% fortune -ae paul murphy

Distributed decision making in IT

When I tell people responsible for anything from tens to hundreds of IT staff serving hundreds to thousands of users that they should change their own roles, cut most of their staff, and change their technologies to make IT more effective, more secure, and more user friendly, they generally think I'm either crazy or joking.

I'm not joking - and while I admittedly don't expect the average CIO to cut his own budget and span of control, the truth is that his bosses ought to do it for him.

The key here is that IT effectiveness is all about people, not technology; but getting IT people into user groups and empowering them to act is always much easier to prescribe than to do - and doing it all with the typical technology in place is, in most cases, simply not feasible.

Imagine, for example, trying to apply the loosy-goosy practices that allow Windows users in small professional organizations to mostly ignore the problems that come with the technology in much larger organizations in government or industry.

Imagine thousands of users with unique copies of proprietary corporate data stored on personal computers; imagine hundreds of work groups with different back-up strategies and, subtly different software choices, all periodically finding resources to re-invent and re-roll the same wheels; imagine the impact of a simple discovery requirement in a corporate IT context made up of thousands of different and competing, user managed, workgroup centric PC operations; or imagine a dozen major groups all insisting on different time share suppliers to run the corporate ERP.

That's a pretty horrific notion isn't it? But in rejecting it are we rejecting the baby or the bath water? or, in other words, is it the user empowerment idea or the technology that's wrong?

Replace all those Windows servants, servers, and desktops with a single central SPARC/Solaris system feeding mostly Sun Ray desktops and it starts to look possible to manage something like this in a smart way. With this technology neither Dick nor Jane's insistence on creating and managing their own data causes any problems: the sysadmins working in their groups simply create new schemas in the central RDBMS implementation and everything -backup, data security, multi-point access, code re-use - works as it should.

Go down the list: from document sharing to authorizing selected customer access to the MicroLath schedule in Materials Testing, the sysadmins do the work, co-ordination comes with the resource, and the net effect is to provide services in a Fortune 1000 context with the speed and flexibility we normally associate with the IT guy running a few servers in a 30 man real estate office.

It sounds absurd, but all it comes down to this: the PC doesn't scale, and throwing manpower at that problem just makes it harder and harder to control the risks that go with the thing; but Unix does scale, so putting educated super users (sysadmins) into user groups and telling them to just get the job done subject to the minimal rules needed to sustain the central system, can work.

Paul Murphy wrote and published The Unix Guide to Defenestration. Murphy is a 25-year veteran of the I.T. consulting industry, specializing in Unix and Unix-related management issues.