x'23' (35): TSS and Virtual Memory on the S/370

posted Dec 14, 2014, 10:22 PM by Jeff Ogden
This is the second of two excerpts from a 37 part oral history interview of Humphrey Watts by Grady Booch for the Computer History Museum. Watts was an executive at IBM where for a time he was responsible for all software development at IBM. Later he was a Fellow at the Software Engineering Institute at Carnegie Mellon, where he provided the vision for, and early leadership in the development of, the widely used standard for assessing an organization's software development capability, the Capability Maturity Model (CMM). He is the author of several influential books on the software development process and software process improvement. Watts passed away in October 2010.

All 37 parts of the interview are available online, see:  http://www.informit.com/promotions/interview-with-watts-humphrey-137746.

This excerpt doesn't mention MTS directly, but covers two issues that were important in the history of MTS, IBM's TSS operating system and the decision to include Virtual Memroy support on all models of the IBM S/370 computer line.

An Interview with Watts Humphrey, Part 10: The Fortune Interview, IBM Lawsuits, and Virtual Memory

April 26, 2010

http://www.informit.com/articles/article.aspx?p=1574479

In this transcript of an oral history, Grady Booch interviews SEI Fellow Watts Humphrey. In Part 10, they discuss IBM in the mid-1960s, including Humphrey's unfortunate Fortune interview, teaching lawyers about programming, and making the decision to use Virtual Memory in the IBM 370.

This interview was provided courtesy of the Computer History Museum.

 . . .

[TSS]

Also I think towards the end of that year we had a meeting -- the TSS system had gotten into trouble, as I mentioned. We had gone and added a whole lot of function to it. And instead of coming out three months late, it was about six to seven months late with performance problems and everything. And so basically the system 360 was coming along and people were starting to buy it and they were happier with it. And so people had switched back from TSS to 360. So 360 was now starting to go full tilt. And the management decision, the division presidents and that whole crowd all said, “No, we’re going to kill TSS, period.” And I objected because I thought it was a system we should have available, but no -- they were going to shut it down. So they did shut it down. It cost us about $30 million. We did actually get the system running, and it was installed in a few places with the Model 67, but it was stopped. But it was the early virtual memory system, and it was really a very good responsive system, but it was not compatible with 360, and that was a real problem that people were concerned about, and it was out.

So in any event, I remember meeting in the board room. I was talking about software and software phase plans and the whole thing, and Tom Watson interrupted me at one point. I had gone through what the phase plans were and when you announce things and when you do various stuff and he said, “Watts, I’m confused now. You did a marvelous job with the FAA and you’re probably the best guy we’ve got to run software. I don’t understand. The FAA was such a tremendous success and you’re doing so very well with the 360.” He said, “How come TSS was such a disaster?” We had just closed it out at a $30 million loss. So I said, “Look, here’s where we announced the 360 schedule.” And I showed him we had running code, we had a whole lot of stuff in place -- at least the beginning code -- and we had plans, and the design was done, et cetera. And I said, “And here’s where we set the schedule for the TSS,” and it was way back at the beginning before we knew anything. And I said a big part of the problem on controlling this stuff is announcing things you don’t know how to build. We didn’t have a good foundation for a plan. So Tom understood that, fortunately. He was quite a guy. He could be really tough. I remember an executive making a presentation to him and he actually reduced the guy to tears at one point. But he was logical. And if you could understand what he was concerned about and really get to the point, he’d switch, and he was great. So that was that.

 . . .

The Virtual Memory Decision

So I was made VP, and I had the architecture stuff. And as I said, Don Gavis had the OS 360 work. And he came to me -- I think it was in about 1969 – and said, "We've really got to go to virtual memory." And remember, the TSS was killed and the 360 didn't have virtual memory. And we had that big battle with Amdahl that… “just add more memory.”

But the programmers had concluded that virtual memory was probably the only way to go. We just had to do it, get out of the constraints of the physical memory. And just about that time, IBM was developing a newer version of 360 called the 370. We'd been out there for a while with the 360 systems, and people were beginning to catch up with us, in terms of performance and that sort of thing. And there was quite a lot of competition. It was still pretty vicious. And so they were coming up with upgrades and higher performance hardware and that sort of thing. And they decided to call it System 370. We were working on 370 development work. And so the recommendation was that we switch System 370 to be a virtual memory system.

Well, that was a radical change. Some hardware guys were happy to do it, but a bunch of them weren't. We had a bunch of microcode machines, which were fairly easy to switch. But the hard-wired, bigger systems were a much tougher problem. And right about that time, they had a re-organization, and a new Division President was brought in. And guess who was made the Division President? It was Bob Evans, the guy I'd had two previous battles with -- on both the timesharing stuff and on the FAA thing -- and I won both of them. And so Bob and I weren't on the best of terms. He'd gone down to run the Federal Systems Division, where they programmed the FAA system, if you remember. And he was brought back as Division President.

And so one of the first things we did was to go in to Bob with my programming team and the architects. The architects agreed that we ought to move to virtual memory. So we went to Bob. And of course, we were fighting with the hardware guys. And so Bob looked at it and went through the story. I mean, I have tremendous admiration for the way he was able to take a multibillion-dollar decision and make it in a day. I mean, he went through this, he looked at all the choices. And he said, "You're right. We'll do it." And here he'd been opposed to it, but he went through the logic and what the guys were talking about. And he was a very sharp guy. I had differences with him, but he knew what he was doing. He made that call and he was clearly right. So that's how we put virtual memory in 370. Bob made the call and it was, basically, Don Gavis that turned us around.

 . . .

Comments