Barry Merrill on SAS, MXG and His Button Collection
Reg Harbeck: Hi. This is Reg Harbeck and I’m here with Barry Merrill who is somebody who has worked with all kinds of interesting data on the mainframe. Barry, maybe you could start by telling us, how did you end up on the mainframe?
Barry Merrill: Well what happened was, my brief career started actually in 1959 when I was a sophomore at Notre Dame and I programmed a 4×4 determinant on an IBM 610. For the first time a digital computer was used for a lab course at Electrical Engineering but I didn’t play with computers again until I got back to Purdue in 1964 on a Navy scholarship. I got involved in Fortran and did quite a bit of programming for the Landsat, the original earth satellite detection programs but that then put me back in the Navy and I was in Navy submarines for a while but got out in 1972.
I happened to go to State Farm where a friend was working and invited me to interview. Dave Vitek had decided it was worthwhile to measure computers. Circa 1972 there were no tools that measured computers. You basically did your capacity planning by the IBM salesman coming to your vice president and saying “Here’s a contract. Sign it; you need a bigger machine.” But Dave had gone to a Boole & Babbage, one of the early Boole & Babbage user group meetings and realized that there might be a way to measure and do it yourself and so he funded a group of ten people. I happened to be the second person that came in because I had some experience from Notre Dame in statistics and in Fortran but I knew nothing about the OS. I took the Deltek course while I was there and actually in the second week I was at State Farm I found this small announcement that SAS, the statistic analysis system, was available from North Carolina State University for $100, over 100,000 lines of code, one third Fortran. I mean I remember the advertisement but I sent off and got the tool.
In 1972 the input statement of SAS was beautiful. There was no preconditioning. You don’t have to preallocate. You don’t have to initialize. It covered your ass completely in programming and was able to handle all of the data formats that I could see in the SMF record except that, as I read through, there was no support for packed decimal. We only have to get four bytes into an SMF record to find a packed decimal field so I called the department of SAS, the department of statistics, at North Carolina State and got Tony Barr—the author of the complier who wrote the program. He and Jim Goodnight, the statistician, had been working on this since 1964 and I said to him, “Now we’re looking at this, but to process SMF data.” He said “Oh wow, that’s an interesting data source.” I was surprised he knew what because in 1972—SMF was announced in 1969 but didn’t really become functional until 1971 or 1972—but Tony Barr in his slow response said “Well it’s like this. We haven’t got around to documenting it yet but if you type in ’PD4.’ it will work just fine.”
Well with that assurance State Farm risked their $100, I had SAS installed in a matter of half an hour and within two weeks we were going to the local Computer Measurement Group. We were going to SHARE and presenting papers on how this SAS was really powerful for processing data but actually our presentations were only half on SAS and the code. The other half was on convincing people they needed to staff a computer management person or a capacity planner, that there was a real need because they could save money; half our studies showed how much money we saved by using SMF data in the role of capacity planner and so that is how I became very heavily involved with SAS, and MXG is written in SAS and it’s the primary tool – that’s basically where I got started.
Reg: Now how long after you got involved with using SAS did you start developing MXG?
Barry: I was working at State Farm but I was also commuting to Champaign, Illinois in a CB500 Honda—by the way on my Ph.D—and so I ended up using the development—State Farm allowed me to use the code development and analysis as a part of my doctoral dissertation. They then allowed me to carry that code when I left State Farm and they allowed me to carry the code to Sun [Oil] because it was a part of my doctoral dissertation. So when I was at Sun, I made quite a bit of an enhancement but Sun regarded it as my code when I got there even though I was an employee and even though a lot was built up. I had obviously very generous people who recognized what I was going to do with this and so especially when they heard the price, they couldn’t believe it. Our initial price was $500 and we had a discount for a second site of $350.
Reg: Nice, very good.
Barry: That was the connection that let us create and what really happened was in 1984. Well in 1980 SAS put on—I lost thread there of what I’m trying to say. You can erase that if you need.
Reg: No, we keep it.
Barry: I know. In 1980 I got a call from the publisher, the vice president for publications at SAS Helwig, who asked if I would write a book, that they were getting a lot of calls and people and so I put together a 350-page book that was called “The Merrill’s Guide to Computer Performance.” Actually the original and working title all the way up to New Year’s/Christmas Eve when it went to the publisher was “The Analysis of SMF and RMF Data Using the SAS System.” Cathy called me and said we don’t like your name. Can you handle “Merrill’s Guide to Computer Performance Evaluation?” I decided my ego could and that book was sold for $395 a book.
Reg: Wow.
Barry: And it came with a sample tape but it was not supported software. There were only like 55 programs on the sample tape but it was so pervasively easy to use SAS with my examples that 1,500 copies were sold in a matter of probably two and a half to three years.
Reg: Wow.
Barry: That’s what really opened up and Goodnight recognized that I’m responsible for half the sales income. I mean at least, because we really penetrated the computer market through from the back door through system program where all of the other products were going over to the application development folks who didn’t seem to have the awareness of tools that could be used but coming through the back door and then making it available. That’s what we did at State Farm. They had previously used a product from University of Illinois the SPSS, not SPSS, a predecessor to SPSS that had been used and as soon as they got SAS in house, that whole group switched over and began to use it.
Reg: Okay. Cool.
Barry: Of course now what SAS has done that’s really moved them forward I think is that now you can create your output not only on a SAS dataset but you can build an Excel spreadsheet. I see lots of my users do that because many corporations require Excel graphs as a standard tool and so you pop it into Excel, you get the spreadsheet and you don’t have to know a word about SAS.
Reg: Okay.
Barry: That’s an advantage for some people.
Reg: I find that a lot of the projects on the mainframe that involve doing something that processes data involve quite often downloading a chunk of that data and processing it using Excel, changing it in some way and then sometimes sending it back to the mainframe. You’re seeing that a lot as well I gather.
Barry: Well not the processing of the volume of Excel. What I’m seeing is that they are taking—the volume processing is the gigabytes that we’ve created—
Reg: Right. Yeah.
Barry: That builds a small selection and they simply create their essential PROCPRINT and send it over to Excel or SAS builds it directly in Excel.
Reg: So a small selection for a particular purpose.
Barry: And then they typically use that for their graphs.
Reg: Now you of course have a long involvement with SHARE and also a bit of a reputation as having a lot of personality including your collecting of buttons. Tell me how did you get involved in SHARE and how did you start collecting all those buttons?
Barry: Well good point because the very first SHARE I went to—people had been attending SHARE from State Farm but no one had been in this—there was a project called Computer Measurement Evaluation and to get into that project unlike any other SHARE project, you had to present a paper, two to five to ten pages to prove that you knew something about performance measurement.
Reg: Sounds a bit like CMG.
Barry: Yes, but this preceded it quite a bit and also we ended up from about 1972, 1974, 1975 and 1976. Each of those years we had a 500-page annual document of the papers for people trying to get in CME.
Reg: Wow.
Barry: And we had massive fights with the board to fund that by the way. Well, not so massive fights, but we used to go to the 4:30 open board meeting on Thursday, 20 of us from the project area all standing up and, “publish our notes!” But at the very first SHARE I went to, I came prepared. I’d sent my note in on SAS and I did a 15-minute presentation on SAS. And it was really great because I would have a graph of something and it would be the three lines of SAS code that generated that graph. It was really beautiful.
So I did my 30 minute presentation and Tom Bell, who was a project manager, looked at that sergeant-at-arms and said, yup general, which meant that yes, you’re going to do a general. They were ready for it for the next session so the next SHARE which happened in Chicago which was the spring of 1973, I was scheduled to do a shared presentation with Bill Tetzlaff who had the SGP, this Statistics Gathering Package, that read three SMF records, the four, the five and the 14 and had fixed canned reports but no language. IBM at least was getting out there. Bill was tasked not to create a language but to create some kind of reporting. Well, he got up, being IBM, and he gave his pitch first. He finished his 30 or so minutes. Then I stood up and I gave my pitch on “Here’s how we use SAS at State Farm.” At the end, I have no idea who he was but he stood up and he says “Well, Mr. Tetzlaff, would you come back to the microphone? Now that you’ve seen Barry’s description for SAS, is there any reason now you’d recommend to us to get your statistics gathering package?” I mean a little bit hostile but it was night and day for the attendees. There were 1,400 people in that room.
Reg: Oh, wow.
Barry: Which is how many people who got introduced to SAS. Over the next years obviously SAS exploited that, used that and people got involved.
Reg: Cool. Now how about the buttons? How did you end up being a well-known button collector?
Barry: That same very first SHARE I attended physically, there was this big burly guy from a big chemical company in West Virginia that I can’t remember right now, nor his name, but he was just a really big guy. He was wearing this purple sash and on this purple sash were about ten buttons. I’m trying to remember what some of the early buttons he had on his jacket were. These were buttons that had been created before SHARE to lobby for some point, typically 100 buttons had been made and they were sold for maybe $.50 or $1.00 to recover the cost. They generally attack IBM things in a humorous manner. These have been created by individuals really for the purpose of humor. They weren’t there to make money although much later a guy did come in and basically mass produced 20 or 30 buttons each year but all of the early ones were really lobbying for a point especially to lobby IBM for improvement of things. Let me pause for a moment. I’m trying to remember what an early issue was.
Reg: There was a source code issue was one of them.
Barry: I’m sorry?
Reg: Source code.
Barry: Oh, yes. Yeah, “We want to fix it in the language it broke in” was one of the preaches for that. There was also the OS360 was up and down and my DOS was very smooth. There was also the—I know what I was trying to get to earlier. At that time there was a merger attempt between Guide and SHARE. The merger failed. It failed actually because there was not enough quorum to vote but everyone knew that Guide voted in favor and SHARE voted against and SHARE voted against with buttons that were anti: “Divided we stand, together we fall.”
There were several other buttons that were lambasting Guide. “Guide loves JCL” was another. So these were the buttons printed in small quantity but after I saw his buttons, what would happen is these buttons would show up. People would get them, take them home, and they weren’t there the next SHARE so I decided I took my first buttons that I got that first meeting and then the next meeting I carried them. About the third meeting I had a friend who was adjacent to an old butcher company that had failed and they had some really nice white lab coats that the butchers wore. He gave me one and that’s what I put all the buttons on. The full button collection is about 2000 buttons, fully documented with pictures and where they came from. For everyone who wants to know at www.mxg.com in the lower left corner you’ll see “The Buttonman Collection,” (https://www.mxg.com/thebuttonman/) so my whole collection is online.
Reg: Awesome. Well this has been a great discussion and clearly it is just barely scratching the surface. You’ve got such a depth but thank you so much for sharing all of this with us on this. Any final thoughts you wanted to share with people?
Barry: Just that SHARE is the most incredibly valuable resource I’ve ever had. I mean even today I learned there’s a new SMF record I’m supposed to know all about.
Reg: Cool.
Barry: So yeah, the recommendation is anybody needs to go to SHARE if they’re going to be a z/OS systems programmer or capacity planner. One week at SHARE is worth a year of college education in these subject matters.
Reg: Awesome. Well thank you so much Barry.
Barry: My pleasure.