In-Depth

Enterprise Management – The Next Generation

• 09/01/1998

As traditional mainframe enterprises reposition themselves as internal and external service providers, managers are moving toward a centralized, holistic approach to managing global applications. By the year 2000, the focus will be on managing applications and higher-level services, with systems and networks viewed merely as components of the overall applications infrastructure. It is time that we welcomed the mainframe to the overall enterprise management fold, not as a subordinate, nor as the center of all activity, but rather as a peer, another cog in the wheel.

Mainframes will be integrated into an overall enterprise management view focused not on monitoring systems or networks, which are today’s predominant methods, but rather focused on applications. And ironically, the concept of total, insular control of networked applications hearkens back to the days when mainframes were the center of the universe, and the tools to manage them provided an all-powerful view of the enterprise.

Why this shift away from network and systems management, away from SNA management and TCP/IP management, and toward a higher-level, more conceptually monolithic (but as we will see, practically yet radically distributed) applications management approach?

Simply put, while the underlying systems and networking protocols may be drastically diverse, the most important, relevant and constant aspect of an enterprise network is, and will continue to be, the applications that run on them. With enterprise networks branching out in so many kinds of unexpected ways, it will no longer makes sense from a management perspective to debate the fine points of SNA vs. TCP/IP and SNMP. The application’s the thing.

Where management is concerned, we are moving up the stack. By the Year 2000 and beyond, the focus will be on managing applications and higher-level services, with systems and networks viewed merely as components of the overall applications infrastructure. Protocols - be they SNA or IP - and platforms - be they UNIX, MVS or Windows - are just part of the equation.

Managers, under pressure from the business-oriented folks, will want to know what is causing an application to perform poorly, or if an expensive business service has a tendency to become unavailable to users. Until recently, managers just kept an eye on discrete factors like network performance or system performance. But now there are other variables that must be considered as well.

Why focus on managing applications? Because there are so many variables involved with enterprise networks these days – clients, servers, routers, wide-area networks, switches and so on. The network is extremely diverse and point management solutions may provide stellar views of one or several of those components. But that view will not necessarily give the manager a good idea of what is causing Joe User in the Topeka office’s order entry applet – which forwards data to the mainframe – to be crashing so often.

It wasn’t always like this, of course. But the enterprise computing landscape has changed drastically in the last 10 to 15 years. The distinct concepts of network management and systems management, which made so much sense for so many years, are slowly becoming outmoded. Even a rudimentary combination of the two, which was a hot concept in the industry beginning in 1994, will not be enough in the 21^st century enterprise network.

Perhaps a look back in time will give us an idea of where we have come from, where we are now and where we have to go to manage the enterprise effectively. In a strange way, we’re almost looking to go back where we started, but in a much more powerful and versatile form – an overall view of everything that goes on in the network.

Once upon a time, in a corporation far away, there were no Java applets. There were no Ethernet switches or branch office routers or Web servers or client/server applications or virtual private networks either. There were hardly any PCs, and whatever esoteric things those PCs may have been up to, it didn’t really matter because well, they were running word processing programs or spreadsheets or solitaire. And from the perspective of the real IT people, the people in the so-called glass house, those rogue machines weren’t doing anything that really impacted the operation of the business anyway.

No, way back when, in the hoary days of 1986 or thereabouts, most of the important stuff, the transaction processing, the data collection, the payroll, was being done on the mainframe. And IBM had made the mainframe a very insular and efficient platform. You had the mainframe and the front-end processor and the controller and the terminals and the SNA protocol.

And while it was all quite expensive, and sometimes very complex, and IBM sometimes seemed clueless about the true needs of its customers, it was all fairly elegant, secure and manageable. And the PCs – well, the department heads that insisted on buying those could keep an eye on them.

Fast-forward to the client/server revolution of the early 1990s. The focus shifted away from the stodgy old totalitarian mainframe to the democratic notion of "fat" clients, distributed computing, file sharing, collaboration. Suddenly, all of these LANs had to be managed, and IBM, Novell and others stepped forward with wildly popular PC LAN management products. UNIX had become the predominant midrange server platform and TCP/IP was increasingly being used on the LAN.

Network management platforms from Hewlett-Packard, Sun Microsystems and IBM kept track of servers and routers that had SNMP agents, which would alert managers to a failure or problem on the LAN. Meanwhile, although the PCs were finding ways to connect to the mainframes through terminal emulation, the SNA network was still managed through an entirely different glass house system, the same as it had always been.

Jump to today and beyond. The mainframe is becoming just another server, and TCP/IP is not only the predominant LAN protocol, but also is running on a large number of mainframes, WAN connections and the Internet. Most importantly, applications and data are being delivered in a wide variety of ways, over LANs, through Web-to-host interfaces, over the Internet and Frame Relay networks, to employees, and increasingly, customers, suppliers and business partners and suppliers.

With the systems and networks becoming wildly heterogeneous and more difficult to control, the applications are emerging as the key element of the enterprise environment. All users care about is how their applications and business services perform. Hence the need for an applications-based view of the enterprise, and the need to integrate SNA management information with IP information, the need to integrate mainframe information, database information, SNMP information, WAN information and so on into a holistic application management view.

"The world’s gone distributed in a big way," says Jonathan Eunice, analyst and IT advisor at Illuminata, a research and consulting firm in Nashua, N.H. Eunice sees complex management "frameworks," as exemplified by Computer Associates’ Unicenter-TNG and Tivoli Systems’ Global Enterprise Manager, as the types of platforms people are beginning to use to create application-specific management systems.

"Take Tivoli, for example," says Eunice. "They are pulling together UNIX and mainframe management. They’re taking an application-centered view. You’re not thinking from a mainframe side, a UNIX side, an NT side, but from a line of business, application side. It’ll no longer be this platform against that platform. It’s not a system focus but an application focus."

Tom Nolle, president of CIMI Corp., consultants in Voorhees, N.J., takes an even more extreme view of the enterprise management infrastructure of tomorrow. "Over the next 10 years, enterprise network management is going to disappear," says Nolle. "The LAN elements of the enterprise network will be high-capacity and switched to the point where they will be self-managing. The WAN management component will be outsourced to carriers. What you will be left with is application management."

Network statistics from the WAN, posits Nolle, will be integrated into the application management infrastructure through "proxy views" provided by the carriers.

Even though there has been a movement to integrate systems management with SNMP-based network management platforms, such as HP OpenView, Nolle says such integration is inadequate. "SNMP as a general management strategy is wildly overestimated. No one really recognizes the problems at a high level."

Eunice sees this trend as going hand-in-hand with the move by many companies to implement enterprise resource planning tools from vendors, such as SAP, PeopleSoft and Oracle: "This is all about strategic management of resources. We’re moving beyond managing PCs, routers and hardware."

Eunice and Nolle agree that the mainframe will be integrated into this strategic management strategy and is likely to serve a key, if not terribly exciting, role. "The mainframes will be the big data repositories that store all of this information," says Eunice, "They will be the principal enterprise management database."

And Nolle adds that for the foreseeable future, the core applications in the enterprise will remain centered on the mainframe. But even so, don’t look to the mainframe as the central management platform – just as a component.

The notion of one central console that views all of the activity in the enterprise is the Holy Grail of management, and there are products that have evolved in that direction. There may be super-administrators that do have a full view of the enterprise, but they will not necessarily have to do it from the glass house. Distribution of management function and responsibility will be a key feature of the new infrastructure.

"There won’t be one single view," says Colin Mahoney, an analyst with the Yankee Group’s Internet Computer Strategy group in Boston. Instead, Mahoney maintains that managers of different divisions, user groups or departments will look at their own customized views of the enterprise, tailored to give them an idea of what is going with their own corner of the enterprise world. Even non-employees – or, in the case of service providers, customers – will be privy to look at enterprise resources and application availability. As supplier networks and extranets are implemented, managers will push for ways to access resources and applications.

The SNA vs. TCP/IP issue will likely die down, as IP is implemented on more mainframes and as more mainframe applications are delivered via Telnet and Web browsers. Mainframe management applications will likely stay in place, and a view of remaining SNA management data and applications is likely to be integrated into a separate holistic application-centric view of the enterprise.

To make matters easier for mainframe managers, steps are being taken to integrate a more universal view of their applications from their management platform as well. In fact, if users choose to go in this direction, mainframes will be able to monitor much of the enterprise network activity, both SNA and IP.

A step in that direction is represented in the new release 1.3 of Cisco System’s CiscoWorks Blue Internetwork Status Monitor (ISM) for System/390. ISM allows mainframes to monitor and control Cisco routers. Cisco ambitiously states that ISM reduces or can in some cases eliminate the need for SNMP management consoles to manage routers by providing management of SNA and IP devices from one screen. ISM integrates with some of the framework architectures as well.

If anything, ISM is an example of how the boundaries between network, systems and applications management are blurring, and how the boundaries between SNA and IP environments are blurring. In fact, many see the problem working out in a opposite way from Cisco’s solution: with SNMP platforms gaining the ability to monitor, or at least view, mainframe networks, not vice versa.

Vendors will soon be introducing ways to give non-SNA managers a way to monitor the availability of mainframe applications by collecting errors, alerts, events and alarm information from SNA’s built-in monitoring systems. This can be accomplish by format that data in a normalized database that allows an association between SNA network outages and availability of applications and databases.

Again, this trend represents a step in a constructive direction: an agnostic, holistic approach to managing and controlling enterprise applications, the lifeblood of a company. This goal is shared by many managers in many companies. One such manager is Bob Rubenstein, director of Global Network Services at LSI Logic in Silicon Valley.

"Over the years we’ve introduced fantastic complexity into the management problem," says Rubenstein, "You have to find a way to correlate all of these different events, such as the state of disk array, reliability, usage rates and so on. The old management model isn’t valid anymore. The days of network groups managing the network and systems groups managing the systems and Oracle database analysts and SAP experts managing their own systems are over."

"You are managing a global, distributed system," he continues, "You have to have a service view of the state of the system and be able to view all of the components required to deliver the applications to the user. This includes, but is not limited to, the LAN the server is on, the LAN the user is on, the user’s PC, the WAN connection, the operating system on the server, the database, the application on the database and so on. All of these can provide a management system with more information than any human being can understand. You have to have a holistic application to manage global applications, something to correlate results into a composite picture that tells you which events are important. There is a lot of process development, workflow involved. That’s the hardest part."

Rubenstein says a data warehouse populated with various management objects is a good way to identify and prioritize a management system, and establish the correlations needed.

Netscape President Jim Barksdale has recently been referring to "enterprise service providers (ESPs)" as he tries to explain how corporate IS will increasingly act as an accountable service provider to its "customers" around the world. The key criteria for measuring their effectiveness will be reliability, uptime, performance and responsiveness.

Rubenstein agrees with Barksdale’s assessment. In fact, he sees LSI Logic’s IT department very much as a developing, service-oriented ESP. "You must focus on the perception of the end user of system responsiveness," says Rubenstein.

And the applications that the ESP provides, while they may not run on mainframes, are mainframe-like in their service requirements and sheer scope. "The model is shifting and we’re moving back to virtual monolithic computing platforms," says Rubenstein. "They can be managed as a single entity. We’ve started to turn servers in the client/server model into mainframes. For example, to run SAP on UNIX it costs $5 to $10 million in server processors and you may need to connect to an EMC database holding five to 10 terabytes of data … It’s beginning to sound like a mainframe, isn’t it? And mainframes are being repositioned as servers."

While Rubenstein agrees with the analysts, revealing that LSI Logic is looking to shift away from a systems and network management system to an application management system, he sees other options beyond CA and Tivoli.

"CA and Tivoli are attempting to, and claiming to, deliver an end-to-end solution," says Rubenstein, "But we prefer to use best-of-breed point products. We’re looking at products from up-and-coming vendors like BMC Software, Micromuse and Concord, in which we’ll use a best-of-breed solution for each component of the management problem so that we can implement global, end-to-end service-level management."

In effect, LSI Logic is building its own framework, slowly and one piece at a time. "You have to build this incrementally," he says.

Rubenstein’s first task is establishing service levels for LSI’s globally distributed SAP application. "We’re looking to establish service-level agreements with user groups around the world," he says. "We’re shifting out of being network people and becoming service providers."

Rubenstein’s work illustrates that, while mainframes may not be the center of the enterprise universe as they once were, the reliability, security and guaranteed service-level ethics of mainframe-based computing are translating into a dynamic new distributing computing model.

Management is not just reactive maintenance anymore. It’s serious business.

ABOUT THE AUTHOR: