Predicting the Unknown: E-Commerce Drives Demand for Capacity Planning

Enterprises are increasingly relying on network-intensive applications to support business processes that integrate employers, partners, suppliers and customers over the Internet. In the past, business-critical applications used dedicated network resources and simple growth projections for network capacity planning. Strategic capacity planning delivers substantial savings by minimizing unnecessary bandwidth upgrades and increases customer satisfaction by improving application performance.

Application performance assessment during pre-deployment planning is an essential step for ensuring customer satisfaction, particularly in the e-business era where the financial stakes are high. The recent growth of Web-based business-to-business, business-to-consumer, and intranet applications and re-centralization trends, are driving the explosive demand for network bandwidth.

A larger pipe can alleviate network congestion and increase throughput, but it has limited effect in improving performance problems caused by latency. Capacity planning can help IT organizations right size the infrastructure the first time and reduce costly errors. An ounce of prevention goes a long way in keeping pace with the dynamic networking environment. Deferred infrastructure upgrades can result in dramatic savings over time. To take full advantage of the power of prediction, however, IT organizations must commit to strategic management of resources, instead of the reactive management that is now pervasive.

Enterprises worldwide are increasingly relying on network-intensive applications to support business processes that integrate employers, partners, suppliers and customers over the Internet. In the past, business-critical applications, such as mainframe applications over proprietary protocols, used dedicated network resources and simple growth projections for network capacity planning.

The Internet has now become the public nerve center that interconnects global users internally and externally, and transports a multitude of applications with different priorities: Web, multimedia, CRM, ERP, e-mail and legacy applications. Traffic flows are erratic and unpredictable. Business-to-business, business-to-consumer and e-business transaction paths include the extended enterprise beyond the control of IT organizations.

By opening up the network to support e-business activities, the existing IT infrastructure must support more users, transactions, applications and servers on a network that is already suffering from performance problems. Without a well-defined capacity planning methodology and tool set, network managers must resort to trial-and-error fixes if performance degradation occurs.

Strategic capacity planning delivers substantial savings by minimizing unnecessary bandwidth upgrades, and increases customer satisfaction by improving application performance. Examples of the economic advantages are:

• Reducing the cost of expensive WAN carrier services that represent a significant percentage of the network’s total operating budget.

• Reducing the cost of response time problems that impact the productivity of mission-critical business processes.

• Reducing the cost of down time with sufficient capacity that accounts for growth and peak load.

• Reducing the cost of downtime with redundancy plans to address network connectivity failures.

• Reducing indirect costs associated with IT staff and consultants in trouble-shooting application performance problems during deployment.

From Reactive to Strategic

Despite these benefits, network capacity planning remains an ongoing challenge for IT organizations. In most cases, IT has taken a reactive, rather than proactive, approach for two major reasons: First, gathering the data needed to determine what applications are running on the network, who is using them, and how they are performing had been difficult. Second, corporate executives had viewed IT organizations as cost centers rather than value creation opportunities to gain competitiveness and reach new markets.

In reactive mode, IT operations would address a problem only after users complained about slow response times. Network managers would diagnose the device interfaces with a network element manager, capture network trace files with a portable protocol analyzer, and manually sift through the output to identify the source of the problem. This process requires technical staff with considerable expertise to yield useful information and is therefore expensive. At best, it is a time-consuming and error-prone process.

Industry acceptance of distributed protocol analyzers and remote network probes made obtaining network traffic utilization statistics easier. Network performance reports automated the data collection and report generation process. As the network performance monitors improved to analyze the traffic information at all seven layers, they helped the network managers isolate and resolve problems faster. However, the mode of operations was reactive, fire-fighting problems after the fact. Network managers would learn to use the historical reports to justify infrastructure upgrades.

Network capacity planning focuses on preventing problems before they occur. Modeling and simulation tools identify network design changes that are required to improve performance levels by increasing segment bandwidth or creating shorter traffic paths to reduce round trip delay. Also, these tools can estimate the impact of new applications on existing infrastructure, helping to determine whether the new traffic load will cause performance problems.

In some cases, such exercises find that it is not new traffic that will cause problems, but the fact that the baseline network capacity has already saturated the critical point. Network capacity planning takes into account the network topology and traffic flows in predicting future capacity requirements. Linear projection of historical trends falls short of accurately predicting end-to-end performance in a dynamic networking environment.

In spite of these technical advances, the successful application of network capacity planning tools often depends on the commitment of the IT organizations to proactively manage the network. Strategic planning requires cooperation from IT network operations (to analyze the current state of the network performance), from development (to understand how new applications will use the network), and from executive management (to define the business goals and priority). This process requires proactive management of resources, which is a trade-off between available resources, cost and business demands, rather than playing catch-up.

Understanding the network impact of applications prior to deployment is becoming increasingly important as enterprises become dependent on e-business applications. Traditional companies, in addition to dotcoms, need to be concerned about the performance of Web-based business-to-business, business-to-consumer and intranet applications, under peak conditions, with the greater visibility that e-commerce brings.

Assessing network readiness by analyzing the interaction between the application and infrastructure on a transaction by transaction basis eliminates guesswork and ensures customer satisfaction by maximizing application performance.

Establishing realistic service levels is an essential step in successfully managing customer expectations and proactively planning for undesirable conditions. Also, rightsizing the infrastructure allows IT organizations to control spending and get full value for the money.

Capacity planning has become easier for network managers and network architects with tools that automate the process of populating the baseline models. Streamlining the set-up process allows for dynamic forecasting using reality-based operational data. The output from predictive analysis is used to set future alarms that alert IT operations of performance problems before they occur. Emerging capacity planning solutions integrate performance data from the application service layer that spans across the client, network and server, and manage the complete performance management life cycle.

Plan Your Future

Is network capacity planning worth the effort? Yes – particularly with the arrival of products that make it easier and require less time setting up the model. Capacity planning is an indispensable tool for preventing performance problems as the infrastructure needs grow at the frenetic pace of e-business. The proactive approach helps IT organizations get it right the first time and reduce costly errors, rather than the hit-or-miss, "throw hardware at the problem," reactive tactics. u

– Rieko Sato is Director of Network Performance Product Management at Compuware Corporation.