FRAMINGHAM (11/03/2003) - Applications at Best Buy Canada/Future Shop can be so complicated that analyzing performance takes no fewer than eight monitoring tools. Jason Kennedy, a systems management analyst at the electronics retailer's headquarters in Burnaby, B.C., says he needs such a large tool set to ensure the core features of point-of-sale, inventory and other business-critical applications are working efficiently across 130 locations. Kennedy is says he hopes to ease his management burden by using Hewlett-Packard Co.'s OpenView Operations application performance management software, although he has no illusions about that tool belt getting any lighter. "I don't think there could be just one tool for application performance management," Kennedy says. "If someone did manage to build that tool, it would be so complex, it would be impossible to use."
In Best Buy 's case, OpenView Operations will provide a collection and correlation point for data gathered by other OpenView products, a Computer Associates International Inc.'s workflow management product and SNMP traps built into the company's homegrown applications, among other third-party monitoring tools. "If I can efficiently get to one pane of glass by bringing that data together, I can get closer to seeing the performance of the entire service connected to an application," Kennedy says.
Understand, then control
Kennedy is tackling one of the more challenging management tasks in networking today: understanding application behavior enough to control it. This isn't a new issue, by any means. But it is one growing in importance as companies continuously intertwine their business processes with those of customers and suppliers.
Users need strong application performance management capabilities to diagnose trouble quickly and eliminate finger-pointing among business partners. This has shifted demand from tools that monitor distinct network components to software that can measure the end-user experience, diagnosis the source of performance degradations, and understand and visually map the relationships between applications and the network components that supports them.
"Business managers only care if their apps or services are working, not about servers or network devices," says Jasmine Noel, principal with JNoel Associates.
At Virginia Polytechnic Institute and State University in Blacksburg, Brian Jones, manager of network engineering and operations, employs homegrown tools alongside Smarts' InCharge Service Assurance Manager software. He integrates network and availability data with SNMP traps and other information that might be pertinent to application performance. But as the importance of application management grows at Virginia Tech, Jones says his team will evaluate more commercial options that are open enough to integrate with the tools he uses today. "Streaming video, (voice over IP) and other (quality-of-service) demands make it more necessary than ever to be able to quickly diagnose and resolve issues. We are looking into systems that could serve as one correlation point," he says.
The choices are mind-boggling. Application performance management underpins such lofty technology goals as utility computing, service-level management and distributed Web services. So loads of network and systems management vendors recently have enhanced their products with features that address this need. Integration, automation, root-cause and traffic-analysis features figure prominently in application performance management systems, as do configuration, asset and inventory-management capabilities.
Integration helps network managers more quickly bring data from proprietary systems together, and automation removes the manual effort of that and other IT tasks. Root-cause analysis speeds problem diagnosis, telling a network manager, for example, that the reason a sales application is not available is because a router is down. And traffic analysis, when used with auto-discovery tools, can help network managers to see how applications touch network resources and to build a topology map.
"You can't talk seriously about performance management without knowledge of your technology inventory and how it is changing," Noel says.
Network management giants Computer Associates and HP promise new products to help companies track application performance outside the corporate firewall by way of Web services. Both will get to that end at least in part through acquired technology. HP says it will improve its OpenView software suite with Web services management technology it gained with the acquisition of Talking Blocks. The OpenView suite serves as the management software component of HP's larger Adaptive Enterprise initiative.
For its part, CA acquired Adjoin, with the plan to incorporate the start-up's Web services technology into its broader Unicenter platform. Later this year, CA is scheduled to release Unicenter Web Services Distributed Management, which uses agents to discover Web services automatically and monitor service characteristics of Web services transactions.
Start-ups such as Collation, Relicore and Vieo have targeted traffic analysis as the best way to get a sense of how an application performs, while Web-centric companies such as AltaWorks, Dirig Software and Wily Technology incorporate .Net and Java knowledge to help users see how Web-enabled applications behave (see chart, left).In the meantime, traffic management and acceleration vendors argue that their products defy bandwidth constraints and speed application performance, and companies such as start-up Reflectent take the end-user perspective even further by exploring application performance on the desktop.
Despite all the vendor attention, the technology to manage application performance just isn't there yet, contends Kim Ross, CIO of Nielsen Media Research in New York. Ross uses a tool belt of "classic monitoring tools" to track availability, but recently rolled out Collation's Confignia software. He says the software is starting to capture and monitor the many-to-many relationships between applications and network and system components, but still he says more work is needed.
"Missing are tools that let you keep track of the ever-expanding number of data center components and the information that can tell you the relationship between the applications and the components they run on," Ross says.
While management vendors say they hope to fill in those missing pieces, the best approach to application management performance really lies in application development.
As companies push toward Web services, in which they build applications to share with partners and customers, they're finding that legacy applications could be better managed if they were built differently. Using languages such as .Net and Java, developers can tell an application how to inform management software about its performance. If an application is built with management in mind or instrumented to export management data to third-party tools, tracking its performance on LANs, WANs, the Internet and partner and customers networks could get significantly easier, says Stephen Elliot, a senior analyst with IDC.
Best Buy's Kennedy agrees. "The real core applications - point of sale, inventory - we've written ourselves," he says. "Ideally, you build measuring capabilities into the application. Ideally, but not always."