May 24, 2004

Where should you use open source? A scenario-based analysis

Author: Robin 'Roblimo' Miller

ORLANDO, Fla. -- This was the title of the last session I attended at GigaWorld 2004 last week. It was led by Forrester analysts John Rymer and Uttam Narsu and was run as a roundtable discussion with heavy audience participation. A majority of the 30 people in the room said they had tried at least one open source program at some point, but it was apparent that this was not necessarily a gathering of open source boosters.To start, note that GigaWorld 2004 had more than 700 IT managers, CxOs, and vendor employees in attendance, and only 30 of them showed up for this open source discussion -- including a rep from Microsoft's PR firm, Waggener-Edstrom. Obviously, most people at this conference decided other sessions being held at the same time were more important to them.

Use case: Health clubs go from Unix to Linux

The first attendee scenario brought up was a chain of health clubs that started moving their servers from Unix to Linux about two years ago. Their first experience with Linux was with their Web servers, a migration that took about three months. Now they're 90 percent Linux, running mostly on IBM blades. They are still in the process of working with PeopleSoft on moving that software package to Linux from Solaris and hope to complete this process and shut off their remaining 20 Solaris servers within the next few months.

Training was not a big issue. Solaris/AIX admins self-trained on Linux by installing Red Hat on their desktops. They played with it and liked it. While some resisted the change at first, they were told they were getting new hardware and that it was going to run Linux, so get used to it. And most of them did, although a few couldn't handle the change and left.

Support turned out to be even less of an issue than training. With IBM hardware running Oracle on Red Hat Linux, they found they could open three tickets -- one with each vendor -- and take a "whoever gets there first wins" attitude toward finding a solution to a problem. "We get more and better support now" was the key statement.

Developer training was the hardest part of the move. Much of the resistance was along the lines of, "But we've always done it this way," combined with worries about moving from a variety of environments, each one a "pet" of the developer who used it, to J2EE. The solution was to bring in an outside development firm familiar with J2EE, not to do development work but to mentor the company's in-house developers during the transition phase.

The company's most critical day-to-day computer function is verifying memberships (members are issued cards) when people show up at the clubs to use the facilities. This must happen so quickly that customers don't feel they're being held up, at any time of the day or night, at any one of thousands of locations. Writing the (Java) software to accomplish this through the network was expected to be a sweat-producing job, but the IT manager describing the process said they were surprised how easy it was once everyone was properly trained; that the job took "20 percent as long as we expected."

Using Linux on desktop 'getting easier'

Asked whether the company was considering Linux on the desktop, the IT manager said, "It's getting closer, but it√Ęs not something I want to put out yet." He said that all his admins were now using Linux on their desktops by choice, but that Exchange was still integral to the company's operation, and was likely to keep operating 200 servers there for a while, but that they -- and Windows desktops -- might go away before too much longer, because using Linux on the desktop "is getting easier. In a year or two it won't make much difference for users" which OS they're using.

At this point, the company claims it is doing 55 percent more computing work with the same budget it had before the Linux transition started.

This brought up an interesting question: How much of the savings could be directly attributed to Linux, Java, or other open source software, and how much to the transition from expensive, Unix-specific hardware to generic Intel-based units? And couldn't they have saved just as much (or almost as much) by using "relatively low-cost Windows" instead of Linux?

The answer was that they weren't sure exactly how much of their savings were directly attributable to which parts of their migration, but that it would be impossibly expensive to use virtual machines and a whole stack of blades with anything but Linux, so it's an integral part of the new system, and no, Windows wouldn't do.

Other attendees chime in

One said, "We're behind the curve on Linux, but using about five dozen open source applications." He then pointed out that porting Unix apps to Windows "is non-trivial, but porting them to Linux takes hardly any work. Some just run, and don't even need to be ported. It's a lot more expensive to move from Unix to Windows than from Unix to Linux."

He was asked about Microsoft's Unix-to-Windows migration tools (some of which are open source). He said his experience with them had been poor; that they didn't make porting significantly easier.

The next topic that came up was Java vs. .Net. A voice from the back of the room said, "The big variable is that .Net is the Microsoft big thing today. What will it be tomorrow? I don't trust them to stick with one thing."

Then came an almost inevitable discussion of security-through-obscurity vs. open source-based security, and the Waggener-Edstrom, guy jumped in. He claimed their were no "zero-day" exploits on Windows; that all exploited vulnerabilities were known and customers hadn't been keeping up to date. I questioned this statement. Session leader John Rymer told me and the Wagg-Edd rep to keep quiet and let the actual users have their say, which we did and they did, but it was still the same discussion you've seen 1,000 times on NewsForge and elsewhere; one claimed Windows is attacked more because it's more popular than Linux, another claimed that if popularity was the criterion, Apache would get attacked more than IIS, another said Microsoft gets attacked because Microsoft is seen as a "bad guy" in the industry. And so on. You've heard it all before.

Open source Java development tools?

This turned into a sub-discussion of its own. A development manager said, "We're using Borland Java tools now and struggling with whether we should move to Eclipse."

He's faced with a short-term project in one of his company's business units where the budget won't stretch to cover additional Borland licenses, but he's scared that the long-term maintenance and training cost of having multiple development tools will exceed the short-term savings he'd get by using Eclipse on a single project. So what should he do? Is this the trigger he needs to move everything to Eclipse? Has anyone used it on any large-scale projects? If so, how did it work?

(No one suggested a move to .Net, not even the Wagg-Ed rep. This may be a crowd that uses Windows on the desktop and on at least some servers -- at least right now -- but they are Java, not .Net, users.)

The consensus was that the best course, regardless of license cost, was to choose one IDE and stick with it. Some people in the room said they were doing well with WSAD (IBM's WebSphere Studio Application Developer), while others had been using Borland so long, and they and their people were so used to using it, that they were scared of a productivity falloff if they switched to anything else.

A point mentioned was that Eclipse and its plugins were more advanced in many ways than any of its proprietary competitors. Heads nodded. Apache Geronimo was also mentioned as something that needs to be thought about, if not now than not terribly far in the future.

Rough consensus: If you're happy with what you have and it's meeting your current Java development needs, stick with it. But if you feel it's time for a change, (open source) Eclipse is probably the best bet.

A session that meant different things to different people

Imagine a face-to-face email discussion list. That's what this felt like. Some people had comments to offer, but most sat quietly most of the time. Many -- especially those who didn't put up their hands when the "Who's tried open source?" question was asked at the beginning of the hour -- took notes.

There was even a bit of flaming, same as you'd see on an email list, but it was polite -- and got shut down quickly by the session leaders.

This was the first open source "roundtable" discussion hosted by Giga/Forrester. It was a learning experience for them, too. Hopefully they'll have similar sessions at future conferences. Hearing about open source migration experiences directly from people who are running IT departments helps everyone -- especially their peers -- learn more about what they can and can't expect when they start using open source software.

One note about this session -- and possibly the most important thing I took away from it myself -- was that neither license subtleties nor "free as in beer" pricing were considered nearly as important when selecting software as usability, stability, security, and support.

In the pragmatic world of enterprise IT management, price is only one factor among many. You can say, "But it's free" all day long, but if that free program isn't as good as (or better than) proprietary competition in other ways, it isn't going to become popular with people running sizable corporate IT operations.


  • Migration
Click Here!