It's Alive!

Network management almost always begins with SNMP (Simple Network Management Protocol). Gillen, for example, uses some of the latest management software tools from companies like Micromuse, Aprisma, and RiverSoft. "These are all SNMP-based," she says.
SNMP is software that, since its humble beginnings in the late 1980s (see sidebar), has made it relatively cheap and easy to collect data about your network. The problem then became making all that data easy to see and use.
"Your typical router can produce tens of thousands of events in an hour," explains Gillen. "SNMP makes it easy to collect all this data — the challenge is in understanding how to filter, correlate, and present it."

Tackling this problem was the beginning of the classical age of network management platforms. Hewlett Packard's OpenView product was one of the first. OpenView gave network managers a GUI (Graphical User Interface) and some correlation capabilities.
Before long there were others, and the market is today still dominated by the big three in network management: HP with OpenView, IBM with Tivoli, and Computer Associates with Unicenter.

Stretching the Definition

The problem with the classical definition was that no one could ever quite stay within its narrow confines. A broader definition quickly began to take hold: Network management is about managing all the stuff that sits underneath applications that run in a distributed, networked, computing environment.

And this, as Jeff Case, founder and CTO of SNMP Research in Knoxville, TN, explains, introduced more complications. "The line separating an application from everything else is somewhat fuzzy," says Case. "The standard joke is to say that the application is in the layer above where you stand."
So, almost from the beginning, all the major vendors in this space were forced into promising a final solution, namely one console that would give managers a unified view of the network and all the systems layered on top of it.
It never quite happened. "The problem these vendors ran into," explains Will Cappelli, analyst with Giga Information Group in Cambridge, MA, "is that network management is always dealing with yesterday's technologies. By the time they had gotten client server networks reasonably well under control, we had already moved to IP- and Web-centric networks."
Take Gillen's work at GTP, for example. She specializes in consulting with service providers such as telcos and ISPs. "Many of our clients are deploying optical and IP-based networks," says Gillen. "These are often connectionless networks. Classical SNMP monitoring tools were not build for these kinds of topologies."

Nevertheless, in today's more conservative economic climate, Gillen says that her clients insist on getting as much mileage as they can out of the tools they already have. "Our job almost always begins with a gap analysis," says Gillen. "We get a clear picture of the gap between what they can effectively manage now and the network they are building. Most clients do not want to throw out the network management tools they already have in place."

Top of the Pyramid

But they do want to extend their management reach. "It truly is a vast domain," says Pierre Champigneulle, managing director in the infrastructure solutions practice at KPMG Consulting in New York. "We go far beyond the classical definition of network management."
Champigneulle puts classical network management at the bottom of a three-layer pyramid. Next come applications and services such as change management. "Change management applied to the client's IT department is really happening for us now," says Champigneulle.
Application management is all about making application performance more transparent. "Network management here means providing a visualization of how the infrastructure affects an application," says Champigneulle.

At the apex of KPMG's pyramid is the business management layer, which can leverage everything underneath. "This is where we can really help the IT executive build an organization that is service-driven."
This level lies far above the world of routers and switches, and the reason it even exists is that today's technology applications are more tightly woven into business processes.
And this means that management is now more abstract. Ric Hughes, managing partner for PricewaterhouseCoopers' IT architecture group, explains it this way. "It used to be primarily about monitoring things, routers, hubs, switches, etc.," says Hughes. "Now it is about monitoring service levels. What clients want today is to correlate all this information and roll it into a level of service that IT can measure and maintain."
This is the apex of Champigneulle's pyramid, and the vendors also sense that it is where you need to be if you want to avoid that dreaded curse that eventually haunts all software and services — namely commoditization.
It is where Gerry Roy, director of product marketing at BMC Software, a leading network management vendor in Houston, has his eye. BMC started out in application management, a place that used to be fairly near the top of that pyramid. "We now think the trend over the next two years is going to be the next layer up, namely services," says Roy. "And this is also where things will be happening for consultants."
Consultants will play a critical role here, Roy thinks, because, unlike classical network statistics, services and business processes are unique to the customer. "We can provide the tools to measure just about anything on the network," says Roy. "But these things are only important to the customer insofar as they are components in a service. How do you define a service? How do you identify the appropriate metrics for it? There is no way to do this automatically."

It is a sign of the times that network management software vendors now admit they can't do it all. It was only a few years ago that Computer Associates, for example, was touting neural network and intelligent agent technology as the ultimate in network management automation. CA pushed Unicenter as the one product that could manage everything, and at one point even went as far as to instrument a Formula One racing car with SNMP.
But today, CA — like its competitors — is far more reserved. "We have modularized Unicenter," says David Hochhauser, vice president in charge of infrastructure management solutions at CA in Islandia, NY. "Customers are now more likely to want to buy management solutions one piece at a time."
Hochhauser sees particular opportunity for consultants in an area largely left untouched by classical network management software. "I think wireless will be a big challenge, and it is an ideal place for consultants to get involved. And there will be a lot of business issues here rolled in with the technology."

He says that he is already seeing customers wrestling with wireless LANs built around the 802.11b standard: "There are big network management challenges here."
Another area that could soon heat up is in managing wide area wireless networks. Antonio Conati Barbaro of Accenture Italy is working on the cutting edge of network management to give Wind, an Italian mobile telecommunications operator, a competitive edge.
"A few years ago, we saw that the next wave in mobile services would introduce new management problems," says Barbaro. "So we started building up our practice in this area and also identifying technical solutions."
The problem Accenture is trying to solve for Wind is wireless IP mediation. Specifically, Wind wants to collect data that will enable usage-based billing. "These are IP-based services," explains Barbaro, "and, to date, they have almost all been billed at a flat rate. In order to compete more effectively with larger carriers, Wind wants to bill clients based on the type of service, the time of day, and the volume of data."
Sounds logical, but the technology to do this is still in its infancy, which is why Accenture picked Xacct Technologies, a relative newcomer in the network management space, to provide the software. The Santa Clara, CA, company specializes in collecting and processing a lot of the stuff that classical SNMP tools can't get.
"We don't really compete with CA, HP, or Tivoli," says Eran Wagner, senior vice president of technology at Xacct. "We focus specifically on collecting network data that is related to revenues using a wide variety of methods. One of our customers is the largest cable operator in Europe. We enable them to identify network usage."
Wagner says that Xacct's software is 98 percent Java code and very distributed. "We have to be distributed. The amount of data we collect could easily bring the network to its knees if we had to bring everything back to a central location for analysis."

It's in the Process

This is the kind of thing that Paul Mason, an analyst with IDC in Framingham, MA, says ensures that network management will provide consultants with plenty of opportunities for some time to come. "Today's systems are multitiered structures," says Mason. "And when you factor in the requirement of supporting external users as well as internal ones, you have management complexities that are orders of magnitude greater than anything we have seen before."
And what about the Holy Grail of network management, namely a piece of software that can handle all your distributed systems? With vendors like CA backing off, the vision seems further away than ever.

But Giga's Cappelli thinks that this is only temporary. In fact, he thinks the big problem at the top of the management pyramid is going to put vendors back on the quest. "It is true," Cappelli says, "that management technologies are still only loosely integrated at best. But, convergence is coming about because both the network side of the world and the systems/application side of the world need to be mapped into business processes."
Cappelli admits that his view of a renewed drive toward convergence in management products is somewhat contrarian. GTP's Gillen, for example, sees a more modular future with a vendor shakeout coming soon. But they do agree on one thing: The most compelling opportunities for consultants in network management will be around business processes.
"The biggest opportunities for consultants," says Cappelli, "will be in building maps between business processes and underlying network events."
Gillen puts it this way: "The best network management software can't set policies and correlate them to your services and your business. The consultant who can do this will do well."

Sidebar: Classical Network Management in a Nutshell

It all began in 1987, when the Department of Defense was about to hand something called ARPANet over to the National Science Foundation.
This was the early Internet, and there was already a significant problem. "The NSF wanted to move to a more open technology architecture," explains Jeff Case, founder and CTO of SNMP Research in Knoxville, TN.

In particular, the NSF wanted a nonproprietary way to monitor and manage a relatively new piece of equipment called a router. "I was at the University of Tennessee at the time and got involved in writing what became SNMP (Simple Network Management Protocol)," says Case.
It was an instant hit. SNMP's greatest selling point was that it was nonproprietary. Case and others worked with the major router vendors of the day to hammer out standards that they could all agree on.
SNMP gave Internet managers the ability to do things that had been limited to the domain of proprietary networks, primarily from IBM and DEC (Digital Equipment Corporation). "We developed a graphical iconical map that would show the real-time status of your network," says Case, "and also the ability to do instrumentation of your routers, and set and change various parameters."

It started a revolution in network management. HP, IBM, and Computer Associates quickly started building software platforms on top of SNMP. "These companies added a lot of stuff you don't get with basic SNMP software," says Case. "SNMP can gather the data for you, but if you want to correlate, do number-crunching and display it, you need something more. This is what the classical network management vendors added."
By the mid-1990s, SNMP was everywhere, the standard for instrumenting, monitoring, and managing your network infrastructure. But already there were rumblings that this wasn't enough.

Applications and systems were still separate worlds, and users were increasingly clamoring for a way to bring these into the same orderly world that SNMP had created for networking gear.
It is a tension that is still unresolved. Systems vendors, the makers of disk drives and servers, along with application vendors, claimed that SNMP was not really suitable for their products.
Case has always disagreed. "SNMP was open and the market embraced it early," he says, "but then the vendors saw that with systems and applications they could compete with more expensive proprietary systems."
He says that there is no reason SNMP can't be used to monitor a database or a disk drive. "It is just the same old tension between standards-based, commodity stuff, and high-priced proprietary solutions," says Case. "Just follow the money."
— Mark Leon