opinion: The End of Mainframe-Only Software
The past two years, we at Longpela Expertise have been creating a database of past and present mainframe-related software: www.lookupmainframesoftware.com. While we've been researching these products, it's been interesting to see how the focus of mainframe-related software has changed over time. One fact in particular stands out: the days of new mainframe-only software are over.
When I talk about mainframe-only software, I'm talking about software developed on the mainframe to run exclusively on the mainframe. In the past, this was pretty much all mainframe-related software. Mainframes were separate islands that didn't really talk to anything else. So mainframe software had to run on a mainframe.
Today, most of this software is getting old, and their days of large-scale development are a memory. New changes only add small functional improvements, keep it supported and relevant, and minimise reasons to remove it. Take DFSMShsm for example. This is a great product for backing-up and archiving datasets. But it hasn't really changed since MVS introduced SMS and Management Classes in 1988. Today, it still cannot backup or archive individual z/OS UNIX files or directories, and is missing an easy-to-use interface. I still use the traditional TSO line commands.
It certainly isn't one-way traffic. For example, IBM still develops great new features for strategic products such as IMS, RACF, CICS and DB2. And many mainframe-only products are getting a web-interface makeover. But work adding major new features to mainframe-only products has been slowing down for years. And some of these new features add communications with other platforms: removing the mainframe-only tag.
But a more interesting fact is that there have been few new mainframe-only products released in the past 10-15 years. Try searching through past product announcements. You'll certainly find IBMs z/OS Management Facility. You'll also find some CICS management tools, and a couple of offerings from smaller vendors such as Brazil based 4bears. But not much more. Most announcements are a mixture of new releases for existing products, renamed products, or products acquired from other vendors. This doesn't mean that there is no new mainframe software. What it does mean is that new mainframe software is no longer mainframe-only.
And to think that this all started in the 1990s.
The end of mainframe-only software began with GUI screens introduced with the Apple Mac in 1984 and Microsoft Windows in 1992. Mainframe 3270 screens couldn't compete, and vendors quickly developed ways to access mainframe applications from these GUI-capable PCs. One of the simplest was to create a GUI application from 3270 screens by screen-scraping. In other words, using software as a middle man: converting incoming 3270 screens to GUI screens, and outgoing mouse-clicks to 3270 keystrokes. Early software used PC-based software to present the GUI; later software web browsers.
Today a browser-based GUI is standard for mainframe-related software. Many vendors continue to offer tools for web-enabling 3270 applications. These include IBM Rational Host Access Transformation Service and Rocket Shadow z/Presentation.
Access Mainframe Resources
In the 1980s, a criticism of mainframes was that they didn't talk to other platforms. Take file transfer. One of the few products at the time to transfer files to and from the mainframe was Barr/RJE. This software emulated a JES RJE device, connecting with a special purpose PC card. Yes, you could communicate with mainframes, but it was hard.
This changed with SNA LU 6.2 in 1984 and TCP/IP in 1989. Non-mainframe platforms now could more easily communicate with mainframes. In particular, they could run mainframe applications, and access mainframe data. Software to make this easier soon followed: middleware.
Websphere MQ is the first middleware product to come to mind, but there are a lot of them. IBM provides the basics with DB2 Connect, IMS Connect and CICS Transaction Gateway. Other vendors duplicate or enhance these basics. Current software includes CONNX, Rocket Shadow z/Connect, and Oracle Tuxedo. Middleware software products today do many things: from connectors (such as JDBC/ODBC/DRDA) to data translation, data mapping, and encryption.
This no doubt has made IBM very happy, as mainframe isolation was a big reason for moving away from mainframes. But opening up the mainframe has been a two-edged sword. More mainframe-related software can now be developed and run on non-mainframe platforms, communicating with the mainframe as needed. So it's tempting to go the cheap option and maximise development on UNIX and Windows; scaling down mainframe development to the bare minimum. And many software vendors are doing exactly that.
SAPs ERP products are an excellent example. The older SAP R/2 could run as a CICS/VSAM application. The replacement SAP R/3 is a UNIX based application supporting DB2 on z/OS as the back-end database.
MVS/ESA 4.3 in 1994 introduced a new concept to MVS: UNIX. z/OS UNIX is a POSIX standard environment with all the expected C and Java features. Now UNIX applications can be ported to z/OS with only a fraction of the development effort. But not only that, enhancements and fixes for this software can also be developed away from the mainframe, and simply ported in. Minimal mainframe development required.
From the late 1990s, vendors have grabbed this opportunity to add z/OS software to their portfolio. Websphere Message Broker is a classic example. IBM itself has enjoyed this ability, providing free ported tools and languages including PHP and Perl. You can even port Apache Tomcat to run as a free J2EE platform on z/OS.
IBMs Websphere Application Server takes this to another level. Now Java J2EE applications on Windows and UNIX can be ported to z/OS. Products enjoying this functionality include banking software Oracle FLEXCUBE and insurance application FINEOS Claims.
UNIX and Microsoft Windows users will be aware of some amazing software for their platforms. For example, Xenos Enterprise Server to process output, Metron Athene for capacity management and planning, and ASG-Entact ID for centralised security administration. And this software supports many different computing platforms, providing a central point for management.
In recent years, some of these products have expanded on the middleware concept, and developed mainframe clients. All the products mentioned are examples. These clients can be simple data pipelines, or more sophisticated units performing serious mainframe processing.
This is a major change of tide. In the past, these types of applications would run on the mainframe. Now they can run on UNIX and Windows, taking what they need from mainframe clients. There is a huge range of this software. Take for example, IBMs strategic Tivoli Monitoring framework. This is a Windows/UNIX based solution with clients monitoring everything from IMS on z/OS to Microsoft Exchange on Windows.
Today, most new mainframe-related software products were not developed on mainframes. In some cases they don't even run on mainframes. Rather, they were developed on Windows and UNIX systems, and either ported to mainframes, or run on other platforms using middleware or mainframe-resident clients.
There are still new mainframe software products in development. But most are only parts of a multi-platform solution. New mainframe-only software is rare, and getting rarer.
And this makes economic sense. Existing non-mainframe software can be expanded for a fraction of the mainframe development costs. Their development is usually cheaper, and there is a larger pool of available skilled developers.
And the reasons for mainframe-only software are fading. Web front-ends are the norm, and mainframes systems now must communicate with other non-mainframe systems. Mainframes are no longer the islands they used to be.