As a service provider, NEON continues to see strong demand for bandwidth. To keep up with this strong demand, and ensure bandwidth is in place when needed to meet our customer's needs, NEON has developed a practical capacity planning process and method that uses historical growth rates combined with forecasts for future demand to predict and validate future capacity needs. Although the process and method described below grew out of NEON's need for an objective and credible approach to capacity planning, it nevertheless can be used effectively in the enterprise space to determine when and how much capacity is needed. The method outlined below is primarily for networks that use pre-determined circuit routes, deterministic networks. However, the principles can also be applied to switched non-deterministic networks, where the circuits are set up automatically on an as needed basis.

What is capacity planning?

The U.S. Marine Corps defines capacity planning and management as, "The process by which measurements of current resource utilization are combined with projections of future resource requirements to allow management decisions to be made as to what computer and data communications resources will be required in the future, and how best to allocate existing resources so that they are used in the most efficient and effective manner." (MCO 5233.2, CCIS-47, 12 Apr 90)

Although the above definition is from an unusual source, it succinctly lists the two key pieces of information that are required to do capacity planning: 1) how much bandwidth (capacity) is installed or used, and 2) how much bandwidth will be needed in the future. The definition also states the purpose of capacity planning: how to best provide and use that bandwidth.

Historical growth rates and forecasts are used in access and core networks alike to predict future bandwidth requirements, but in slightly different ways. For enterprises, their internal demand set can be viewed in two ways: the campus network is considered an access network, and the network connecting multiple campuses is the core network. In the access network there is often only one way to route traffic back to the core, regardless of system topology. Because of this, the statistical analysis for both the access network and the end office is fairly straight forward and often the same analysis. By plotting circuit installations and cancellations month-by-month and analyzing them over six-month intervals, predicted growth rates can be calculated and compared across the six-month intervals. Confidence in the predicted growth rate can be achieved by comparing the predicted growth rate versus the forecast. This is a critical step and one that cannot be overlooked. The predicted growth rate validates the forecast and the forecast keys the capacity planners into significant changes in bandwidth consumption that may or may not affect the overall growth rate, but could significantly reduce, exhaust, or add bandwidth to the access network or end office.

The growth rate table and graph shown in the center box are examples of the statistical analysis NEON uses to determine growth rates. The table shows the STS1 (synchronous transport signal level 1) bandwidth (about 50Mb/s) used in six month intervals, while the graph shows when, and in what quantity, the bandwidth growth occurred. The data points plotted by the graph are the monthly provisioned bandwidth to the site. From these data points a trend model is derived. In this case it is a linear model, but often it is a quadratic model - a curve. As can be seen from the graph, the bandwidth did not grow smoothly, but instead grew in chunks. If we were to derive the growth rate based on the first six months or the last three months, we would have derived a growth rate in excess of actual demand. This would have resulted in NEON adding bandwidth to the network when it wasn't needed. In the example below, our unit of measure is the STS1s because this is a convenient SONET building block, but this is not the only unit of measure that can be used. For enterprises, the unit of measure could easily be Mb/s (Megabits per second) or Gb/s (Gigabits per second). The end result is the same however; plotting bandwidth growth over time and deriving the statistical growth rate.

The same basic process is used in the core network; however, the growth rate cannot be calculated from the circuit installation and cancellation dates. This is due to the multiple routing options usually found in the core, which allows circuits to be groomed across the network in response to customer needs and network activity. Because circuits may not always reside on the same link the installation or cancellation dates of circuits cannot be used as data points, since it is possible that a circuit could predate the installation of the link. The result could be a very low or very high growth rate; both are undesirable. Instead, bandwidth consumption from the link installation date is used to calculate the growth rate. In addition to the growth rate and forecast, a preferred routing table is also required to model how the core links could be impacted by the forecast. Once again the forecast is used to validate the predicted growth rate and vice versa, based on preferred routing for the different "A" to "Z" combinations in the forecast. For many enterprises, calculating the growth rate as a function of consumption over time will be very close to that of an access network unless the enterprise has a private, high capacity network carrying with it a fair number of circuit moves.

The analysis outlined above not only works for a deterministic network, where the end-to-end routing of a circuit doesn't change based on usage, but also on non-deterministic networks, where circuits are set up as needed using the best route available. For non-deterministic networks the bandwidth used on the links between locations would be measured and recorded over time, and a growth rate derived from the resulting data points.

Now that all the hard work is done, the final step in the process, which is the same for both carriers and enterprises, is to calculate the dates when new incremental capacity should be added to the network and determine how much capacity should be added. The amount of capacity to be added is derived from the respective growth rate of the link or end office being analyzed, the timeframe over which the bandwidth is expected to last, as well as the time required to install the additional capacity. At this point other network factors, such as costs and equipment constraints, may need to be considered when developing the capacity solution.

Benefits of this approach

This approach to capacity planning provides a number of benefits. First, the process and data analysis establishes credibility within the organization and company that the numbers and subsequent recommendations derived from the numbers are not the result of guess work or an anomaly. Second, it establishes a proactive, instead of reactive, approach to identifying the parts of the network that require or will require additional capacity. And third, the process and analysis can be built cost effectively from off-the-shelf applications, as the chart and graph demonstrate.

In summary, we have outlined a practical capacity planning method and process that can be used by companies both large and small to credibly predict a network's future capacity needs by using statistically derived historical growth rates and forecasts for future demand in concert with each other. Using one without the other is like only reading every other page in a book - each page may be interesting but you are only getting half the story. Using this disciplined approach significantly reduces the chance of starving the network of capacity in some locations while carrying excess capacity in others.

When evaluating bandwidth solutions from service providers, ensure that they too are actively managing their own network capacity and have a defined process in place for determining when additional capital investments should be made to accommodate the demand. Without this assurance, you could be left waiting for additional capacity when your need is most critical. NEON provides the assurance you need and takes the guesswork out of capacity planning.

About NEON Communications Group, Inc.

NEON is a facilities-based wholesale communications provider, supplying high bandwidth fiber optic capacity and comprehensive end-to-end telecom services to communications companies and enterprise customers on an intercity, regional and metro network in the 12-state Northeast and mid-Atlantic region. With 4,800 route miles and over 230,000 fiber miles from Maine to Virginia, NEON is providing unparalleled capillarity and central office connectivity in the world's most demanding telecom market.

For more information, please contact Maura Mahoney, Senior Director of Marketing at NEON Communications, Inc., 508-621-1869;

email: mmahoney@neoninc.com; web: www.neoninc.com.