• United States
Senior Editor

Server magic

Dec 23, 200210 mins
Data Center

New automated server configuration management software promises to lessen your staff’s workload in the data center while boosting server security, configuration consistency and availability.

Network downtime is not an option for . The Web-based “name your price” travel services company – widely known for its comic ads featuring “Star Trek” legend William Shatner’s crooning – cannot separate business from technology. Priceline’s IT infrastructure, applications and supporting software tools are the business.

Specifically, the company’s bottom line depends directly on 500 Microsoft Windows NT and Unix servers staying available to support millions of customer page views per day.

“Technology is the foundation of our business. If the Web site is closed, then the store is closed,” says Ron Rose, Priceline CIO. The Norwalk, Conn., company estimates that one hour of downtime costs about $50,000.

So last year when a slew of new companies started making claims about reducing manual tasks, speeding problem resolution, automating server configuration, and providing consistency and security across data centers, Rose paid attention. “We take mean time to repair on our servers so seriously that I had to investigate. The store cannot appear closed to any customers at any time,” he says.

BladeLogic, a systems and server management start-up, stood out among the companies Rose investigated. The company’s BladeLogic Configuration Manager  software modules provision, analyze and manage networked servers from a secure console. The software automates many configuration processes that normally are handled manually, such as the application of patches or the collection of inventory information. Rose says most appealing is that BladeLogic provides cross-platform management.

“Data center architectures are very problematic from an administrative perspective, so tools that help unify the administration of these heterogeneous environments are the wave of the future,” he says.

Priceline runs two Web site infrastructures across three physical sites – corporate headquarters and two data centers – with a total of 400 production Wintel or NT-based servers and about 100 big Sun servers for Oracle databases. Rose deployed BladeLogic about nine months ago in the data centers and has run the software in full production mode for about six months. With BladeLogic, “there are fewer detectable errors and a greater level of security across servers,” he says.

And, while Priceline has yet to perform a return-on-investment study, from a cost standpoint, Rose says he can control more servers with fewer people because manual configurations have been eliminated.

Moonlit servers and spinning plates

Besides BladeLogic, companies such as CenterRun Jareva Moonlight  and PlateSpin  have emerged in the past 18 months or so promising to reduce staff hours and eliminate configuration errors when provisioning data centers.

And established management software vendors such as Hewlett-Packard Novadigm  and Marimba  have enhanced their software to distribute patches automatically or configure servers consistently across data centers, for better security and availability. These management vendors are responding to a number of factors: the shortage of skilled IT workers; the need to eliminate server configuration errors, which can lead to network downtime, on e-business and e-commerce Web sites; and the evolution of software toward automation features.

Managed services provider Loudcloud even changed its business model and its name to get into the automated configuration management game. The company, now called Opsware, sold its managed services business to EDS, which also is Opsware’s largest customer. Opsware’s System 3 platform provides the data center management and automated server configuration Loudcloud once delivered as a service.

These companies all offer one form or another of server configuration and change management software, an IT area that vendors and users either overlooked or underestimated in the past as they rushed to provision servers and set up data centers for growing e-business initiatives.

But as network managers look to squeeze more out of existing staff and servers, industry watchers say automated configuration management tools will become a must-have in large corporate networks.

“Right now IT staffing is inelastic. Staff cannot solve the server provisioning and management problem for the long term,” says Ronni Colville, a research director with Gartner. “People don’t scale in the distributed way that e-business servers scale. I don’t see any way to staff up to adequately support the servers that support e-business.”

While specifics vary among vendors, the software works with a “master” centralized server and software agents deployed on managed servers throughout the data center. The server software communicates with the agents, which capture configuration and performance data on each managed server. In some cases, the software includes scripts that can kick off automated actions to report configuration errors or even dynamically change the configuration based on the knowledge built into the software. The master server can contain a data repository in which all user, application and server changes and actions are stored, or users can save the data elsewhere.

Software applications that perform specific automated tasks, such as deployment, comparison, notification and transport, also reside on the central server. And some vendors offer applications or software modules specific to third-party products, such as BEA Systems’ WebLogic or IBM WebSphere application servers, that provide automation out of the box.

These software tools can be pricey, ranging in cost from BladeLogic’s $25,000 entry price to Opsware’s $250,000 low-end implementation. But they promise to reduce staff hours by automating server provisioning and configuration – both labor-intensive tasks, especially as the number of servers in a network grows. The products also can provide security by scheduling and distributing software patches on a one-to-many basis, eliminating holes that human administrators potentially could miss.

In the country’s best interest

This security aspect is what led the U.S. Department of Energy to implement Opsware’s System 3 automation platform. The agency wanted to better secure technology assets and manage software licenses across its 130,000 users, says Karen Evans, the department’s CIO, because it felt “an increased responsibility to protect essential national cyber assets.”

In the past, the Energy Department manually provisioned servers, which didn’t ensure the configuration would be the same across all resources and didn’t create a central point of control. System 3 lets the department provision servers easily and maintain robust configuration control throughout the agency, and it gives Evans a handle on the software licenses across the department’s multiple sites. Opsware’s collection technology has helped “gain increased savings” by centrally administering the software licenses, Evans says.

Such also is the case at Royal & SunAlliance (R&SA) USA, a property and casualty insurance company in Charlotte, N.C. R&SA USA, winner of our 2002 User Excellence Award, tapped Novadigm’s Radia software for automatic distributions, then discovered its usefulness in getting a handle on software licenses. In a money-saving project, the company is negotiating higher-volume agreements on duplicate software licenses that individual offices had bought, such as mapping software, says Roger Thibodeau, chief network executive at R&SA USA.

Other benefits of server automation software are centralized control and volume distribution, which also serve network managers well when it comes to software application updates. Not only can these tools roll out application updates in volume, but they also quickly roll back to the previous version when an application configuration error threatens to take out one or more servers.

At Priceline, BladeLogic has tripled the speed of application rollbacks, Rose says. He notes that the company rolls out upwards of 300 application updates per month to its two cross-platform data centers. Some rollouts keep the company up to date on the latest technologies and others involve distributing patches, but until recently, application rollouts – and the rollback process when application errors threatened to cause downtime – required considerable manual work.

“Most people don’t focus on the speed of rollback, but you realize it’s important to cause a minimum impact on the Web site and customers,” Rose says. “The greater the speed you can roll back a problematic app, the more likely your Web site and services will remain available.”

While server configuration and change management software is not new, the automation and dynamic features included in these tools are, says Corey Ferengul, a program director with the Meta Group. He says these vendors are trying to eliminate server and application failures caused by human error and improve the ratio of administrators to servers, which tends to range, depending on network size, from 1-to-30 to 1-to-100 or more.

Automated server configuration software would lessen the time needed for the administrator to touch all the servers with updates and maintenance. The software ideally also wouldn’t include any errors. Users would input the configuration data correctly once and be able to reuse and apply it to all servers.

“The two processes that network management vendors did a terrible job of automating are configuration management and change management,” Ferengul says. “About half of the issues that happen when a change is made are manual errors. Automation will save human operators from inputting errors.”

While Ferengul says these tools meet a real customer demand, many vendors, because they are new, lack long lists of proven implementations – a must-have in today’s IT buying environment. Another drawback is convincing network administrators to change their processes to support these automated tools.

The market for intelligent configuration management is young, Ferengul says, noting that Meta predicts that by 2006 corporate users will look to automate much of the management of their complex application and Web server environments. And some big vendors, such as HP with its Utility Data Center software and IBM with its autonomic computing initiative, are working toward providing an intelligent network of hardware and software that can provision itself dynamically – with the knowledge and understanding of the environment programmed into the tools (see story, “Clash of the Server Titans” ).

Still, Ferengul says, most of today’s vendors that do bits and pieces of automated change and configuration management might lead the way toward broader implementation of intelligent automation tools. Patch management for Windows platforms will be a driver, he says.

“Say you have three or four patches per month, taking three or four hours per patch, per server, on 120 servers with, say, four administrators – you do the math,” Ferengul says. “There aren’t enough hours in the week to apply patches. Numbers like that will get people looking into automating these IT tasks.”

Getting smart

Established and start-up vendors alike have added automation features to their server provisioning and configuration management software tools. The list that follows is a sampling of what is available.

Provisioning and maintenance


Anduva, BladeLogicCenterRunContradoJareva (being acquired by Veritas), MoonlightMarimba and PlateSpin

What they do:

Automate the deployment of hardware, operating systems, patches and application.

How they work:

Maintain environment configuration data, building and maintaining it by tracking changes and updates made through the tool itself.

Resource virtualization


Blackstone ComputingHewlett-PackardPlatform Computing and Terraspring (acquired by Sun)

What they do:

Dynamically allocate technology resources among applications, with no regard to the physical boundaries of the devices.

How they work:

Take existing and new technology resources under their management once they are built and reallocate them in part or in whole across the infrastructure as required.

Dynamic provisioning and administration


NovadigmOpsware (formerly Loudcloud) and Think Control

What they do:

Provide provisioning and maintenance, and then intelligence to automate tasks, such as allocation or removal of servers.

How they work:

Interact with performance data to determine which application tiers require additional resources and then allocate infrastructure accordingly.
Source: Meta Group