Which makes more sense: fighting for the experienced programmers and managers with deep experience in Windows, Linux, and Unix platforms, or building up a modest staff of mainframers to support a set of applications? The costs of experienced IT people on mainframes has, according to Vince Re, chief architect at software maker CA, has remained remarkably stable over a long period of time.
This cannot be said of other platforms, where turnover is higher and compensation is forced upwards through competition.
Re has a slightly different angle about how an impending shortage of IT personnel might play into the hands of the mainframe, and he shared his thoughts with us recently.
Q: Anytime anyone says that the mainframe is going to have some kind of advantage when it comes to a labour shortage, my ears go up and I want to hear this theory.
A: You sound like a skeptic.
Q: Well, I am, professionally speaking. That is my job. But seriously, IBM is doing a bunch of things to address the skills issue by making mainframes easier to use and administer by putting more wizards into the operating systems. You seem to be taking a much more broader stance on the issue than this, however.
A: There are a handful of things that I look at to come to this position. One of the things is that when it comes to the mainframe, it is not really an ageing population problem. The real problem is that, as an industry, starting in about 1990, we did not really create any new mainframe talent. Everybody was lured into the client/server world and distributed computing - whatever you want to call it - and there was never that crop to fill in. An interesting thing there is that a lot of times, when people went into those other platforms, they had mainframe skills, they just made different career choices. We see this here at CA. A lot of our senior management and product development folks were from the mainframe. So to an extent, there is still a big pool of talent out there. They just may be hidden in a lot of organisations. It is not that those folks just vanished overnight back then, they just went into different directions. I think this fact tends to lessen the near-term impact of labour shortages in the mainframe space.
Q: And when you back former mainframers out of the projects they are working on now in distributed systems, it is then presumably easier to fill in the gaps you have for employees in for Unix, Linux, and Windows programmers or system administrators, since these skills are in a larger supply.
A: I think that this is certainly one of the angles here. It is easier to find a Unix or Windows developer, or even at a higher level, a Java developer who is transferable across platforms.
Admittedly, that is a small piece of the equation here. The bigger piece, in my mind, is that we have this view that demographics are catching up with mainframe computing when, in fact, as we dig deeper and deeper, this is happening for all of IT, not just the mainframe.
Our government tracks statistics that show how many computer science graduates we have produced as a country each year for the past eight years, and it is down fully 70% from where it was in 2000. I saw a survey of college freshman recently, and fewer than 2% have indicated that they wanted to study computer science. What we have been telling these kids for quite some time is that these jobs are going to be offshore, that in the global economy there is no future in it, and this has dissuaded many of them from studying computer science. The US Department of Labor cites IT as the fastest growing part of the economy after healthcare, and they project that there will be another 600 000 new jobs available in the next five years, but there will be only 200 000 qualified Americans to get those new jobs.
So that tells me there is a big gap in the supply and demand curves.
These are not mainframe jobs, obviously, and I have got to believe that a very small percentage of them are mainframe jobs. If I am a CIO that is trying to deal with this issue today, one of the things that I have got to think is that, at least in North America, there could be a shortage of trained IT workers. And certainly places like India, China, and other offshore areas are a wildcard in this. It is hard to guess how they will come into play. But you have to remember that these are not just software development jobs, but management jobs.
Q: Some of those IT jobs move a lot more easily than others, right?
In terms of skill sets for mainframes and legacy platforms like the AS/400, there are a lot more people with skills in North America and Europe than there are in India and China. The quality of the work becomes an issue, and in the case of the AS/400 base, which does not do a lot of offshoring, what I have heard is that the quality of the work is sometimes lower and when you add in the extra hassle of coping with remote programmers and managers, offshoring does not buy you much.
A: You bring up a very important point here. CA is a big software development organisation, and one of the things we have found is that if you are going to do development, say, in India, one of the ways you make that possible is by having a pretty big staff.
And it is expensive to do that. You have extra infrastructure costs, and time zones to deal with. You have to absorb all of these things just to get productivity out of your people. And on a small scale, it might be more economical to compete for the limited talent pool here than to try to go offshore for it.
Q: But then there is the other side of the coin. Take IBM, for example, which has 55 000 employees in India, and its presence in that country has allowed it to get a dominant position in the services market inside India. In IBM's case, it is doing big business in India, so being there is being local, not offshoring at all.
A: That makes my case for me. If you are offshoring on a large enough scale, it is a good option. It is just not clear how many organisations that are facing a scarcity in IT are going to be at a big enough scale to invest in offshoring.
Q: City and state governments have been using small mainframes for decades, and I doubt they have the budget or the political permission to offshore a lot of what they do. They do not have budget to move to a new release of the operating system and some new hardware, I cannot imagine they have budget to do offshore programming.
So let us bring this back around to a skills shortage playing out in favour of mainframes.
A: With everything that we just said as background, it seems to me that if I take a cold, calculating view of the whole environment, the mainframe has the lowest human cost per transaction. It does not much matter if a mainframe shop has a couple of thousand MIPS or a hundred thousand MIPS of processing power, a small staff can manage that. We are a pretty good example of this. We have the largest data centre on Long Island, north of 50 000 MIPS, and we have a dozen people managing it. Our SAP ERP systems runs on mainframes. I am not counting help desk and network administrators, just the people who run the systems and application software.
And they support the machines that support around 15 000 employees.
I think CA is typical here. In our case, a lot of people in the IT department are ex-system admins. I do not think we are unusual in the amount of capacity we can manage with this few people.
If I am in that CIO's shoes, one of the things that is going to occur to me is that if I have a choice between putting an application on the mainframe or pushing it to some other platform, I might want to stop and think about where the human costs are going to be lowest over a long period of time, particularly in light of this coming skills shortage across all of IT. I think they are going to think long and hard about taking younger people and training them really well on mainframe systems, and then I have the situation covered for the foreseeable future without a huge investment.
Q: Do you think that this trend is already hitting the mainframe?
IBM has had eight consecutive quarters of MIPS sales growth, even if the revenue has not grown very quickly because of the advent of speciality mainframe engines with lower prices that support Linux, DB2, and Java processing. Pricing is not perfectly elastic on a mainframe, but it is certainly more elastic than it has been.
But over 100% MIPS growth year-to-year and 45% growth over all kinds of engines is still pretty impressive.
A: By our best projections, we think that over time, IBM is going to continue to grow that platform on the order of 15% per year - I am talking about capacity - and what that translates into in terms of revenue is hard to say.
Q: Probably flat to declining sales, like everyone else in the server racket. The good news for the mainframe is that it went through the virtualisation crunch already, starting in 1990.
A: The mainframe has a key advantage here in that it is a shared resource platform. Mainframes do not have to plunk everything inside of its own LPARs, since it was designed to be shared to begin with.
I am also a believer in supply and demand, and one of the things I look at is that if it were indeed true that supply and demand were out of whack for mainframe skills, the costs would be climbing.
We would see salaries rising, other compensation rising. The truth is that if you look at the data in constant currency, from the late 1980s until now, the cost of a mainframe resource has been exceptionally stable over that time. This is simply not true on the distributed computing side, where the supply and demand for skills created all kinds of weird conditions during this time.
Q: But is not true that H1-B and offshoring have had a mitigating effect on rising IT salaries? There must be some kind of downward pressure on salaries and compensation from these factors.
A: On the mainframe, I do not know. First of all, I do not know that there is a big pool of mainframe offshoring anywhere in the world. I can tell you personally that CA tried to find that in India, and we could not find very many mainframe developers.
The problem I have with this is that it would fly in the face of the capacity growth we have seen with mainframes. If people were really successful with mainframe offshoring, we would see capacity declines. We just do not see that. I believe that the biggest mainframe companies are smart organisations, and they are growing their mainframe capacity very quickly, and we see it from the revenue it generates for us. If there was an easy path off the mainframe or the economics did not make sense, these companies would have had a very different strategy by now.