Network Management: An Untapped Market

NetLabs president describes the drive for a self-managed network

The network management applications market is largely untapped, according to Andre Schwager, president of NetLabs Inc., Los Altos, CA, a developer of such applications. The market for network management applications will rise from $50 million in 1993 to $300 million in 1995 to $525 million in 1997, he told a recent meeting of the Software Entrepreneurs Forum/UniForum Open Systems SIG in Palo Alto, CA.

But as a five-year-old company with about 60 employees, NetLabs finds itself having to sell products that sit on top of and work with large-vendors' products such as SunNet Manager, Hewlett-Packard's OpenView, and IBM NetView 6000. "We're really out there dancing with the elephants," Schwager said. "It's very rewarding. The question is, how do you dance with these elephants?"

NetLabs' estimate is that HP and IBM are together in the "top tier" of network management, and therefore their management products are the most important platforms for small companies' applications. But others such as AT&T and Siemens/Nixdorf also bear some attention. A lot of integration and consolidation of products will occur, he predicts. "The ability to accommodate change is most important," he said. "We will attend all weddings and bring presents, but we're not going to get married. We're just going to watch all this happen. You need the ability to move quickly. A small company has the advantage of moving quickly and that's worked well for us."

Self-Managed Networks

Most network users, network managers, and corporations have the same goal in network management, Schwager said. "Everybody's fantasy, and it's been around for a long time, is that they want the network to be self-managing. They want it to heal itself. But I don't think anyone's accomplished that."

To make the network self-managing, users have to be given more information about what's going on in the network, Schwager said. "You want to be involved in managing. You are the one who has demonstrated that you can manage what's out there, historically. So we would like to develop a system that keeps the user informed."

The three components in a self-managed network, all of which interact, are company-defined policies and procedures, informing users and network specialists, and the ability to diagnose, correct, and tune the network, Schwager said.

But in approaching a self-managed network, "We're really not very far along. We have a long way to go," he added.

Increasing Complexity

One problem in approaching that goal is that the more components, and the greater variety of components that are added to the network, the fewer trained people know how to manage that increasing complexity. "They're caught in a spin today-" Schwager said, "a problem management spin." So 95 percent of a system administrator's or manager's time is spent just keeping the network up-monitoring and detecting, isolating and diagnosing problems, and taking corrective action-not in planning. Just getting information is most of the battle, Schwager said. "The biggest problem you have out there today is how do you collect meaningful information?"

Getting information would be simpler if data could be exchanged more easily over telephone lines. Therefore, telephone companies would like to get into the network management market. "Data people and phone companies have never been able to talk to each other," Schwager said. "If the telco people had understood the data problem correctly, we would not have the hub business today. The hub is a data PBX [private branch exchange-an internal phone switchboard].

"Phone companies are now saying they can manage your nets for you, because they have managed the most complex networks in the world. Whoever breaks through that barrier successfully and reliably and with a cost that's reasonable is going to make a lot of money."

Developing Building Blocks

NetLabs' approach has been to develop a nerve center for handling device-level events and a problem manager for handling network-level problems. On the way to developing those solutions, the company has had to progress through a series of management building blocks, Schwager said. First there had to be agreement with the major companies on management protocols. Then his company had to develop device agent software for distributed management applications. Now the company is producing applications for such platforms as HP OpenView and SunNet Manager. Those platforms need applications because "They're just collectors and they have very simplistic thresholds in terms of what causes an alarm," Schwager said. Next on the agenda is the development of network management middleware.

Schwager divides the driving factors in network management between the "pull" and the "push." Pulling development are the downsizing movement, rapid changes in communication technology, and connectivity through internetworking. Downsizing has led to a dramatic increase in lost revenue caused by system and network outages, he notes. Whereas IBM used to keep systems under control, the increase of diverse distributed systems has brought management difficulties with it. In 1989 the estimated losses caused by outages was $4 million, mostly in lost productivity; in 1993 losses totaled $11 million, mostly in lost revenue. "Trying to manage this net is just much more complex," Schwager said.

Pushing the market are the factors of no single company setting the net management rules, multivendor complexity, individual vendors' agendas, and customers' strategies. "It was really IBM that taught the industry how to administer and manage a system," Schwager said. "Almost everyone could price themselves to make a lot of money under the IBM umbrella. Who is the IBM of the '90s? What company is leading the way to manage this great array of communication and computer equipment? It isn't there. That creates the problem of how we move forward and train the industry."

The result is what Schwager calls platform wars. Hewlett-Packard is moving forward as the leader he said, but "You can't count IBM out. HP doesn't view anyone as a major competitor except IBM. I've been expecting a major move by IBM for some time, but it hasn't come. And there's always Sun-I'm not sure where they're going to go. Companies are using network management as a vehicle to get in and control the network and then control the system that you're running it on." Sun has reportedly lost millions to HP because of customers' preference of OpenView to SunNet Manager, he said.

DEC, Schwager said, "probably has the best net management product out there for an all-DEC network. Unfortunately, it's not open."

The Distributed Management Environment (DME) once touted by the Open Software Foundation had great promise, but in the end it delayed the development of usable technology, he said. "DME was an incredible disturbance in the industry. It cost everybody about two years," Schwager said. "It was very honorable. Everybody was waiting for this big thing in the future that would solve everybody's problems. If they could have delivered, it would have been incredible." However, the development process was tied down with too much input from too many sources. "If you had any market momentum, you had some input." One developer alone could have done the job, he said.

Eventually, network management will develop to the point where portable databases such as Oracle and Sybase are today in allowing builders of open systems the freedom to shop around for the platform and applications that suit their needs, Schwager said.