In-Depth
Top Data Center Issue: People
Skills, whether in distributed or mainframe computing environments, are in increasingly short supply.
In response to my most recent column on data centers, I received e-mail from several readers that focused on my comments on mainframes.
One reader wrote, "I read your article on data centers and I am particularly interested in what you had to say about MFs [mainframes]. We are in an effort right now to move off our mainframe to a distributed environment. Although you discussed mainframes from an operations perspective, you did not say anything about them from a development perspective.
"One of the concerns we have with MFs is the availability of developers in the next 5 to 10 years. Our average MF developer age is 52, the youngest being 36 and the oldest 65. No one is teaching MF skills these days. In a 5- to 10-year window, getting MF development resources will become a challenge as the baby boomers who make up this labor pool retire. I would be interested in your comments regarding this."
My response to this reader is summarized in the following discussion. There are actually three fundamental issues here. First is the question of whether labor shortages will impact the availability of operations and development personnel who are required to keep mainframes viable in the contemporary data center. Second is the issue of labor availability and cost in general. Third, and too often unmentioned, is what should guide the "platforming" of applications.
Considering the first issue -- the availability of qualified personnel -- the logical place to start is with the statement made in my previous column (see http://esj.com/enterprise/article.aspx?EditorialsID=3374). Even in the current employment climate, which by all accounts promises to worsen over the next 12 months, projections of job creation in the technology field remain mostly unchanged. According to the Bureau of Labor Statistics, the U.S. is expected to create 650,000 to 800,000 new IT jobs between now and 2010. Quoting CA's vice president of innovation and strategy, Vince Re, institutions of higher education are expected to graduate only about 100,000 IT job candidates in the same time period, creating a shortfall of 550,000 or more qualified applicants for IT job positions.
The decline in IT-oriented degree program enrollments has been discussed for some time. Not long ago, Microsoft and other tech companies were arguing that they needed more offshore workers to be allocated work visas because of current and future labor supply deficits. The matter became politicized and the discussion was beset by populist rhetoric that raised the ire of many who worried that their jobs would go to foreign workers willing to work for less pay.
Truth be told, whatever the ancillary motivations of the tech companies in seeking workers from abroad, their statistics on labor shortfalls appear to have been correct. Despite the increase in investments from tech vendors in colleges and universities to fund new department chairs and technical disciplines within existing departments -- including many mainframe technology-oriented curricula -- the number of students choosing technology degrees continues to fall to pre-1980s levels.
Part of the reason for low enrollment, many experts observed, was the increased enrollment of women in universities. With IT perceived as a "bloke culture," as one education official told me, women have demonstrated significantly less interest in the IT discipline. Adding to the demographic issue was the diminished interest in pursuing technology degrees among students who witnessed the impact of the dotcom bust in 2000. The chances of graduating from college and immediately realizing a highly paid position in a high-profile tech company were very low. (This perspective might change, some say, because of the job crunch expected to hit the global economy over the next year.)
Finally, discouraging data on rates of pay in tech (IT worker salaries have barely kept pace with inflation, unless you are an IT auditor, according to recent data) and no shortage of disgruntled commentary from contemporary IT practitioners regarding the discomfiture associated with their jobs are also factors. Cost-cutting efforts in the front office have seen significant consolidation and dislocation of IT personnel, which in turn places a greater task burden on fewer staffers while increasing fears that their jobs might evaporate in the next downsizing interval.
Bottom line: IT has lost its appeal to the young and they are responding by enrolling in other degree programs. The coming labor shortage is systemic.
The Age Issue
The reader's claim that his development group on the mainframe is aging and that this is an incentive to move applications on the mainframe into distributed systems, where presumably skills are more available, is one I hear too often. Truth is, managing, operating, and developing applications in the mainframe world requires fewer people than the same tasks require in the distributed world. Conservative estimates place the difference at about 10 to 1, which is to say that for every one person working in the mainframe side of the glass house, there are 10 doing comparable tasks on distributed platforms.
Statistically, as Re correctly observes, this translates to more problems for the distributed environment than for the mainframe environment. Assuming that the systemic gap in qualified IT workers applies to both environments, distributed computing environments may be harder hit than mainframe shops by the coming labor shortfall. I would be very concerned about moving applications off the mainframe, were I the reader, and hosting them on platforms that require more people to support them.
Re also observed that a senior mainframe operator/developer is probably less expensive than a distributed administrator. A rarely analyzed trend shows that mainframe personnel pay, adjusted for inflation, has not increased markedly since 1985. By contrast, the salaries paid to workers in the distributed environment, including the need to retrain periodically to capitalize on cool new technology that appears every six months or so (especially in the world of the Internet) have accelerated overall labor costs.
Re is correct when he points out that mainframes have matured into simpler and more automated platforms. Current struggles to consolidate distributed platforms using X86 virtualization run afoul of the complexity and proprietary aspects of the disparate heterogeneous technologies that have been deployed over time -- typically in the near-complete absence of unified management. This observation isn't made to deride client/server technology, only to underscore a key truth in Re's commentary: human costs are greater in distributed environments partly because of the need to hire personnel to audit the work performed by other personnel and the jobs they are doing to administer distributed infrastructure. Re claims that, in his data center (the largest on Long Island), it is 50 percent less expensive to run Linux on a mainframe than on other server platforms.
Choosing a Platform
Whether or not you agree with the preceding arguments about skilled labor availability and differences in labor costs, the third issue -- selecting an appropriate platform to host an application -- is really the most important. Hardly anyone would disagree that applications have their own requirements (and nuances) that must be considered when designing an appropriate hosting platform. The business process being supported also dictates how data must be handled, both from the viewpoint of budgetary constraints and regulatory requirements. These two sets of parameters dictate an appropriate hosting environment.
A problem encountered in many shops I visit -- especially in those where high-performance transaction processing apps, for which mainframes are excellent hosting choices, have been migrated onto X86 server platforms -- is that the resulting poor performance and high labor costs offset whatever cost savings were anticipated to accrue to the move. In one financial services company I recently visited, Oracle had convinced the CIO to embrace the X86 hosting alternative for a database containing over 100 years of financial trade reconciliations. On the new hosting platform, the database was taking over an hour to load in production and even longer to load in the CIO's test/development lab.
His question was simple: what could be done to improve the load time? I suggested that he should begin by firing whoever came up with the bright idea to migrate the app off the mainframe. After he left his chair and exited the room, the mainframers gathered in the conference room broke into unbridled laughter.
The point isn't that the mainframe is a one-size-fits-all application host. Different applications and business processes require different platforms. The key to containing the costs of infrastructure is using the right platform for the right app -- including, in the distributed world, the right kind of server, network, and storage.
This view flies in the face of vendors who market their wares as a "one-stop shop" for all applications -- a characterization that was resoundingly rejected when IBM used it in the context of mainframes 20 years ago. In the distributed world, so-called SANs, for example, were once described by Veritas (now Symantec) as "an aqueduct that gives all applications a drink of storage." We are only now waking up to the fact that not all applications want the same drink of water, that some want Gatorade instead, or a Red Bull, or a beer.
Approaching the issue of cost in the data center requires that we revisit the rationale that we used to build infrastructure over the past decade or so. Technologies, such as server virtualization, are forcing many shops to do this today. We need to ensure that the platform is still aligned to the requirements of the application and the business process, and fix them when they are not.
Of critical importance in all cases is management. With skills, whether in distributed or mainframe computing environments, in increasingly short supply, Re is correct in his observation that software will need to pick up the slack by simplifying operations and improving the productivity of a shrinking staff.
Thanks for the encouraging feedback to the inaugural column. Your comments on this installment are welcome: [email protected].