With multiple point products available, managers must mix and match to find the best approach for their networks.
Cosmetics giant Mary Kay is in the midst of an extreme makeover – one that will overhaul and enhance the company’s application performance management strategy.
Steve Moore, technology leader for Mary Kay Information Services and Technology in Dallas, says his organization is evolving its processes and fine-tuning technologies to be able to predict application performance issues, rather than simply react. While acknowledging it might sound clichéd, he says transitioning from a reactive firefighting group to a proactive application management team is an IT priority for the retail sales company.
“We care if a core router goes down, but we also want to be able to predict the other things that will be affected if one does,” Moore says. “For us, application nirvana will be becoming more predictive and understanding the behavior of applications directly in relation to our business processes.”
Moore is among those enterprise network managers who have been charged with adapting their basic element and device monitoring tools into intelligent application management systems that will dynamically respond to changing end-user and business demands and deliver optimized application performance. Unlike in the past, when availability management tools – which reported up/down status for devices, servers and so on – sufficed, today’s IT managers demand more information about potential performance bottlenecks across multiple networks and devices.
“Application management needs to bring together a lot of different data points and perform automated analysis,” says Stephen Elliot, a senior analyst with IDC. The market research firm estimates that performance-management software revenue will experience a 7.4% annual growth rate over the next five years, reaching $4.48 billion by 2007. Application management tools aren’t new, but in the past year or so, the demand for products grew and vendors began to attack application management from different angles.
Management veterans such as BMC Software, Computer Associates, Compuware, Concord Communications, Mercury Interactive, Micromuse, NetIQ and Smarts each offer multiple products that promise to address application performance management. The software typically measures response time across multiple infrastructure components and provides the option to identify the most likely cause of the problem, after the slowdown has occurred.
Newcomers such as Collation, Entuity, ProactiveNet, Troux Technologies, Vieo and Wily Technology also deliver tools to track response time metrics, potential application code errors and more. New vendors tend to talk more of monitoring all else for the sake of the application, and integrate with network management tools such as HP OpenView to bring those network fault and events into the analysis of application performance.
WAN optimization companies such as Expand Networks, NetScaler, Packeteer and Peribit Networks address application performance with their high-availability gear that speeds packets across WANs. While their focus is on the wide area, traffic monitoring tools also can watch application traffic on the LAN for performance issues.
CA, for example, acquired technology from SilentRunner in 2003 to add traffic monitoring and packet analysis capabilities to its Unicenter management software line. BMC recently partnered with Packeteer to add the latter vendor’s network traffic management technology to BMC’s application and systems management tools.
Industry heavyweights IBM and HP, which have been stumping for their respective utility computing product road maps, offer management software and services to help customers tackle application management. These vendors recognize that intelligent and dynamic application management is a critical component of their ultimate goal of automating data center operations, provisioning and resource allocation.
“The application performance management market is fierce; every vendor wants a piece of it,” says Jasmine Noel, principal at Ptak, Noel & Associates.
Despite the many ways to manage applications – or perhaps because of the multiple approaches – achieving an affordable and maintainable application management system continues to pose a challenge to enterprise network managers.
“Network managers are being inundated with events set on 40 different thresholds,” Elliot says. “They need correlation diagnostics, a tool that says all these events mean there could be an application performance problem at this time in this environment.”Monitoring application behavior is difficult for IT managers and vendors alike because applications typically don’t take the same route to fulfill each end-user request. For example, an online ordering system would send the Web site customer’s request for more information through a router to a Web server, which then needs to pull data from either a database or storage device. Once the information is gathered, the delivery could take a different route back to the end user.
Because these tools need to capture the data at each step, it becomes difficult for IT managers and vendor products to understand, control and then predict performance problems. Typically, capturing data at multiple points requires agents on managed systems, a time-consuming (and costly if considering product upgrades) process. “Right now for application management, network managers are defining one key service and starting there,” IDC’s Elliot says.
Adding to the challenge is that today there is no one guaranteed approach or one commercial product to manage applications. While vendors say otherwise, network managers seem to accept the fact and cobble together multiple tools to get data on how their critical applications behave.
“I can’t imagine having the financial resources or the staff to maintain that type of product if it did exist,” Moore says.
Yet Moore says he still is working to reduce the number of tools needed to monitor application performance and to do so cost-effectively. He uses products from BMC, Micromuse and NetIQ, among others, to track application performance across 600 or so Windows servers supporting Mary Kay’s Web environment. Still, when it comes to getting the simple red, yellow or green light to indicate overall application status, Moore says the dashboard capabilities of the products used at Mary Kay “at this time are not implemented on a wide scale.”
“We had multiple different screens, multiple different monitors all reporting on different metrics,” Moore says. He says he was able to reduce the number of management consoles and get more pertinent data by tying Micromuse and NetIQ closely together. “We’ve got them set up to go to the application layer and monitor response time within that environment,” he says.
Kim Jahnz, lead WAN data network engineer for Aurora Health Care in Milwaukee, reports a similar situation on her networks. She uses Concord’s products alongside Aprisma software and other tools to track the performance of core business and medical applications. Jahnz uses Concord’s Application Response feature to track specific response times and to better understand how the application interacts with end users. Application Response monitors the response of general repetitive functions within an application and overall user experience. She uses robot PCs in the field that run scripts performing certain functions to get a read on application response in different business units.
Now five years into working with Concord, Jahnz says application response monitoring is not a quick install, and configuring Concord and other systems such as Aprisma’s Spectrum is an ongoing process that involves IT and business managers. Like Moore, Jahnz needed to narrow down the monitoring metrics to get a handle on managing applications.
“Some [business managers] want everything monitored while others want one small function,” she says. “Too much monitoring and the numbers start to fade into each other; too little, and you aren’t getting a real ‘user perspective.'”
The team at Quest Diagnostics, a medical testing and diagnostics firm in Teterboro, N.J., uses BMC, Concord and HP OpenView software in conjunction to first prioritize applications and then to ensure service-level agreements (SLA) are met. The software monitors Windows 2000, Unix, mainframe, Oracle, Sybase, SQL, Exchange, BEA WebLogic, IBM WebSphere and more in Quest’s network.
For capacity planning manager Boris Gdalevich and senior analyst Sean Haines, working with business managers to define and then put tools in place that fulfill pre-set service levels is critical. Business goals, profits and potential losses play a direct role in Quest’s application management strategy.
“The highly profitable and visible projects, the ones that follow tough SLAs, are the first in line for proactive, intelligent enterprise monitoring,” Haines says.
Gdalevich adds, “Stockholders would not accept nor understand a situation when the company would pay out an SLA fee or when an investment shows no profit.”
While not business-related, IT managers running networks supporting military and government operations also struggle to keep applications in line and service delivery up to snuff.
The Marine Corps Network Operations and Security Command (MCNOSC) in Quantico, Va., houses all IT services for the Marine Corps. The Marine Corps Enterprise Network (MCEN) is for internal users only and contains about 14,000 devices and 8,000 routers, switches and bridges.
The MCNOSC uses HP OpenView for WAN monitoring; Concord eHealth for trend analysis; CiscoWorks 2000 for routers, switches and bridges; and NetIQ AppMan-ager for managing services, desktops and applications. MCEN runs 1,293 legacy applications – 60% desktop and 40% server.
Marine Corps Capt. Sabre Schnitzer says he needed to deploy a Network Common Operational Picture (NetCOP) in “Marine speak,” or Manager of Managers in technology talk. MCEN connects and controls all bases, posts and stations for the Corps, so “if service goes down, the entire Corps feels the impact,” Schnitzer says.
He purchased Smarts InCharge products to serve as the NetCOP to integrated data from multiple management systems.
“Microsoft Exchange 2003 and the Defense Messaging System are the Marine Corps personal and organizational messaging platforms. When it takes more than 30 seconds for a page to load, users go berserk, so we needed a notification system that sends an alert when response times for SAP hit 25 seconds or greater,” he says.
Schnitzer says his group worked with program managers to identify the metrics they wanted monitored and set thresholds for performance. He says if a threshold is set for Exchange at 75%, and the level is at 76%, the light is green; however, if it drops to 75% or 74%, the light goes red.
“Technicians need only the relevant information. They don’t need to track percentages,” he says. “They simply look at colors. Green is good. Red is bad. Simple.”
Ellis Admire, director of emerging technologies at Direct Holdings Services in Virginia Beach, Va., makes customer-facing and Web applications a top priority in the Candle (recently acquired by IBM) and Mercury products deployed on his network. He says the products don’t monitor “end-to-end” application performance, and he has to manually aggregate data from traffic monitoring and LAN monitoring systems into his application management system. He says the management process will continue to require upgrades and constant fine-tuning to adequately track application performance.
“Application monitoring is tricky because most tools require an agent on the platform and it needs a module specific to the application you are monitoring,” Admire says. “To come, there should be industry accepted standards of monitoring for applications such as the standards that exist for LAN and hardware monitoring.”
Groups such as the Datacenter Markup Language Organization, the Organization for the Advancement of Structured Information Standards and others are working toward standards that would let applications and the tools that monitor them more easily integrate and collaborate to enable automation. But while the industry works on standards and vendors add more intelligence and automation to their products, network managers say they will continue to manage applications with multiple tools, manual effort and more time than their schedules should allow.
“Good monitoring of applications takes time, a lot of planning and understanding of how the application actually works,” Aurora Health Care’s Jahnz says.
And Mary Kay’s Moore adds, “Don’t expect to ever get one tool to do this. There is no product that can give you the application management viewpoints you need from a technical and business perspective.”
