An Eye on Your Apps

Application monitoring software is becoming a critical part of many organizations' systems management portfolios.

It's midnight on the campus of Wayne State University in Detroit, but many students and faculty members are still at work, entering test grades, uploading notes, registering for classes or filling out financial aid forms using the university's online systems. John Camp, Wayne State's CIO, says it's critical that students and staffers be able to complete their tasks unhindered by sluggish Web servers or database bottlenecks.

Access to university systems is so important, in fact, that Wayne State recently installed the Vantage application monitoring tool from Compuware Corp. to sound an alert should any systems falter.

"It's strategically important that we make it convenient for people to register for classes, check grades and interact with professors," explains Camp, noting that students, like everyone else, expect round-the-clock access to online services such as e-mail and financial accounts, as well as their class notes and other university-provided systems.

"The drive toward self-service applications over the Internet has changed everything," Camp says. "People have very high expectations of availability now."

These days, there's virtually no business process that isn't automated by software, be it payroll, purchasing, inventory management, customer service or any of thousands of other daily activities. The increasing reliance on computers to conduct critical business transactions has motivated more organizations to invest in application monitoring and management technology, in addition to the traditional network- and hardware-monitoring products they already own.

Another factor is the increasing complexity of distributed applications. This interdependence of applications makes it more difficult to identify problems and often leads to finger-pointing between IT departments and outside vendors.

Henry Yiin, manager of network administration at CDC IXIS North America Inc., the U.S. arm of international banking firm CDC IXIS, says most people blame the network when something goes wrong. So he relies on Network Physics Inc.'s NP-1000 appliance to help pinpoint the actual source of failure in application performance.

"Two or three times, we've had a major server outage, and [the NP-1000] provided evidence that it wasn't the network's problem," says Yiin. The NP-1000 monitors trading applications and the company's Exchange e-mail server. A separate product, BMC Software Inc.'s Patrol, keeps an eye on the database.

Gartner Inc. analyst Laurie Wurster estimates that worldwide sales of application monitoring and management tools currently total $484 million annually. She has identified at least 58 vendors of application monitoring and/or management products. Wurster's research shows that sales of the tools grew by 30.7% from 2002 to 2003. "We're starting to see spending on things that will increase productivity, decrease downtime and make an organization run better for less," she says.

One way such tools improve productivity is by enabling less-technical employees to troubleshoot problems -- a boon for those lone database administrators or ERP experts who are tired of being on call all the time.

For instance, Shivaji Huttler, database manager for the Boise, Idaho, municipal government, replaced his collection of homegrown diagnostic scripts with BMC's Patrol for PeopleSoft to enable other IT employees to troubleshoot the city's PeopleSoft applications.

"It puts all the information in one place and makes it easy to drill down into the problem," he explains. "So people at the help desk and other IT managers can see at a glance the root cause of a problem."

Patrol was particularly useful when Boise migrated from PeopleSoft 7.5 to 8.0 and payroll processing slowed to a crawl. Patrol helped Huttler quickly identify the bottleneck. "I would have gotten to the root cause on my own, but with Patrol, I can do it in seconds," he says.

Where's My E-mail?

The traditional assumption in monitoring products is that if the server is up and the application is responding, then everything is fine. In recent years, however, IT workers have come to realize that problems are often the result of software glitches, not necessarily the server or network hardware.

"Monitoring a server doesn't necessarily tell you if everything is fine. It may be running at 25% CPU utilization, but users are suffering a 5-second delay in response time," says Gartner analyst Cameron Haight.

Because server monitoring alone won't catch every problem, many tools now track user transactions from beginning to end. On an e-commerce site, a tool may run a script that mimics a user logging in, selecting a product and paying for it. It checks that all the steps are completed correctly.

Marc Rieger, consulting and systems manager at HypoVereinsbank AG in Munich, says he appreciates the end-user view that Segue Software Inc.'s SilkCentral Performance Manager provides through screen captures of errors. "I can see what was on the screen when the error occurred. It's a root-cause-analysis function, which makes things easier."

Active vs. Passive

Some monitoring products do active monitoring, which involves constantly testing the application with synthetic user transactions, while others do passive monitoring, meaning they alert administrators only when an actual transaction fails. Some tools do both.

Sorin Fiscu, a project manger at Berkshire Life Insurance Company of America, a subsidiary of The Guardian Life Insurance Company of America, chose Empirix Inc.'s OneSight tool specifically for its active monitoring capabilities. He has set it to monitor a handful of critical user transactions. "If it finds a problem, I am -- hopefully -- alerted before the users are," he says.

Towers Perrin, a human resources consulting and management firm, uses Mercury Interactive Corp.'s performance management software for active monitoring and TeaLeaf Technology Inc.'s RealiTea for passive monitoring.

Michael Boyer, Towers Perrin's director of enterprise systems and data management, says he generally prefers the passive monitoring approach because it doesn't require updating of synthetic use-case scripts. Also, he says monitoring actual transactions gives him a more accurate picture of what's happening. "Only by watching all of the activities that real users conduct can you know for certain that your applications are performing the way they should," says Boyer.

Agents vs. Agentless

Some tools install agents on the monitored system to collect data. Others use an agentless approach that involves repeatedly contacting the application for updates. Alex Beardsley, IT monitoring manager at Navitaire Inc., an application service provider that provides reservations software to the airline industry, says both have pros and cons.

"There may be more talk over the network [with agentless technology] if you're going for heavy application analysis," he says. But for his environment, Beardsley prefers the agentless approach of Mercury Interactive's SiteScope because it requires less maintenance and consumes fewer CPU resources. "You don't have bulky agents everywhere, all of which have to be maintained and upgraded," he says.

Steven Lee, senior consultant at Tembec Inc., a forest products manufacturing company in Montreal, chose the agent-based approach of Heroix Corp.'s Heroix eQ Management Suite. "For us, the agent was important. A lot of products were server-centric, sending queries across the wire. We didn't want to transmit passwords or use heavy encryption," Lee says.

Agents tend to gather more in-depth information, says Yiin. "You need agents if you want to collect very granular system information," he explains.

A Tool for Every Task

Of course, a basic criterion for selecting an application monitoring product is support for the operating systems and applications that it must interact with. Tembec needed a tool to work with Oracle Corp. and Citrix Systems Inc. applications, as well as the VMS operating system. Towers Perrin wanted support for Unix, Oracle, Windows and legacy mainframe systems but found that many products supported only Windows, according to Boyer.

On the other hand, some companies want a tool that specializes in a single system or application. Both multivendor and single-vendor support have trade-offs, notes Gartner's Haight.

"The best-of-breed approach gives you better domain knowledge, which gives you better time to value," says Haight. "The rub is that at some point you may need to integrate it with the rest of your monitoring and management infrastructure."

Towers Perrin needed integration with reporting applications and archival software. "We made sure that it didn't use a proprietary data engine, but one based on Oracle, DB2 or SQL Server, and also that it could do data management -- either through proprietary mechanisms or support for third-party tools," says Boyer.

Diagnostics or other problem-identification features may also be required. For example, Wily Technology Inc.'s Introscope Transaction Tracer keeps a list of all transactions that exceeded a performance threshold, as well as a component-level breakdown of each. Other vendors, such as BMC, Segue and Mercury Interactive, also provide some form of diagnostics.

Just having a log of system events at the time of failure can be helpful. OptionsXpress Inc., an online brokerage in Chicago, relies on Identify Software Ltd.'s AppSight Black Box tool to find errors when an application fails. "It takes a snapshot of the event. We can walk through the steps the user was doing," explains Vlad Karpel, executive vice president of IT at OptionsXpress.

The choice of a monitoring tool depends on internal technical factors, but it should also be based on business requirements, such as which applications are so critical that they must be monitored and at what level of sophistication.

As Camp points out, Wayne State purchased a monitoring tool because of the importance of keeping the university's systems available around the clock. "It's not about the tool, but what we're trying to achieve, which is making it easy to do business with us," he says. "When we moved our systems to the Web, we knew that people would expect them to be always available. It's not like the mainframe days, when a system might be down for hours and nobody minded. We're in a different world now: exciting, but more challenging to manage." Hildreth is a freelance writer in Waltham, Mass. She can be reached at Sue.Hildreth@comcast.net.

Copyright © 2005 IDG Communications, Inc.

  
Shop Tech Products at Amazon