Computerworld - Most corporate IT organizations have steering committees to craft strategies for new technologies, chief technology officers to assess new products, and IT policies and procedures for developing and buying new hardware and software.
But where are the review committees for obsolete technologies? Who's looking at what's in the data center, on desktops and in briefcases to see if they still make sense? Who's checking to see if spare parts, vendor support and employees with the right skills will be available next month—or next year?
In most companies, no one is doing those things in any rigorous way, says John Parkinson, chief technologist for the Americas region at Cap Gemini Ernst & Young in Chicago. "I know of very few companies that actively manage sunsetting their IT," he says. "They think it will last forever."
It doesn't, of course. But in most cases, there's no need to rush: "No tool is really outdated if it serves the needs of end users," says Eric Goldfarb, CIO at PRG-Schultz International Inc. in Atlanta. However, IT managers who wait too long may risk being forced into expensive last-minute changes to accommodate new technology initiatives as business needs change. That IP telephony call center application won't fly if you have to replace not only the private branch exchange but also update network cabling and those nonswitched, shared-media Ethernet hubs.
Parkinson says that for each type of software and hardware installed, companies should have an estimated cost and date to replace it and an estimated cost to retain it. "You really should have this in the plan when you [buy], otherwise you won't know what ROI to expect," he says.
Of course, some technologies need closer scrutiny than others. So Computerworld asked corporate IT managers and analysts what items they would put at the top of their lists. Some of them may justify an immediate rip-and-replace strategy; others should be put on your "endangered" list. Here are five submerging technologies to watch in 2004:
1. WINDOWS 9x
Why it's sinking: Can 92 million users be wrong? Yes. Declining support, reliability problems, security issues and incompatibility with new applications should drive the remaining installed base to Windows 2000 or XP.
Credit: Red Nose Studio
But eradicating Windows 9x won't come easy: IDC in Framingham, Mass., estimates that by year's end, there will still be 17 million Windows 95 installations, 48 million Windows 98 users and 27 million machines still running Windows Me. And the majority of those are business PCs, claims IDC analyst Dan Kusnetzky. "In the long term, it will probably be less costly to upgrade [to Windows XP], just because the NT kernel is much more reliable," he says.
But what if your organization has waited? Should you go directly to XP, wait for the next generation (code-named Longhorn) or choose something else?
Don't hold your breath for Longhorn: It isn't due to arrive until 2005 at the earliest. Linux is a widely touted option, but for many the idea of replacing thousands of Windows installations, training users on a new operating system and getting it to work with existing Windows applications is a nonstarter.
Tom Pratt, information systems manager at Coastal Transportation Inc. in Seattle, says he has no plans to abandon Windows 98. The applications running on his boats won't run on anything else, and it's perfectly satisfactory for his office applications as well, he says. Can Pratt stay on Windows 98 forever? "Forever is a funny term," he says. "Let's just say 'indefinitely.' "
But John Montgomery, chief technology officer at Marine Terminals Corp. in Oakland, Calf., is making the move to Windows 2000. "Our newer applications are not going to run on [Windows 9x]," he says. "The main reason why Windows 9x is still out there is for legacy applications that it took us quite a while to get rid of."
2. CLIENT/SERVER COMPUTING
Why it's sinking: Two-tier computing with fat clients had its day, but there are now better ways to distribute data and computing power for flexibility, ease of maintenance and business continuity.
The original client/server scheme—where the application's visual presentation and business logic reside on the desktop, and data resides on a server—is an idea whose time has passed. It's being replaced by Web browser clients, n-tier systems and Web services.
Why should users replace their two-tier systems? "Flexibility would be the big reason, and some issues around business continuity and disaster recovery," Parkinson says. "Also, a lot of that software was built with second-generation client/server tools, like [Sybase's] PowerBuilder and SQL Windows, and things have moved on a lot since those days." It's becoming harder to find people with those skills and to get the object code to run well on newer technology, he adds.
Randy Heffner, an analyst at Forrester Research Inc. in Cambridge, Mass., makes a distinction between mission-critical enterprise systems and "low-affordability," or departmental, applications. "Strategic applications should never be developed in the traditional fat-client, two-tier client/server model," he says. "The business logic becomes inaccessible and hard to maintain. Instead, the right approach is a service-based design—that is, to build a business services layer in the application that can be exposed via Web services or any other mechanism."
But two-tier computing with fat clients is still the best approach for companies such as PRG-Schultz, a $500 million "recovery audit" company that analyzes clients' accounts payable records to see whether they've overpaid. "Because of the creative work our auditors do on their desktops—data mining, analysis—we keep a lot of computation there," Goldfarb says. Response time is enhanced by keeping the power local rather than relying on a network, he says.
Jim Honerkamp, CIO at Clopay Corp. in Mason, Ohio, has found a way to extend the lives of his two-tier, fat-client applications. He replaced desktop PCs with Windows thin-client terminals from Wyse Technology Inc. and moved the client code to a MetaFrame server from Citrix Systems Inc. "This allows you to put a very thin client on the desktop, and it greatly simplifies support on the desktop," he says. "The application thinks it's still running in a client/server environment."
3. IBM SNA/PROPRIETARY NETWORKS
Why it's sinking: Proprietary network protocols are so 20th century, getting shoved aside by the power and ubiquity of TCP/IP. Migration will ease support and interoperability concerns.
"All proprietary protocol stacks—such as SNA, DECnet, AppleTalk, Novell IPX/SPX—are in great decline," says Jonathan Eunice, an analyst at Illuminata Inc. "Over time, TCP/IP has taken on their jobs—a process that continues as TCP/IP products continue to improve with multigigabit switches, quality-of-service techniques and so on."
You don't have to scrap those old IBM mainframe/SNA/3270 applications right away. You can run Data Link Switching over an IP network, which encapsulates SNA traffic in an IP wrapper, and leave the application and the SNA hardware unchanged. But you'll pay a performance penalty to do that.
"There are a lot of Band-Aid solutions, but ultimately the idea is to move the application to Unix or something and make it IP native," says David Passmore, an analyst at Burton Group in Sterling, Va. "If you can't rewrite or replace the application, consider software that maps old IBM 3270 terminal to Java-based browser interface."
David Pensak, senior research fellow in advanced computation at Du Pont Co. in Wilmington, Del., says he still has printers on an AppleTalk network. "Why get rid of it? The printers are fully depreciated," he says. Pensak says he also has an AppleTalk network at home, adding, "I won't replace it until I get a new computer, and maybe not then." But he says he has dumped his old DECnet applications. "TCP/IP is so much better; it does so much more," Pensak says.
4. TAPE BACKUP
Why it's sinking: Tape is cheap, but disk technology is closing the cost gap. For day-to-day backups, disk-to-disk systems that use inexpensive ATA technology make sense.
Although magnetic tape's cost per megabyte will give it a role in keeping archival records for years to come, better technologies and techniques are eroding tape's dominance for day-to-day backup and recovery tasks. "It will be replaced by other kinds of protection, like journaling and/or replication, snapshots or point-in-time copies," says Dave Freund, an analyst at Illuminata in Nashua, N.H.
Several technologies are changing the basic approach to data backup. Disk-to-disk backup systems based on relatively inexpensive ATA storage can rapidly back up and restore entire networks. Snapshot features such as Network Appliance Inc.'s SnapMirror allow rapid imaging of a system. One indication that such software tools are becoming mainstream is Microsoft's inclusion of its Shadow Copy snapshot feature in Windows 2003 Server.
The more sophisticated tools can also briefly "quiesce," or pause, applications such as databases and flush the caches for copying without bringing the system down. "We believe that five years from now, most medium- and large-sized customers will be using snapshots on disk as the primary recovery media," says Bob Passmore, an analyst at Gartner Inc. "But that doesn't mean tape is going away in the next 12 months." Alternatives just aren't well known yet, he says.
"I wouldn't consider tape old technology," says PRG-Schultz's Goldfarb, whose company backs up 165TB of data onto a gigantic IBM 3590 tape system. He says tape will be his medium of choice for three to five more years. But Goldfarb says it will eventually be replaced with disk-to-disk or Flash Erasable Programmable read-only memory backup systems.
5. VISUAL BASIC 6
Why it's sinking: As Microsoft gradually withdraws support from Visual Basic 6 and programmers abandon it for Visual Basic .Net, those old VB 6 applications will get harder and harder to maintain.
Visual Basic 6 may be the most popular programming language, but its days are numbered. "VB 6 is the dinosaur of old. There's tons of legacy code out there, but no self-respecting developer wants to go there anymore," says Dan Mezick, president of New Technology Solutions Inc., an IT training firm in North Haven, Conn. As a result, the talent pool for maintaining VB 6 code is already shrinking. And Microsoft will phase out support for VB 6 in favor of Visual Basic .Net in the next two to four years, says Mezick.
PRG-Schultz has a number of VB 6 applications but is writing all new ones in VB .Net. Goldfarb says that in 12 months, he'll have more VB .Net applications than VB 6 applications. "Right now," he says, "I'm straddling both sides of the stream."
Mark Hall and Frank Hayes contributed to this story.
Free Insider Guide