Larry Ellison’s vision of a cheap appliance that would provide the user experience and functionality of a PC but without the cost and the high management overhead was, well, visionary. They’d be cheap to buy. The user’s data, the machine’s OS and apps would all reside on centrally managed, supported and backed up servers. If a network computer (NC) failed, it could be replaced with another identical NC straight out of the box and its user could be productive again within minutes rather than the hours involved in rebuilding a PC. In remote sites, this swapping of appliances could be done by the user rather than by an expensive technician. It all seemed too good to be true.
It was. My first experience with real, live NCs was about five years ago with a bunch of Sun JavaStations, an uncharacteristically ugly (for a Sun Microsystems product, anyway … most of their kit wouldn’t be out of place in a designer electronics store) brick with a noisy fan, a Java OS that ran (well walked) really, painfully slowly and a major lack of serious productivity apps. The JavaStation2 (aka Mr Coffee) looked like a Sun product and had had a fan-ectomy but it was still reliant on the same awful Java OS and still lacked in terms of what it could provide on the user experience and productivity front. The very worst thing, from my point of view at least, was that the JavaStation had no non-volatile memory so the OS and apps had to be downloaded across the network at each and every bootup. For machines out on the WAN this meant you either had to have a local server (an absolute no-no according to Larry’s One Big Server, One Big Database schtick) or you had to wait a very long time for the damned things to get going. It just didn’t work.
The clincher for me was that these “cheap” appliances were selling in New Zealand for more than name-brand PCs. Whilst I can appreciate that 50% to 75% of IT total cost of ownership is in the “soft” costs like ongoing management and maintenance, the whole thing just wasn’t living up to the hype. Despite being a longtime follower of the way of Sun, I had to admit defeat. The JavaStation was a pretty terrible implementation of what was (and still is) a really good idea.
You talk to anyone in Sun or SolNet nowadays and they tend to blush when you mention JavaStations. And, to be fair, In IT terms at least, it was all a long time ago. But what, exactly, have Sun done about this stuff over the intervening four or five years? Well, their architecture has changed significantly. Their next generation appliance, the SunRay, doesn’t have any processor or RAM. It’s the ultimate in thinness – everything runs back on the server. That’s good, no OS and apps to load across the network. They’ve got the price right. Entry-level SunRays are about half the price of the cheapest business-level PCs. Since acquiring StarOffice, they’ve even got some pretty damned good productivity software. Sound ideal, right?
Nope. You need a 10Mbit/s or preferably 100Mbit/s switched network between the server and the appliance. Whilst this might be okay in a campus environment or in a huge government department or SOE, in NZ’s small, often very widely distributed real life business environment, it’s neither practical nor cost-effective. For large branch offices it might be okay, but for a business like the one I support, with less than half a dozen seats in each of fifteen or so branches, it’s not practical to go putting little servers everywhere. I’ve just spent the last two years consolidating as much as possible back to the data centre -- putting little servers everywhere again would a great leap backwards and would do some serious damage to my budget.
To be fair, neither Sun nor SolNet pitch the SunRay as being suitable for everyone. If I were providing desktops to an organisation which concentrated its users into a few large locations, they’d be ideal. Unfortunately, that’s not what I’m doing right now.
Swanson is IT manager at W Stevenson & Sons. Send email to Jim Swanson. Send letters for publication to Computerworld Letters.