Like veteran poker players, network executives have long depended on a mix of skill, instincts and just plain luck to predict application performance. With business success hanging in the balance, however, today's stakes are too high for a gamble on application optimization.
Fortunately, today's network executives can deliver optimized application performance across local- and wide-area networks using a variety of technologies, and can spot slowdowns in real time using tactical tools. In fact, when planning their New Data Center infrastructure, network executives must incorporate application optimization technologies. The challenge is collecting the right mix of technologies and tools for a winning hand. (To read about perfecting application performance, see related story here.)
"Application management software is great when you are working with a constrained pipe. But if you are dealing with application traffic on the Internet, it becomes a whole new game," says Bruce Meyer, senior network engineer at ProMedica Health System in Toledo, Ohio.
Meyer invested in a Nortel Application Switch 2208 to get a better view of the application traffic consuming bandwidth among some 9,000 users at 209 healthcare facilities. He couples it with an open source, Linux-based reporting tool to get stats on top talkers and bandwidth consumption. The application switch, deployed between a gateway router and the firewall, can classify more than 90% of network and application traffic and apply predefined use policies to guarantee that critical application traffic, such as radiology services, gets priority over traffic such as peer-to-peer chats, he says.
"If you know what's traveling over your network, you can better control application flows and offer higher availability and better performance for business applications," Meyer says. "No one has told me about the business need for iTunes yet."
Meyer certainly is not alone in his quest to understand applications better. Across enterprises, the urgency of that task continues to multiply as data center consolidation thrives, telecommuting grows and complex distributed applications flourish. Whether used on their own or in concert with others, the tools available today should in one way or another tell network executives how to design their networks for applications and instrument applications for their particular networks. "There is a lot more intelligence available today about how the application uses the network and how the network can make the application more responsive, robust and reliable," says Joe Skorupa, a research director at Gartner.
Accelerators, switches and software - oh my
With such stakes, it's no surprise that network gear and management vendors are pitching product packages for eliminating application performance woes. For example, Cisco has its Network Application Performance Analysis push, while HP has Adaptive Enterprise. And that's not stopping newcomers, such as Certeon and Symphoniq, from pursuing the optimization market.
Opportunities should be plentiful, regardless. "Most network managers are going to need two or three clusters of tool sets to adequately attack this problem in the data center and across the WAN," says Dennis Drogseth, a vice president with Enterprise Management Associates. "They need visibility into application performance, usage and behavior as well as active capabilities that can take action in real time. It will be like Lego pieces coming together - a monitoring product sees the performance, an accelerator triggers a policy to ensure service delivery and an analytics tool comes into play for planning."
Application acceleration and WAN optimization devices perform a series of tasks, from compression to caching to server load balancing. The goal is ensuring remote and branch office users don't suffer performance degradations.
Traffic-analysis products from network-savvy vendors including Network General, Network Physics, NetQoS, NetScout and Visual Networks (recently acquired by Fluke Networks) can deliver application flow data, detailing the paths and hops application traffic takes between user requests to back-end systems.
More hardware options come in the form of application switches from Cisco, Extreme Networks and Nortel. An application switch acts as a proxy for Web applications and servers sitting behind it. Users can virtualize an unlimited number of back-end Web applications and servers, providing better availability, scaling and performance.
In the software realm, application performance management software from BMC Software, CA, HP and IBM provide infrastructure monitoring, critical to understanding the resources applications consume. Now the big four - with EMC SMARTS and Opsware - are augmenting their suites with application-discovery and dependency-mapping technologies (see related story). These tools, similar to tools from specialty start-ups such as nLayers, promise quick views of the sources of performance slowdowns through topological maps.
Tools that monitor users, from companies such as Reflectent (acquired by Citrix), Coradiant and ProactiveNet, address the most critical element to many network executives - the customer experience with an application. And don't forget predeployment profiling, testing and emulation tools from Apposite Technologies, Compuware, Mercury Interactive and Shunra. They equip network managers with knowledge of network and application performance before an application rollout.
"IT managers are warming up to the idea of spending more time with an application upstream, in predeployment, to ensure better performance on the live network," says Jean-Pierre Garbani, a research director with Forrester Research. "But still it's not enough, because often the rush to roll out the app limits the testing, and inevitably performance problems crop up later."
Be the user
Among the din of product pitches and business demands, network executives must hear one voice above the rest - that of the user.
"We have come a long way from how we used to manage applications. In the past, we really did more element management, focusing on managing the servers that hosted the application," says Jean-Philippe Draye, a system architect with Avaya's IT department, just outside of Brussels, Belgium. "Now we realize we need to be the end user."
Capturing the experience a user has with an application has long been the Holy Grail for application managers, but until the past few years the technologies involved in doing so required a lot of manual work and human correlation of data gathered from disparate IT resources. Niche vendors such as Reflectent and ProactiveNet work to deliver that perspective by using advanced analytics and other technologies, such as client-side agents.
"Coupling traffic monitoring with application performance management from the end-user perspective is the most interesting approach to tackling this problem," says George Hamilton, a Yankee Group research director. "Regardless of how you build the network to support applications, tools that monitor traffic will give you a better look at how the application performs, and client-side monitoring will do the same for the end-user experience."
Network managers have been using various software packages to capture the user experience and stop problems in their tracks. At Avaya, Draye uses a combination of management software products from HP's OpenView portfolio with Packeteer PacketShaper appliances to gauge application performance locally, at distributed locations and over the WAN. To get ahead of performance problems, HP OpenView Internet Services and Transaction Analyzer software applications let Draye set up about 30 workstations around the world as sample users. The PCs loaded with the HP software monitor about 50 applications by performing more than 50 simulated transactions. With this information, Draye ideally gets a heads-up to performance degradations before users experience a slowdown. By coupling information from Packeteer's devices with the alerts from HP software, he says he can identify the nature of a performance problem more quickly. "We get a yellow flag from OpenView Internet Services, and based on metrics we can determine if the application is slowing down globally or just for one location - and whether it's a problem with the application or with the network," Draye says.
While many network managers work to tweak applications in the data center, others focus on speeding the application over WAN links. Today these functions remain relatively separate, but industry watchers say a poorly designed application can wreak havoc on a LAN or WAN.
"Unless you do a really good job structuring the application on the back end, it can increase traffic and become a performance nightmare on the network," Gartner's Skorupa says. "A well-structured app can actually decrease traffic and become a dream."
Some say the slew of today's point products for performing a specific task, such as protocol optimization, will become part of other network equipment. For example, Cisco's Application Control Engine is a blade that slides into its Catalyst 6500 switches and performs several functions typically handled by load balancers, compression devices and application-acceleration devices, the company says. The blade resides in a switch deployed between a server and the WAN to improve traffic flow.
"A lot of the acceleration technologies will eventually become functions of the network, so investments in them today are mostly tactical," Yankee's Hamilton says.
Until the worlds collide and acceleration becomes part of the plumbing or developers discover how to design applications for optimal WAN travels, network managers have plenty of choices in acceleration and optimization technologies. Large vendors such as Juniper Networks down to newer players such as Crescendo Networks are hammering away at the problem, delivering products hand over fist to address chatty protocols and network latency issues.
And companies such as Packeteer are adding to their capabilities through acquisitions. Packeteer, known as an application traffic-management vendor, did so by buying Tacit for its wide-area file services technology.
At Rockwell Automation, 110 of Packeteer's PacketShaper appliances help manage traffic worldwide, says Dan Hanke, global network infrastructure manager at the Milwaukee company. The appliances let him give priority to business-critical traffic while still allowing employees to use noncritical applications.
"We didn't want to lock out all application traffic - if an employee does online banking during his 12-hour day, we are OK with that," Hanke says. "But we wanted to size our links appropriately so spikes in one type of application traffic from one user wouldn't affect performance of other applications."
Hanke represents a good example of how application optimization should be considered. While the products, approaches and performance demands vary, industry watchers say network managers need to remember the technologies available today are ultimately designed to enable the peaceful coexistence of many applications.
"Application management, acceleration or optimization tools should help customers prioritize and customize how their network and applications work together," Yankee's Hamilton says. "No one application should be set above the rest and sacrifice the performance of the others. These technologies should help network managers accommodate many applications traversing the network."
Learn more about this topic
Network World, 04/19/05
Network World, 03/06/06
Network World, 06/14/04