Guest Column: June 2021
COMPUTING HISTORY, Part 2
EGGS, CHICKENS, DATA
Arkansas’ poultry industry flocked to computers early
Past President, Arkansas Academy of Computing
THE 1950S AND 1960s were a pivotal time for the poultry industry in Arkansas. Both local entrepreneurs and northern producers moving south were keen to take advantage of our state’s milder climate, lower costs, and a farming population eager to meet the growing consumer demand for chicken. The state supported the rising industry with research into nutrition and poultry genetics, as well as a robust rural roads and bridges program to facilitate the growing line of trucks filled with feed and chickens.
As usual, underneath the business structure was data. How many birds in breeder flocks produced the eggs that hatcheries nurtured into day-old chicks, which then went to farms and into broiler houses? How much corn, soybeans, and milo went into nutritional feed for those broilers, and how many pounds of meat were produced? How much money was spent, invested in living inventory? And at the end of the process, how much profit would there be?
In the beginning, most of the records were manual spreadsheets and ledger cards. Records for each grower and flock, each set of eggs, each feed delivery were recorded, tabulated with adding machines and desk calculators, and cross footed for control. This was the story at Peterson Industries in Decatur when Donna Amos—a charter member of the Arkansas Academy of Computing, and a strong source for some of the stories in this column—began working in the office there in the late 1950s. At Peterson and elsewhere, manual work was replaced first with office machines specially designed to handle ledger cards and similar documents, combining some aspects of typewriters with those of adding machines.
As companies grew, the equipment changed and Peterson and others, such as OK Feeds in Fort Smith, moved on to unit-record equipment. With that, data was punched into “unit records”—cards with 80 columns of info—and fed into special purpose electromechanical equipment for sorting, tabulation, and collation. The database was the cards themselves, many thousands of them—12,500 to a megabyte. That’s seven feet of cards!
For anyone to access the data, the cards had to be sorted into sequence and fed into a reader. In the right sequence, the cards could provide totals for hours worked, pounds of corn purchased, total expenses for a broiler flock, or many other applications. The sorter put cards in the right sequence and could select just the ones required for a particular job using control codes punched in each card, part of the database design. Other unit record equipment could total columns on those cards, punch the totals onto other cards, or print details and totals. Collators could merge two sorted decks of cards and sort them into pockets.
By the 1960s, progressive poultry companies were looking at computers to better handle data. IBM had a highly active presence in Arkansas and was selling models like the 1400 series to banks, insurance companies, and others, but the prices were too high for most poultry firms. Then in the mid-1960’s the IBM 1130 came out, arguably the first midrange computer and the forerunner of later IBM S/3x machines, as well as the DEC PDP-11 and Data General Nova in terms of architecture. The 1130 boasted 8,000 16-bit words of memory and a single removable disk that held roughly 1 megabyte of data. It was designed for scientific use, had floating point arithmetic, and the primary programming language was FORTRAN. The 1130 with its card reader and printer cost just over $100,000, a bargain at the time.
In 1967 OK Feeds was the first in Arkansas to buy, and they had a special reason. OK had originated as a commercial feed mill making feeds for many kinds of animals, not just chickens, but it had expanded into broiler growing and a processing plant, then added breeder flocks and a hatchery. As a scientific computer, the 1130 had a linear programming package with the tongue twisting name “Linear Programming Mathematical Optimization Subroutine System”—LP MOSS for short. It could mathematically formulate animal feeds of optimal nutrition and minimum cost, and as commodity prices shifted, so could the feed formulas. The cost savings in feed ingredients would justify the purchase.
OK had a very experienced nutritionist, Ed Sharum, who had learned his craft over a lifetime. Sharum helped develop additional constraints and pre-mixes that made the automated formulation feasible. OK also planned to use the computer and eliminate their unit-record equipment and manual record keeping.
A year later, Peterson also purchased an 1130, but feed formulation was not on their wish list. Peterson was focused on poultry genetics and developing breeder hens and males for the poultry industry. Mr. Peterson wanted accounting and operational systems—a high priority due to the company’s complex corporate structure—but he didn’t trust feed formulation by computer.
AFTER THE COMPUTER purchases, the question was the same as today—where do you find the people to design, program, and operate systems on the computer? The answers were also the same as today: Either hire someone with technical degrees and experience, who would learn the industry, or train industry people on the computer tasks.
OK Feeds did both. Initially they used yours truly, along with my brother Charles. As sons of the OK Hatchery Manager, John E. Chamberlin, Charles and I had grown up in the business and had done hatchery records manually, but our main qualification was knowledge of FORTRAN programming, which we had learned in college engineering programs, and, in my case, also through a course in linear programming.
My brother and I worked summers and winter breaks during our college years. As the fulltime programmer/operator, OK Feeds hired Theron Liles, a person with unit-record experience who passed the programmer aptitude test to qualify and then took IBM courses. An IBM Systems Engineer, Mary Sue Jacobs, was very helpful to the OK Feeds team, introducing us to the IBM Commercial Subroutine package and IDEAL FORTRAN, which facilitated accounting applications.
In addition to feed formulation, early applications included feed and ingredient inventory, and payrolls for feed mill, hatchery, field operations, and the processing plant. Then came broiler records, breeder records, and egg projections for planning.
The 8 kilobytes of memory in the 1130 was core memory, each bit a tiny ferrite donut wrapped in wires connecting it to other bits in three dimensions. The memory assembly was a cube over a foot long in each dimension. Sixteen bits made up a word and there was an extra bit, a parity bit attached to each word. The parity bit changed as other bits in the word changed and provided a check for any errors. A parity error meant memory was corrupted, perhaps physically damaged, and usually shut down the computer.
In the first month of operation, the OK 1130 had occasional parity checks and they were coming more frequently. The IBM Computer Engineer, Joe Partridge, was on the phone to the IBM support center in San Jose, California. After some hours they determined the problem was in one core. Most likely it was cracked. Somehow from the failure pattern, they determined which core the crack was in. They told Joe what tools he needed to fix it. First needed were four screwdrivers to hold the core in place in the x and y dimensions. That required two helpers. When the helpers and screwdrivers were in place, they told Joe the final tool required, and he got out a ballpeen hammer. He placed another screwdriver above the column of cores that included the cracked one and hit it firmly. That was supposed to force the crack together, and it did. The 1130 was up and running again until they could send a new module. In the meantime, the back of the 1130 was kept open, the memory module swung out over the air conditioning vent, and the temperature turned down. That cooled the faulty core and contracted it, ensuring the crack stayed closed.
Within two months of the arrival of the 1130, IBM was bringing visitors to look at the operation, including a group from Japan. Feed formulation dropped the cost of cattle feed so much that producers from the upper Midwest paid attention.
AT PETERSON, THE staff was given the opportunity to take the programmer aptitude test. Five people qualified, and then went to IBM classes on FORTRAN, IDEAL FORTRAN, and the Commercial Subroutine package. Each person was then assigned to convert applications. Very soon, Donna Amos became the lead systems person, displaying a gift for understanding systems, growing technical expertise, and a decade of familiarity with the applications. Donna also remembers the help given by Mary Sue Jacobs. IBM SEs were an important part of the success of rolling out computers in Arkansas.
Peterson’s 1130 was a newer model than OK’s and had two disk drives. Each application tended to have its programs and data on a drive, because of the limited storage space. Data entry was on IBM key punch equipment. Some applications, such as payrolls and inventories, could be written with the master files on disk. This meant using either sequential access or some form of random access devised by the programmer. There were no built-in indexed files. For other applications, the workflow closely resembled that of the unit record equipment. Cards were sorted into sequence, read in through the card reader, and tabulated while reports were printed. There wasn’t enough disk to sort thousands of cards for broiler applications. Instead, reports were printed on the line printer, often with multi-part carbon paper copies. Distributing paper reports was an important part of the operator’s work.
After successes with the 1130s at OK and Peterson, IBM was able to sell to other poultry firms by the early 1970s. By then the machine of choice was the newer IBM System 3, or in the case of Tyson, an IBM 360 model 20. The poultry company System 3s were among the first installed in the state. The tendency for poultry companies to be first, or nearly first, in line for new models continued.
Later, Banquet Foods became the first IBM System 34 user in Arkansas, running a network from Batesville with links to Georgia, Missouri, and New Jersey. Produced by Arkansas Systems, this package was later sold to Ralston Purina de Venezuela. Donna Amos at Peterson realized the great power of the physical and logical file system on the IBM System 38, and Peterson became the first Arkansas user of that system. Peterson had a computing application with many integrated modules and was able to sell their software to other firms, including Perdue Farms in the Delmarva region. In poultry systems, as in many other areas, Arkansas software solved problems not just in the state, but also across the nation and the world.