Client/Server Systems Development

Client/server design skills and expertise are scarce. Organizations that do not have them in-house or on contract may soon be left behind.

by / May 31, 1996
GartnerGroup envisions a revolution occurring in client/server, where less emphasis will be placed on the technology itself. As Darlene Brown puts it, "The next generation of client/server development has less to do with the technology than it does with the design issues." Brown is Gartner's vice president and research area director for the Application Development Methods and Management Area.

Over the next five years, says Gartner, the walls currently surrounding an organization will all but disappear. Through the Internet, everyone will be connecting with everyone else. Application logic will no longer reside on a single box, but will be distributed on multi-tiered servers. Some parts will run on NT, others on Alphas, and some portions on legacy hardware.

Gartner's studies indicate 22 percent of their clients are currently using multi-tiered applications. This will increase to 55 percent in 1997, and up to 70 percent by the year 2001. Brown estimates 40 percent of existing legacy code is reusable in multi-tiered applications. "We're predicting 60 percent of all client/server applications existing today will either be scrapped or rewritten by 2001," said Brown.

The corollary to the 22-55-70 estimate is a critical shortage of client/server design knowledge, support, and experience. This lack
of expertise -- whether in state governments or industry -- will
most likely be remedied through outsourcing.

Changes are looming for the desktop clients. At any given time, most organizations have a blend
of client types. The "fat client" has most of its program logic and data stored on its own hard drive. The "thin client" obtains most of its operating software from the server. GartnerGroup has coined the term "ultralite" for Web clients without program logic or data. Ultralite clients don't have to be anemic. They can be high-powered, graphical machines that operate from a remote server.

The ultralite client is appealing because of its ease of administration. Applications and data are easily maintained in a central location. However, this methodology consumes large amounts of bandwidth. LAN-attached users will have to move up to faster topologies and switching hardware. Remote offices with ultralite clients will no longer be able to operate across 56Kb lines.

The movement of images, graphical data, and program code is already slow on T1 lines. Adequate performance requires even faster transmission mediums, and their accompanying higher costs.

At the opposite end, the fat client model presents different problems. Data replication and synchronization between thousands of workstations is a fundamental issue. This problem is further aggravated by replicating to dial-in mobile users.

The thin client provides middle ground for many organizations now. Applications run on the client, and the data resides on a remote server. Both the thin and fat clients require individual software licenses. Neither can take advantage of software metering from an application server.

Client/server development skills are already scarce. Preparing for the year 2000 date rollover problem will make these scarce resources even harder to find. By GartnerGroup estimates, organizations must have their plans in place by mid-1997. Those who delay beyond that date run the risk of coming up short handed when the millennium rush hits.

As technology sophistication increases linearly, the design skill requirements increase exponentially. Users are seduced by easy-to-use development tools. Today, anybody with a credit card can purchase visual development tools and create applications. When this is done outside the scope of the centralized IS department, these applications become support nightmares. Brown observed, "A lot of those people doing those applications don't have any idea of what they're doing. They need the design skills."

GartnerGroup classifies organizations into the categories of buy, build, and outsource. They identify:

Type-A organizations as risk takers. They are willing to break new ground, and fall back on a contingency plan if it fails. Type-As are application buyers if it gets them finished sooner. These organizations will outsource when it buys them time.
Type-B organizations will build an application when a purchasing option is not available. This is done on an application by application basis, rather than across the board. The pros and cons are weighed, and a decision is made between buying and building. Seventy percent of all organizations are Type-Bs.
Ten percent of organizations fit the Type-C, always-build category. Both Type-C and Type-B organizations will outsource when the required expertise is not available in-house.