• United States
Senior Editor

Net management’s new autonomy

Feb 24, 200312 mins
Cisco SystemsData CenterIBM

Automation has been a long-standing failure in network management products, but the latest generation of such tools might work as advertised.

Amid some of the worst economic times he’s seen in his 25-plus years in networking, Clyde Wilson needed to upgrade his network to better support Markel’s business.

“My budget is getting crunched, I’m not allowed to hire as many people and I need to do more with what I have today,” says the manager of technical services at the specialty insurance broker in Richmond, Va. “It’s just a sign of the times.”

To make his network more efficient without unloading a lot of cash, Wilson added software from HP, Heroix and others to automate management tasks across the network, which supports 12 locations.

The need to do more with less turned Wilson into an early adopter of today’s automated management tools. But he is not alone in his need for automation. Network executives across the board face similar concerns: reduced head count; tight or no budget dollars; and growing demand to support ever more complex business services.

Automation has long promised enterprise network companies a combination of increased efficiencies and cost savings, but the technology never really delivered on its promise. In 2002, IBM, HP, Sun and Cisco, along with some savvy start-ups, sought to eradicate automation’s bad name. Software and hardware vendors launched campaigns detailing product road maps – such as IBM with its autonomic computing and HP with its adaptive management – they say will satisfy enterprise needs for cost-cutting, network-optimizing tools.

“More than three-fourths of the average enterprise IT budget goes into keeping the lights on and maintaining the status quo,” says Zeus Kerravala, a vice president with The Yankee Group. With a meager 20% to 25% of their budget available for new and in some cases necessary IT projects, enterprise network managers such as Wilson decided in late 2001 and throughout 2002 to revisit automation and take a look at the slew of new tools.

In the beginning

Despite what vendors might say, the change they want to drive into every network IT shop this year isn’t a radical, new idea, but one that traces its origins back to the mainframe. Software giant Computer Associates introduced the first commercial job-scheduling product, CA-Scheduler, for mainframe environments in 1981.

Automated products use process rules and product-management information written into software applications to, say, monitor the CPU or memory utilization on a server at a scheduled time – without human intervention. For automation products to act on behalf of a network administrator, the knowledge of the network hardware and software applications needs to be built into the tools. Product developers write “if A happens, then do B” scenarios into the tools, and when a threshold is missed or a rule broken, the automation feature launches an action, such as paging a network operator or rebooting a server. But because of the dynamic nature of networks, early automation products quickly failed.

“Automation in distributed management has a fairly checkered past – lots of overpromising and underdelivering,” says Jasmine Noel, principal analyst at JNoel Associates.

Successful automation requires network configuration parameters included in the software to change as the network changes, and immature products could not deliver that capability. The variety and sheer volume of network devices, events and alarms also overwhelmed early tools designed to automate simple processes. Hence the young technology disappointed users with lengthy deployment cycles, and constant update and maintenance needs.

“But like everything else, the ability to automate management is improving,” Noel says. She says more vendors, such as Magnum Technologies, Micromuse and Smarts, learned to write intelligence into their tools to help automate network discovery, data aggregation and event correlation.

Leaders of the pack

IBM is leading the charge toward intelligent network management tools and automation software. Big Blue became the front-runner in the automation revolution in April 2001 when it announced its eLiza initiative, which rolled into the broader autonomic computing initiative announced in October 2002.

While IBM works to integrate self-healing, self-managing, self-provisioning and self-protecting capabilities across its hardware group and four software brands (potentially five counting the $2.1 billion acquisition of Rational Software at the end of 2002), its Tivoli division already delivers several of the promises – on a limited basis. Tivoli, the network and systems management (NSM) software arm of IBM, today delivers network, configuration, service-level and security management software, which includes self-managing and self-protecting features. Steve Wojtowecz, director of strategy for IBM Tivoli, says today’s automation efforts face the hurdles of user hesitance and vendor integration across heterogeneous environments. While Tivoli, with other vendors, must win users’ trust in automated offerings, software and hardware makers also must improve integration among their tools, either through standards work or partnerships, Wojtowecz says.

“It would be ideal to manage and automate cross-vendor technologies for the sake of the customers,” he says.

Not far behind IBM is rival HP. Also a hardware and software behemoth, HP made news with its Utility Data Center software announced in fall 2001 and its Adaptive Management Platform road map from its OpenView software division. The product strategy, sounding much like IBM’s autonomic computing, promises intelligent hardware and automated software that can ensure applications meet service levels, and free IT staff from constantly monitoring and reacting to network performance problems.

Jim Grant, OpenView’s general manager, says the current stream of automation features in HP’s management tools fall under the umbrella of service management. This moves today’s automation away from past efforts that focused on gee-whiz technology, such as automatically monitoring static devices that don’t support customers or end users.

Grant says corporate users need to target their automation efforts at the customer-facing services, and the elements that support those services. HP software can help clients identify, configure and automate the support of the 12 most-important applications to a client’s line of business, he says.

“Automation today can help enterprises translate business language into technology language and cross over the Tower of Babel that perhaps limited the potential of the tools in the past,” Grant says.

Top software NSM competitors BMC Software and CA could fall significantly behind IBM and HP because they lack a hardware element in their product portfolios.

“The hardware companies, such as IBM and HP, have approaches to drive down the monitoring and management information into the hardware environment,” says Rich Ptak, president of Ptak and Associates. “CA seemed to dismiss this idea last year, saying it was years away from reality, and it made a serious error.”

For CA’s part, the company says it succeeds against its hardware-driven competitors with software that can manage and automate across third-party gear and applications to provide agnostic management options for heterogeneous enterprise networks. But server giant Sun, an unlikely player in the management market, might provide the gear on which CA and BMC can hone their wares, industry experts say.

Sun, with its N1 initiative, might be avoiding the typical management jargon in its pitches, but the company can’t deny it’s looking to sell self-monitoring services as part of its N1 open architecture. The first phase of Sun’s staged approach involves server, storage and network virtualization to enable quick and cost-effective management.

Sun has partnered with BMC, which developed a version of its flagship Patrol software to automate management across Sun servers; and acquired Terraspring, a start-up with products that can automate server configuration and dynamically reallocate resources across data centers. Yet Ptak sees Sun as a distant third to IBM and HP in knowledge and experience in the network management realm. “Sun fails to link its ‘technology vision’ to a convincing chain of events and products that demonstrate the superiority of its vision over IBM’s or even HP’s,” he says.

Other large well-known companies to watch in the automated management space are Veritas and Microsoft.

Veritas closed 2002 by announcing it would acquire server provisioning software vendor Jareva Technologies and application performance management software maker Precise Software. While Veritas didn’t indicate it would expand from storage to a broader automated management approach, Ptak says a company with that kind of cash could make its mark in management.

It has the potential to be a disruptive influence if it does go that way,” he says.

The same goes for Microsoft.

“Management vendors should at least be thinking, ‘What is Microsoft going to do?'” says Jean-Pierre Garbani, a director with Giga Information Group. He points to Microsoft’s success with its Microsoft Operation Manager (MOM) software. He says Microsoft could look to improve upon the tool to address automated data center management. “The company did not show up on the management map at all, and then two years ago, it comes out with MOM and makes $300 million.”

Smart start-ups

New companies not intimidated by going up against the industry giants emerged last year: Vieo, Relicore, Troux and Appilog. These companies share the same philosophy that successful management must attack the problem from the top down.

Until now, most tools attempted to manage business services by monitoring the infrastructure devices that support them, then working their way up. Tools separately manage network devices such as switches and routers, and systems such as operating systems and servers. Then they correlate this information to determine application performance and quality of service. These start-ups propose to redefine the applications, underlying networks and back-end systems as components that support the service, thereby managing the whole vs. its parts. (See story “Never-fail business services“.)

Vieo, for instance, is noteworthy because it will use hardware to reroute application traffic automatically and improve the delivery of business services.

“Clearly a cultural problem has to be overcome. The focus has to move off the network and storage and server resources and focus entirely on supporting the application that supports the business,” CEO Robert Fabbio says.

Vieo’s Adaptive Application Infrastructure Management (AAIM) appliance – expected to ship in the middle of this year – is a Layer 2 switch that will initially manage hosts running Web, application and database servers. AAIM will watch traffic looking for application abnormalities compared with predefined policies that, for example, say that traffic from an online ordering application gets priority over an MPEG file being downloaded to a user’s desktop. Because AAIM is a switch, it can automatically redirect and reprioritize application traffic or reallocate network resources as needed.

Fabbio’s concept that hardware can better automate than software might prove successful. Cisco also builds intelligence into its hardware to enable management, among other things, across the infrastructure. Dubbed the programmable network layer, Cisco says the switches, routers and hubs include automated management features users can launch. (See story, “The evolution of resiliency” .)

Garbani says newcomers such as Appilog, Relicore and Troux also give the application top billing in their management approaches. The evolution of network discovery has evolved alongside automation, he says, from IP discovery to Layer 2 discovery to today’s forays into application discovery.

“By discovering the application on the network, you can now look at all the components of the application and understand how they interrelate,” Garbani says. The relationships between components can deliver a more accurate picture of where performance problems could crop up. Automatically discovering them would prevent network managers from pouring through event logs from multiple network devices.

Man vs. machine

Markel’s Wilson this year slowly will unlock and deploy more automation features. He says he knows the technology can help him make the network more efficient and let him direct his attention toward aligning the IT infrastructure with Markel’s line of business – rather than chasing down problems.

“We’re only at 20% automation now, but we’ll move our efforts closer to that ultimate goal of completely automating management further in 2003,” Wilson says. While he wants to use the technology to his advantage, Wilson says automation is not a cure-all that will replace human expertise.

“When the automation fails, because it will, IT experts will be needed because there is nothing more powerful than the human brain when it comes to IT,” Wilson adds.

Chris Utter, information services and technologies project leader at cosmetics giant Mary Kay in Dallas, agrees that automation can’t do it all.

“The theory that the current automated m anagement software can learn to respond to every scenario is naive,” Utter says.

Utter uses eight software tools to manage across networks, systems, applications and storage resources. Most recently he installed Micromuse’s Netcool/Omnibus management software, which automatically delivers a centralized, enterprise view of all the monitoring and alerting aspects of management in Mary Kay’s network. He says the idea of advanced automation features is exciting and could represent significant cost efficiencies for his organization, but he doesn’t believe automation will replace human intelligence and IT expertise.

“The personless network operations center is about as plausible as the paperless office,” he says. “The idea can actually be detrimental to process improvement.”

Still to come

Vendors still need to prove they can get automation right this time, but have been making progress.

Desktop Dynamos

Automated software distribution tools reach for patch management.

For any ranking of sexy technologies, software distribution tools would fall dangerously close to the bottom.

“Software distribution is beyond a trend, it’s a constant. It’s the plumbing,” says Ronni Colville, an analyst with Gartner. “But automating software distribution is truly a brutal problem to solve.”

Click here for more.

“2003 will not be a year of dramatic revolution for management vendors,” says Corey Ferengul, a program director with Meta Group. “We’re going to see a slow evolution where these automation concepts, seemingly solid in theory, take shape in real-world tools – or not.”

Joe Schinker had been burned by early automation tools, but recently turned to Smarts’ InCharge service assurance software.

“I learned I had to be careful of my expectations when vendors started talking about automation,” says Schinker, a network engineer at West Corp., a call-center provider in Omaha, Neb. InCharge autodiscovers West’s network at a preset time daily and then automatically updates configuration changes.

“Now we get a heads-up on the problem before we hear from an end user. And I’d say 75% to 80% of the automation in this software is out-of-the-box,” he reports. “If you told me last year that software could do that level of automation, I wouldn’t have believed it.”

Some still won’t – until today’s tools prove them wrong.