In-Depth

Gateway Evolution: Tying the Mainframe to the New Paradigm

• 12/01/1998

By linking mainframe data to new Web-enabled client/server platforms, SNA gateway products let enterprises leverage their investments in legacy systems while achieving the strategic flexibility of newer platforms like Windows NT.

The Systems Network Architecture (SNA) was originally used to network IBM mainframes to one another and their dumb terminals. It’s since evolved into "an umbrella term for a set of specifications, a variety of products and the architecture that governs how those products are constructed," says Eric Hindin, Director in Data Communications/Network Solutions, at Yankee Group. With the advent of PCs, though, vendors created products called SNA Gateways that networked PCs to mainframes. Using them both leveraged the power of the PC and was cheaper and easier than proliferating one-dimensional dumb terminals or rewriting mainframe applications to interface with PCs. SNA gateways translated the SNA protocol into the language of PCs, so PC users could access IBM mainframe data and applications.

When client/server architecture emerged, laymen mistakenly believed IBM mainframes were on the wane. Though sales flattened, mainframes were still needed for apps that required high-end transaction processing. And even in situations where client/server systems assumed mission-critical activities like processing insurance claims, the mainframe remained a key application and data server. Why move mainframe data and apps to less powerful servers when you could leave them on the mainframe where personnel could access them using SNA gateways? The SNA gateway market steadily grew - first with products linking UNIX environments to IBM mainframes and then, as Windows NT flourished, with products linking that backend to mainframes.

Audrey Apfel, Vice President and Research Director in Networking at Gartner Group, says most companies now migrating to NT are deploying new apps on NT and linking them to the Internet and relegating legacy apps to mainframes. Mainframe apps still perform traditional mission-critical tasks, but any new era apps dealing with, say, electronic commerce, supply chain management, or intra-corporate communication via intranets are deployed on NT. Apfel calls this new NT environment the "strategic" platform and the older mainframe environment the "legacy-enabled platform."

Though companies’ development priorities have shifted to the "strategic" platform, new apps housed there can’t function without data from the "legacy-enabled" platform. If, for example, in a supply chain scenario where a auto dealership wants to order 500 new vehicles, it still must access inventory information stored on a legacy mainframe inventory control app. The dealership would not port that app and data mother lode to its NT platform - cost wouldn’t justify it, and NT wouldn’t scale to support it.

The residuum of that evolution from mainframe to client/server to electronic commerce computing paradigms is this: The two protocols that dominate networks now are SNA, linking the client/server world to the mainframe one, and TCP/IP, linking client/server platforms to the Internet. And neither will shrink any time soon. To suggest that SNA is a dying protocol because mainframe sales tapered off is like saying that toes on the human foot are obsolete because we don't use prehensile appendages to hang from trees like our simian forbears. Migration to new computing paradigms is evolutionary, and vestiges of previous paradigms usually inhere in the new platforms.

This doesn’t mean that SNA gateway vendors won’t try to leverage their hereditary strengths to augment their customer base. For instance, Microsoft has a vested interest in creating such a powerful, stable and scalable alternative NT platform that IBM mainframes devolve into mere data repositories. IBM, by contrast, has a huge stake in creating SNA gateway products that make mainframes more cost-effective, low maintenance and easier to use so users retain them for transaction-intensive processing.

Of course, these vendor motivations shape the market dynamic governing SNA gateway products. Apfel says Microsoft will proliferate its NT SNA gateway products into enterprises from its stronghold in workgroup computing. From its incumbent position in the high end, IBM must ward off that competition with SNA gateway offerings that render the mainframe platform more flexible. The end game of both vendors is the same - enabling NT users to leverage mainframe apps and data in any way they choose.

Apfel classifies SNA gateway products into four categories:

"1. Those that provide traditional host access/terminal emulation-based connectivity to the mainframe (could be Web-based or not); 2. Those on PC hardware (Microsoft’s SNA Server, IBM’s NetWare for SAA, etc.); 3. Those on UNIX and UNIX-based products (Communications Server for AIX, CNT/Apertus); and 4. Those on or in network equipment (routers, etc.). Of these, the first two categories are the growth ones."

Hindin says, with the popularity of NT, sales of UNIX SNA gateway products have peaked. SNA gateway functionality in network equipment is relatively new, so it's yet to impact the more popular PC- and browser-based gateway products.

Of the PC- and browser-based SNA gateway products, Apfel estimates IBM’s NetWare for SAA (sold by IBM and Novell) has 70 percent of the market, Microsoft’s SNA Server 15 percent, and the rest 10-15 percent. Those percentages represent how SNA gateway products prospered within the computing paradigm for which they were created. NetWare for SAA services UNIX and other early client/server environments; SNA Server grew steadily as NT became the OS of choice for client/server and e-commerce environments.

So, though IBM’s NetWare for SAA was first to market and dominates market share, Apfel says Microsoft has a couple of factors in their favor. First, it makes the NT platform to which most companies are migrating. Second, SNA Server is resident on NT and comes bundled with Microsoft BackOffice, so it will gain in power, scalability and stability as NT does and increase market share as BackOffice sales grow. As usual, Microsoft is driving sales of a less mature product by letting it ride sales of more established ones. Given Microsoft’s sales model and marketing clout, market share could balance out quickly.

The inverse is true with a second IBM SNA gateway product called Communications Server for Windows NT. Recently launched to compete with SNA Server for the NT/mainframe connectivity market, Communications Server is the upstart in its niche. But it should win quick sales in browser- based NT connectivity on the coattails of IBM’s popular e business electronic commerce platform. And Internet connectivity is where SNA gateways are headed.

That market "is evolving more and more to support the integration of Internet functionality into SNA," says Hindin. "Users, rather than running traditional terminal emulation software … will be running a browser with a Java-based [or other Internet] application. And it’s that Java application that will want to talk to the mainframe which houses the data the Java [or other] application needs to access," he continues, "so the SNA gateway will be the go-between the Java [or other] application and the mainframe."

This strategy lets corporations leverage their investments in intranets while capitalizing on free and pervasive thin clients. As Hindin explains, "if a browser is a ubiquitous piece of software running on everybody’s desktop, then ... if you provide universal access to your internal applications via the browser, you’ve saved yourself the trouble of having to install application-specific software on everybody’s PC. Instead, you just leverage the browser that is already there," he adds, "and that’s obviously got a lot of cost savings associated with it." Both SNA Server and Communications Server accomplish this by linking their Java or ActiveX or other browser-based apps to the mainframe via SNA or TCP/IP over NT-based intranets.

But, while Internet connectivity has arrived, what Apfel calls "host integration" via SNA gateways is still on the horizon. Host integration is possible once SNA gateways accrue enough functionality that they can operate as application servers. As middleware platforms for applets and apps, they will then draw competition from application servers for the host integration market. But this eventuality is not imminent. Hindin says both IBM and Microsoft products encounter performance, administration and scalability problems when forced to support numerous users. For instance, though SNA Server is rated for 10,000 users, Apfel wouldn’t recommend it for more than 2000 or so.

SNA gateway products are naturally concentrated in vertical markets where mainframes have traditionally performed intensive transaction processing like banking, insurance and financial services. But they’ve proliferated across industries into just about all Fortune 500 mainframe-based applications. For example, they’re also prevalent in airlines and hotels, she adds, where reservation systems rely on real-time, high-end transaction processing.

A typical PC SNA gateway application might be a customer service center for a utility company where reps respond to phone inquiries by using customized PC interfaces to access mainframe applications containing billing, customer, and service data. Electronic commerce, by contrast, fosters many Internet-integrated SNA gateway applications. Here, Apfel says, customers with browsers and Java or other apps might access product and price lists stored on a mainframe to perform virtual business transactions.

The next challenge for the major vendors is combining the capabilities of these two types of SNA gateway products. Those who excel will most successfully mine the market’s emerging sweet spot.

ABOUT THE AUTHOR:

To Sidebar #1

To Sidebar #2