemistry computer generated exams on a IBM mainframe at UM which was capable of upper and lower case. The MSU CDC mainframe could only handle upper case. The MERIT network has evolved into the regional public network for the state of Michigan. The network was renamed MICHnet in 1988 to differentiate the name from the parent company, MERIT, Inc. which began operation of NSFnet at that time.

In the same time frame, a wider area project called ARPANET was under way. ARPANET was a wide area research network connecting a number of defense contractors, large research universities, and other defense related entities. ARPANET evolved into what was called the "Internet" by the late 1980's. MSU became involved in ARPANET only shortly before the transition to the Internet. During the Mid and late 1980's MSU participated in HEPnet, the High Energy Physics DECnet network, and the UNIX networks Csnet and usenet which were promoted by the computer science communities. During that time, MSU also participated in the email network Because It was Time NETwork, BITNET, sponsored by EDUCOM. By the early 1990's, these networks had all coalesced into the "Internet." In all cases, these networks were available to the Chemistry Department via the building and campus networks.

Instructional Computing

Atkinson's 2nd Law of Computing: Unfortunately, the understanding of a modern computer operating system somewhat resembles Dante's vision of Hell. There are many levels of understanding and each is circular. That is, every thing on a level is interrelated and it seems that you have to know everything before you can learn anything. As you proceed to lower levels, there is more complexity. For example, to understand the concept of a task(program) one must understand the concept of a file which is used to store the program. To understand what a file is, one should also understand a little about the device used to hold the file. But one should also understand a little about the task(program) that loads a program off the disk. Where does one begin? The only real answer is that the novice must jump onto the "merry-go-round" and go round and round until all the concepts at a given level are understood. Hopefully, the novice will be able to iterate as necessary for she or he to master the material on a given level. One important departure from the Dantian analogy: the bottom level is not the Depths of Hell, rather the State of Heavenly Bliss, i. e. knowing "how it all works". (Well, at least until the next version is released.)

Atkinson's 3rd Law of Computing: If you need to read a computer manual, you can't. If you can read a computer manual, you don't need to.

Atkinson's 10th Law of Computing: The only way to really understand what is happening with a computer is to have all the schemats for the hardware and all the sources for the software on your bedside table and read them every night.

The department's instructional uses of computing have always been bimodal: instruction on how to use computing and communication versus the use of computing and communication to deliver and augment instruction.

As indicated by the lecture sized slide rule shown in Figure 3 and the "Friedan" teaching/research facility shown in Figure 2, the Department has always been committed to teaching the "tools of the trade" to our faculty, staff, and students. This tradition has been carried on in many venues. Stan Crouch and Chris Enke have included the use of computers in their instrumentation course, CEM 838, since the mid 1970's. The topics have included the electronic principles used in computers, computer interfacing, computer programming, and the operation of the computer. Included were units for OS8 for the PDP8s developed by Keith Caserta and Brian Hahn, units on RSX-11, VMS, UNIX, MS-DOS, and Communication developed by me, units on "c" and the National Instruments interfacing technology developed by Eric Hemenway and Jim Ridge. Computer tools have been integrated into the chemistry curriculum by many members of the department. The main thrust of the departmental SGI visualization facility is to teach the undergraduate and graduate students the use of modern chemical computational and visualization tools.

The use of computers to deliver and augment instruction began with Bob Hammer, Paul Hunter, and Harry Eick's efforts in General Chemistry. An exam generating facility helped to create exams for the large enrollment courses. Various grade keeping facilities were developed first on the CDC mainframes and later on the IBM PCs. Gordon Galloway and Paul Hunter have developed a number of stunning general chemistry tutorials that are delivered to the student by Authorware on the Macintosh and PC. In last year's newsletter, Dave Morrisey described the CAPA system for generating individual homework assignments.

An NSF grant enabled the department to install a lab in room 121 with 30 networked Zenith 286 PCs in 1989. This provided a setting for group instruction with each individual having a workstation. Gordon Galloway implemented a corresponding MAC Lab in room 120 in 1991. The PC Lab was upgraded to 486 based PCs in December 1994.

In Conclusion

This article is primarily a personal reminiscence. I certainly have not been able to do justice to all the computing developments that have occurred over the last 5 decades. More importantly, I have not given credit to all the many, many faculty, staff, and students that have expended large amounts of their energy doing, made significant contributions to, and have interesting stories to tell about computing in the Chemistry Department. Maybe they will have a chance to tell their stories sometime. I, myself, have other stories to tell. If you ever get a chance, drop by and get me to tell you the "Vector General/Neptune Van Line" story.

TVA

1. References
2. Jack J. Dongarra, "The LINPACK Benchmark: An Explanation Part I", Supercomputing Magazine, 10, Spring 1988.
3. Jack J. Dongarra, [http://www.netlib.org/utk/people/JackDongarra.html] November, 1995.
4. Acronyms page 7, Volume 24, Number 3, Summer, 1993.
5. J. Sutherland Frame, "MISTIC, The First Computer at MSU." [http://www.msu.edu/clweb/FrameMISTIC.html], November 1995.
6. "Mainframe History" [http://www.msu.edu/clweb/mainframehist.html], November 1995.