Maximizing Network Performance

David F. Carr Avatar

Updated on:

CRISIS AND OPPORTUNITY

If the pitch for application networking or application delivery networking sounds familiar, that’s because you’ve heard it before: A few years ago, these technologies were marketed under the banners of load balancing, Web caching, proxy services, wide-area network (WAN) acceleration or quality of service (QOS) network traffic prioritization.

Today, however, they are being combined in more sophisticated ways. Application acceleration products that were once specific to Web site operations are more relevant to enterprise applications. That’s partly because so many applications are Web-based and the latest generation of acceleration appliances supports a wider spectrum of protocols.

CIOs and executives overseeing corporate IT strategy don’t need to know how to install or configure these devices, but they should understand the potential of the technologies and the difference they can make in the performance of key business systems and user satisfaction. Getting the best results also hinges on understanding the characteristics of corporate networks and applications well enough to choose the right vendors and deploy the technology to the right places. Achieving success also requires understanding the staffing implications for network and data center management, given that the people who configure and administer these devices need to have a more application-oriented skill set than the people traditionally charged with the care of routers and switches.

The convergence of the words application and networking reflects an overarching goal to make network devices application-fluent, meaning that they go beyond routing and switching individual packets to applying optimizations that are specific to a particular application.

For example, WAN acceleration vendor Riverbed says Windows file sharing is one of the applications in which it delivers the most value to its clients. Because it was designed for fast LANs, Microsoft’s file-sharing protocol typically uses more than a thousand network round-trips for simple operations like retrieving a directory listing. That’s a slow process over a trans-Atlantic network connection where the speed of light puts an upper limit on how fast each signal can travel. But because the sequence of these transactions is predictable, Riverbed’s Steelhead appliances can intercept network directory and file copy requests and stream them across the network more efficiently.

Similar techniques can be applied to speed up Web applications, which require multiple network round-trips to retrieve the HTML for a page and its other components, such as images and style sheets. Web applications also benefit from compression to speed the transmission of the Web’s text-based data formats.

These technologies have matured just in time to match a spike in demand for network performance, driven by the centralization of data centers, proliferation of bandwidth-hungry applications, and verbose data formats used by XML Web services and business-to-business integration.

Application delivery networking can be either the solution to the crisis of an overloaded network or an opportunity to stretch the network to support new applications at a lower cost in money and bandwidth than would be possible otherwise.

“When I heard the pitch from the QOS vendors a few years ago, my question was always, ‘What is this solving for me that I can’t put off for a year or two?’” says Robert Whiteley, senior analyst for network operations and architecture at Forrester Research. “But now there’s a real sense of urgency. Workers and applications are evolving to become more network-dependent than ever. That’s driven by Web 2.0 or service-oriented architecture–if you’re going down either of those roads–or by the fact that workers are becoming more mobile.”

Gartner analyst Joe Skorupa echoes this theme, arguing that today’s application networking technology has become “a beneficiary of an unintended consequence of Sarbanes-Oxley compliance.” Because so many of those controls are embedded in computer systems, the law has driven large corporations to consolidate their information systems in centralized locations, rather than distributing data and applications to regional data centers.

Now an employee in a branch office–or even an overseas office–who accesses corporate financial applications is increasingly likely to do so over a WAN. CIOs and network managers are learning to proactively maximize network performance for those applications if they do not want to be buried in an avalanche of complaints, Skorupa says.

The processors embedded in network appliances are now much faster, making it practical to execute more elaborate rules without bogging down the basic network functions of routing packets. That wasn’t necessarily the case when the first differentiated QOS products debuted in the 1990s.

Another change is that application vendors have stopped claiming that their applications work just fine without any changes to the corporate network. “Now you’ll see things like best-practices guidelines written jointly by SAP and F5,” Skorupa says.

“This technology has made the jump from being a nice-to-have solution–or a Band-Aid–to being architected in from day one,” adds Cindy Borovick, research vice president for datacenter networks at IDC.

Ask your CEO and CFO: Is it likely that we will need to consolidate IT operations even further during the next several years?

Ask your branch office managers: Which corporate applications run out of our central data center are experiencing performance problems? If we could improve the performance of those apps, would it have a measurable impact on productivity?