Thin is in

Thin clients aren’t a fit for every organization, but they should be in more widespread use. Faster networks, the increasing prevalence of mobile users, shared work areas, and concerns about client technology being outmoded six months after purchase should put thin back on the map. Other benefits like true lockdown, instant-on, and location-independent sessions are worthwhile no matter what shape the economy is in.

The trouble with thin is that it’s a horrendously expensive technology to roll out. You have to lay in a lot of server firepower before you can take even one department from desktops to thin clients. As more users are brought on, servers have to grow rapidly.

Fortunately, memory is dirt cheap now. Having roughly 512 MB of memory per active user isn’t the show-stopper it used to be. But processors are not cheap. As you get into machines that can accommodate 32 or more CPUs — the kind of hardware you want driving a bunch of thin clients — incremental upgrades in CPU capacity can bust your budget.

Considering these conditions, a thin client network is one of the best justifications I can think of for capacity on demand. HP, IBM, and now Sun have programs that let you add processors to a running system. With the best of these plans, the processors are already in the system, but they’re initially idle. You pay for them only after you put them to use. The second-best arrangement guarantees a quick turnaround on the delivery and installation of additional processors. A well-designed modern server allows the insertion and removal of processor cards while the machine is running.

What typifies a good capacity-on-demand program? The SLA (service-level agreement) is the most important part of the contract. It will define the maximum time you’ll have to wait for a new CPU and what the vendor will do for you if the company can’t deliver a processor on time.

You should also look for plans that adjust costs both ways: Your payments go down when you turn processors off or give them back to the vendor. The vendor should work with you so that processors can be reallocated across servers as needs dictate. And finally, the SLA should spell out what happens when the vendor quits making the servers you’re running. If the box goes out of production and parts become scarce, you can’t let yourself get stuck with an unexpandable white elephant. You have to be able to upgrade to newer technology at an acceptable price.

Thin is not the answer for everything. Some applications are a horrible fit for thin clients. I wouldn’t consider doing publishing, graphics, or design work on any machine that lacks a brain of its own. It’s fair to argue that a cheap desktop PC costs less than most thin clients, and any PC or Mac made in the last 3 to 5 years can itself be a thin client. But if you look at the cost of maintaining a building full of boxes with fragile moving parts, and contrast that against running a small group of servers to feed all desks from one place, thin starts to make sense as at least part of your computing strategy. And capacity on demand fits thin computing hand in glove.