Slashdot Mirror
Unilever Ditches Global IT Linux Migration
GP writes "One to stir the open source debate. The CIO of global consumer goods giant Unilever says in this interview with silicon.com that the company has ditched plans to migrate its enterprise IT platform to Linux running on Itanium. He reckons hidden support costs and security issues have emerged over the past two years with open source and that proprietary vendors have also raised their game in response to the 'threat'."
Hmm, not to troll or whatever, but why do we care what OS does some company use? It's not like I ever work for the company, the most that change is going to affect me is the 404 page will say IIS instead of Apache. And it doesn't mean that one OS is better than the other, just that that company decided to use one. I'm not a company, I'll go with whatever suits me.
Send email from the afterlife! Write your e-will at Dead Man's Switch.
Well, the problem must be Linux.
"They that can give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety."
The snippets of text in the article imply that Linux was the part that no longer makes sense, but I suspect that switching to Itanium was also part of the reason they stopped. I can't believe that attaching massive Itanium use to any major infrastructure would increase its cost competitiveness. Sure, you could argue that Itanium in a few niche areas gave better bang for the buck than x86, but for the global IT infracstructure of a company? It can't be a good idea.
"Unilever CIO Neil Cameron, said the cost benefits of migrating en masse to an open source platform are no longer as clear cut as they were two years ago because of security and support issues."
Sounds more like he got his ass handed to him by an enterprise architecture team after attempting to push through a bad idea based on a flawed financial model.
-- lk t lv ll th vwls t f wrds. T svs lts f tm t wrt bt ts pn n th ss t rd nd mks m lk lk cmplt dpsht.
A few years ago, there's a need for a fair to middling department store chain to develop and deploy an epherimal business monitoring system. The current in place at six test stores is doing well and promises to provide detailed and instaneous headcount monitoring data to the central office which, when fully deployed and combined with sale pricing, inventory and geographical demographic data, offers an unprecedented degree of feedback to the decision makers. Consequently, the decision was made to give the project the go ahead.
In a nutshell, the current system listens to the infrared people detectors that go "bong" when people walk into the store and "bong bong" when people walk out, and feeds the data over the token ring to the store computer. But this won't do for the rest of the stores because they're using wireless networks.
The general idea thus becomes to make these systems wireless and functional out-of-the-box so that a store clerk can take it out of the packaging and situate the device near a source of power and within listening range of the people detectors. And since there was a great deal of buzz about achieving a lower TCO with Linux the company's "Linux on new installations" initiative meant they wanted to switch from Windows (used on the prototype machines) to Linux on the new devices to avoid per-site charges and network worms.
That's when things start going downhill -- not from an inherent flaw in Linux mind you, but from the fact that the original app was compiled Delphi and the compiler was in Norway with Jacques, the former IT developer, who returned to his family to work on their penguin conservation efforts (I imagine a matter of keeping the penguins fed and the polar bears fed with something else.) The current guy, a Linux enthusiast familiar with Wine, figures that instead of trying to rewrite the application from scratch it'd be quicker to wrap the Windows binary in a layer of emulation and wrap all that with a layer of Perl to interpret and route the results over the wireless network.
But the damnedest thing always seems to occur in these situations; it never takes as much time to rewrite as it does to kludge. Everything looks right after a week or so, functionwise -- these were embedded systems and therefore difficult to debug, but the development was done at a workstation that had a .wav recording of the "bong" sound that could be played into the
unit for testing. The system listens, transmits a byte over the wireless
when it gets a hit, and the central computer tabulates the data. No worries.
Except that nobody seems to be leaving the store. 0 counts for exits, average stay is 16 hours (from open to close.)
To say the guy was frantic at this point is an understatement. There were five days to go until the devices needed to be shipped to meet the deadline, and they're only half functional. To add to the problem there is now no time to rewrite, he's no good with a disassembler, and the embedded environment thwarts his further attempts at debugging.
Nevertheless he keeps at it. GCC/GLib are at stable versions, libraries are properly loaded as are the drivers -- indeed, the device isn't crashing and is able to speak with the network. He checked LKML, he stopped by #linux on EFnet, downgraded and upgraded the kernel all to no avail. His last resort was fervered e-mails to Jacques to see if he knew anything about the situation.
Fortunately, at the last minute Jacques was able to let him know what the problem was and that, in hindsight, it was both trivial and obvious, and everything ended up working out. But he swears that next time he'll start with a rewrite and leave the fancy stuff as a last option.
Try not. Do or do not, there is no try.
-- Dr. Spock, stardate 2822-3.
This isn't as much "normalization" as it is "don't take so many drugs when you're designing tables."
While I'm not often a big Linux advocate, I have to question a few assumptions here.
Is security reduced through exposure? I'd say not, so long as the design is fundamentally secure.
Windows is not. Linux, Unix and OS X are. Put the ***x OSs at the same marketshare as Windows and some holes will emerge, but the fundamentals of the OS are more secure. They will not suffer to the same extent as Windows.
Windows is (thankfully) moving in the right direction but it seems to have the same tight turning ability as a supertanker in a narrow harbour.
****
Is SCO a real risk? I'd say not. I don't believe the case they brought will survive much longer, and already they're considered a bit of a joke.
The FUD was strong though.
****
Is software patenting an issue for Linux? I had heard it was covered. Maybe through something like OpenBSD. How is it less an issue for Windows?
Err... scratch that idea. They'd never notice.
it's a blue bright blue Saturday hey hey
I worked for Unilever IT until quite recently.
There was never a serious itent to migrate to Linux. It was invoked more as a threat in 2004 to get big suppliers like HP and MS to cut prices when dealing with Unilever.
I guess it worked, and now Unilever can drop the pretense.
...seems to be a very bad idea in most cases IMHO--at least if it can be avoided. I should hope any CIO that would suggest that sort of thing would have his ass handed to him by his team.
Is it just me, or does it seem that most big, all-encomapssing IT projects are unmitigated disasters? It doesn't matter if it is Unix to Windows migration, Windows to Linux, VMS to whatever...or even the initial implementation of a big system like SAP--it is extremely difficult to pull off. Really, what "financial model" could possibly show that uprooting the entire IT infrastructure of a large corporation all at once would be favourable? Is there no risk analysis done? Hell, does common sense not even come into the picture?
There are only a few situations that I could see where a massive enterprise project like this would be justifiable--and in the case of large corporations I would say that such situations would be due to neglect and incompetence--for example they've got a bunch of elderly Win95 PCs, a VAX that you cannot get parts for anymore, etc. and if anything bad happens to any of it the results would be catastrophic. So even if a massive IT project is not a foolish idea, it was foolishness that led to the need.
The article says that Linux is still part of their plans--it is just going to be used more strategically and selectively. I don't really see where the big argument is here. I'd rather see a large number of smaller success stories than one huge successful Linux project if it means hearing about 4 more Linux-based disasters that Microsoft could use as ammunition (ignoring the fact that the failure rate of massive Windows-based projects would be at least as bad).
You mean open source doesn't solve every software problem?
I found the "religion" comment particularly amusing. I wonder how many managers have been turned off of open source because they have some employee running around screaming about source code freedom and writing stuff in emails like M$.
Two "bongs" don't make a byte.
Try not. Do or do not, there is no try.
-- Dr. Spock, stardate 2822-3.
Anybody considering moving to Itanium really can't be taken seriously.
Anyway, it sounds to me like they were using Linux mainly as a bargaining chip with Microsoft.
"I work in a *very* big company (can't say right now, in the office... but we're possibly the biggest on Earth by at least one metric)"
Microsoft?
Or maybe they have the fattest employees?
My guess is that this guy didn't get his job because he's an idiot.
uh. have you met many IT professionals and corporate executives?
The next comment I write will be ready soon, but subscribers can beat the rush and see it early!