% fortune -ae paul murphy

Heads I win - tails? not my problem.

Imagine that you've magically become head of IT at a major University: you've got 65,000 students and 15,000 faculty expecting reliable email services and three real choices about how to provide it:

  1. Microsoft's Exchange Server 2003 Benchmarks data suggests that a highly tuned, four machine, dual 3.6Ghz Xeon cluster can support a maximum of about 20,000 users. In other words, you need at least five such clusters and four staff plus an email manager to run them at an expected annualized cost somewhere north of half a million - and if this were a business you'd spend nearly two million bucks more just on CALs.

  2. There's some junior guru who claims that a pair of Sun pods each with dual Sun T2000s and a single X4500 either with Lotus or some open source applications would handle the entire job, complete with self serve administrative applications and part time support, at an annualized cost well under $100K; and,

  3. Google says they're eager to host your email domain for free - or $10K a year if you want to devote part of an FTE to calling them when trouble hits - and all your users have to do is point their web browsers at gmail.google.com to get service.

All three choices impose costs on somebody - the Microsoft choice imposes costs on both users and IT, the Sun choice centralises most of the costs in IT, and the Google choice silently places most of the costs on the user.

The Microsoft choice is the default for most organizations, but is indefensible on rational grounds; not just because it's the most expensive or the least reliable, but mainly because it imposes the greatest restrictions on other user and IT decision making. Try to imagine, for example, making an Exchange Server decision for campus wide email - and then not using Microsoft's identity management tools, networking management tools, database tools, or email and web clients.

The other two choices cost less, have better reliability records, and don't impose monopolistic burdens on either IT or its users -i.e. your decision either way doesn't dictate your decisions on issues like identity management or staffing and doesn't force your users to choose between enjoying the wonders of Outlook/IE on Windows or the daily joys of discovering whether your latest patches or upgrades have silently made their browsers or email clients incompatible again.

So how do you decide between them? The cheaper one looks to be both safer and less hassle, particularly since I've loaded the description to make the Sun choice look wonky.

But what's really at stake here isn't the usual quantification of cost and risks, but the question of who bears those costs and risks. The google option looks less risky for IT, but places critical organisational records in the hands of a third party: meaning that a google decision reduces apparent risk to you as head of IT, but does so at the cost of increased risk to your organization. In effect, this choice can make you look good - but you'll last until something goes wrong and senior management catches on that you've stuck them with your responsibilities.

More importantly, a google decision shifts the cost burden from dollars spent in IT, to minutes spent by users - minutes google monetises through paid advertising on email pages.

Basically, you're selling both user attention and your organisation's records to get a reduction in IT cost - something you'd almost certainly consider unethical if google's offer consisted of a check sufficient to cover internal costs in exchange for that same access to user eyeballs and information.

Bottom line: the google decision isn't what they pay you for - and what you need to do in this case is your job: meaning that going Sun is the only responsible choice here.

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.