In-Depth

Sticking It Out on MVS

In spite of IBM’s best efforts, a lot of customers continue to run mission-critical applications on older MVS mainframes

• By Stephen Swoyer
• 07/26/2005

Do you have an S/390 mainframe or zSeries system in your infrastructure? It's no wonder: they're are among the most accommodating systems on the market, hosting a number of application and operating-system workloads and functioning—as lynchpins or periphery contributors—in a variety of different application architectures.

For users of older MVS (S/370) mainframes, the systems that preceded the System 390 revamping and rebranding of the early 1990’s, IBM wonderswhat you're waiting for. zSeries mainframes are available at a price point and capacity to suit almost any customer, argues Jim Porrell, an IBM distinguished engineer and chief architect for zSeries software, and there aren’t many cases in which it doesn’t make sense to step up to zSeries.

“If you’d asked me this two years ago, I would have said maybe [an upgrade would be worthwhile], but z/OS 1.4 really brought a tremendous number of features and enhancements, and so a large number of OS/390 [shops] have come on board, although we still have a few MVS [customers],” he indicates. “Two years ago, it was a big problem, but it’s much less of a problem now, thanks to the economics of zSeries, the new workload license charges—these really have reinvigorated this platform.”

Nevertheless, a number of customers continue to run mission-critical applications and systems (or host mission-critical data) on older MVS mainframes. In spite of the best efforts of IBMers such as Porrell to get them to upgrade to zSeries, many of these customers are staying put.

“I don’t have any first- or second-hand knowledge of this directly,” stresses John McKenny, director of product and program management for BMC Software Corp.’s mainframe business unit, who says BMC doesn’t market any products for S/370 or earlier mainframes. “But take, for example, one of your smaller businesses that has the older hardware that has either gone a different direction from an IT perspective with regard to their architectural strategy, or maybe they just find it prohibitively expensive to move. Their conundrum is that they’re running on older, unsupported hardware or software, but they need to find a way to integrate this with” their Unix, Linux, and other distributed systems.

Of course, the MVS mainframes of the 1970’s and 1980’s weren’t quite as accommodating as today’s zSeries systems, McKenny notes. Instead, they’re the products of an age when the mainframe (or minicomputer, in the case of VMS) was king and distributed platforms were relegated to peripheral roles at best. This has ramifications for customers who wish to involve older MVS mainframes in the application architectures of today.

As a result, says Porrell, even in the late 1980’s, SNA was still the dominant mainframe architecture, and TCP/IP support was far from ready for prime time. “If they’re on MVS, I hate to imagine that they’re on TCP/IP, because the state-of-the-art [for TCP/IP] in MVS was horrendous. That means they’re all SNA, so they’re paying quite a premium for the [SNA] communication controllers and all that. Those boxes eat up a lot of floor space.”

This is important, Porrell notes, because the most common way to expose older MVS applications is by means of 3270-to-XML translation.

Although it’s possible many customers have modernized or stabilized their MVS TCP/IP stacks (via assistance from IBM Global Services or other services providers), most shops will probably need TCP/IP-to-SNA integration middleware to facilitate access to 3270 data streams.

IBM and a number of other vendors provide 3270-to-XML translation solutions, so customers have no shortage of offerings from which to choose. IBM, for example, markets its Host Access Transformation Server (HATS), now available in version 6.0. HATS provides a drag-and-drop application-integration environment, without directly changing source code and without configuring applications on a screen-by-screen basis.

Big Blue positions HATS as a tool for enterprise environments that aren’t rich in IT talent—or that can’t spare valuable IT resources—but which want to expose traditional green screen host applications in new-fangled interfaces. HATS applies presentation rules to green screens as they are encountered in a 3270 data stream, and can also skip through some green screens programmatically, bypassing extraneous screens. Because HATS redisplays 3270 data streams in a Web browser interface, organizations can customize the presentation logic of host data by inserting corporate banners or other information into a flow.

Microsoft, too, markets its Host Integration Server (HIS) 2004, formerly SNA Server, which facilitates access to SNA-compatible applications and services. Developers can also use Microsoft’s BizTalk Server, in conjunction with third-party solutions such as OnWeb from NetManage or Extra! from the former Attachmate (which last month merged with host integration specialist WRQ Inc.), to translate from the 3270 data streams to XML.

Attachmate, NetManage, and WRQ have been talking up aggressive legacy-mainframe, application-integration scenarios for some time now, trumpeting capabilities that go beyond conventional Web-to-host screenscraping to include bona-fide involvement in next-generation service-oriented architectures (SOA). “Customers can extract [a mainframe application] into a series of reusable services that can then be exploited by a developer who’s building a front-end application, usually in a variety of different architectures, usually in Microsoft or [J2EE],” explained Eric Varness, director of product management with host integration specialist WRQ, in an interview last year.

The irony is that host applications are probably better suited for exposure as part of an SOA than many applications based on more modern 4GL object-oriented languages, said Phil Murphy, a principal analyst with consultancy Forrester Research, in an interview last year. “When folks wrote screen-based transactions many months ago, they wrote it at a business function viewpoint: I add a customer, I add an order for that customer, I check backlogs for that customer, etc. So in many respects, those CICS screens of 15 years ago are better suited to service orientation than a lot of the newer, distributed code that’s been written over the last several years, because of their affinity with a business function,” he argued, adding: “What did the object-oriented guys do? They took those screens and they broke them down into a thousand different objects.”

The big takeaway, Porrell says, is that if a customer wants to expose MVS applications running on older S/370 mainframes, there’s no shortage of vendor solutions with which to do so.

“There’s a wealth of vendors that have built servers, either at the application layer, at the presentation layer, etc., to try to provide a glue between the older mainframes and the more modern desktops,” he says. “Again, if you look at the aggregation of costs [of paying for these solutions, along with MVS maintenance and support] I think there could be a very interesting debate about whether it wouldn’t be better for these [customers] to move [to zSeries].”