Optimize VoIP performance: class-of-service management adds value to an IP network - Voice Networks - Industry Overview

Communications News, Jan, 2003 by Gerry Roy

Voice over IP (VoIP) changes the network management paradigm, creating a complex problem for network administrators to assure quality of service (QoS) of this demanding application. The answer to this challenge is the use of class-of-service (CoS) management solutions to monitor, report and configure categories of services based on applications.

CoS management is a way to add value to an IP network, by prioritizing traffic based on its origin, destination or applicative nature. Without it, all the traffic has the same behavior. This also means that a company must be able to manage based on applications, not tech components.

With CoS management, IT classifies the traffic and creates categories. A company can order traffic by application or by a set of IP addresses; once created, IT applies policies to the classes of services.

If an organization wants to make sure that its primary business application is never interrupted by other traffic on the network, it can place that business application in a class by itself. Then, IT sets the policies so that the first class always gets priority on the network through the use of such techniques as priority queuing.

Managing each class of service and the parameters set for each can be complicated. IT can leverage its performance-management solutions, however, to automate this process. These solutions recognize devices and automatically discover the CoS structure, the QoS policies (such as policing, queuing or traffic shaping) and associated parameters that may be currently configured.

In providing a complete overview of all the parameters and configurations, these solutions provide advice to the network manager on how to fine-tune the parameters. In addition, the information gathered from performance-analysis tools can ensure that service-level agreements with customers are met. For example, performance--management solutions track quality-of-VoIP factors, such as where jitter is higher than 20 milliseconds, VoIP path latency is higher than 300 milliseconds and VoIP path availability is lower than 99.9%.

Application-centric network management (ACNM) can assure business availability by directly relating network management to an organization's business processes, focusing on the application itself to direct network policies and procedures. ACNM can help the network ensure that VoIP traffic receives the priority it needs to maintain conversations.

Critical to the success of ACNM is visibility into the entire network to examine how the flow of traffic is affecting the network. IT needs the ability to discover and view a topological map of the infrastructure, which includes the application and network resources and how they are interconnected. This technology can be used for diagnostic purposes m providing traffic visualization, which enables quick pinpointing and understanding of the cause of problems, and real-time observation on a selection of devices and links.

Tuning and diagnostics boost network performance. In addition to network visualization, diagnostic software can automatically find the network parameters that contribute to an increase in response time. This kind of complex correlation is impossible for staff to perform manually, and is essential in a VoIP environment.

Root-cause diagnosis requires information from the individual components of the system. This information is gathered by probes, agents and, more generally, every infrastructure component that globally instruments the system, such as routers and switches. These local metrics (e.g., CPU usage, congestion rate or CoS load) are collected periodically for each monitored device and then stored.

These metrics are useful for reporting on network performance individually on each device and interface, or to determine whether they can be correlated with excessive response time. The analysis discovers anomalies in end-to-end response times by comparing the response time data over a period of time to all the network metrics that are currently being monitored, to determine the metrics that represent the response time degradation.

For more information from BMC Software: www.rsleads.com/301cn-252

Roy is director of product management and development for PATROL network management solutions at BMC Software Inc., Houston.