A large component to a successful voice over IP (VoIP) infrastructure is effectively deploying a quality of service (QoS) model. Real time traffic, such as voice traffic, requires high availability, low delay bandwidth across the routed infrastructure. In order to provide these metrics to assure the reliability of voice calls, quality of service components are configured across the data infrastructure. These components entail classification and marking, congestion management, congestion avoidance, policing and shaping, and link efficiency.
Classification and marking is a system of identifying packets or traffic flows and assigning certain parameters within the packet headers in order to group them. Various methods of packet identification can be based on source or destination IP addresses, input/output interfaces within a network component, or components of an access list. Once the traffic is "identified", it can be marked via several methods, such as the IP Precedence bits (the three most significant bits within a ToS byte of an IP header), or the Differential Service Code Point bits (the three most significant bits within a ToS or Traffic Class byte of an IP header).
Congestion management is intended to control congestion as it transpires. Several queuing methods are available to prioritize traffic based on their classifications or characteristics. Some of the well-known queuing methods are priority queuing, weighted fair queuing, custom queuing, and low latency queuing. Traditionally, data traffic was handled on a first in first out (FIFO) basis. Today, with the performance requirements to transmit voice traffic across a data infrastructure, congestion management techniques are available to carve out highly available, low delay bandwidth across the data network.
Congestion avoidance prepares for and prevents congestion on typical bottlenecks throughout the data network. These techniques all involve dropping traffic during times of congestion, but the differences within these methods are the determining factors behind which packets to drop. Some of the common techniques are Tail Drop, Random Early Detection (RED), and Weighted Random Early Detection (WRED). Within some of these methods, traffic is prioritized with low-priority traffic having high drop probability during times of congestion.
Policing and shaping is used to limit traffic flow. Policing drops or remarks traffic that exceeds limits, but shaping regulates the traffic back to a defined rate by delaying or queuing the traffic. Some common policing and shaping techniques are token bucket, committed access rate (CAR), and class based shaping and policing.
Link efficiency is the practice of fragmenting large data packets on slow links to reduce the "wait" time in order for voice packets to gain access to the network. Link Fragmentation and Interleaving (LFI), for example, is a technique that can fragment a large data packet into five smaller packets and allow a voice packet to be interleaved between the smaller data packets to minimize the delay necessary to reliably transmit the packet across the data infrastructure.
Richard Parsons (CCIE#5719) is a Principal Architect for AT&T with a focus on network planning, design, and implementation. He has built a solid foundation in networking concepts, advanced troubleshooting, and monitoring in areas such as optical, ATM, VoIP, routed, routing, and storage infrastructures. Rich resides in Atlanta GA, and is a graduate of Clemson University. His background includes consulting positions at International Network Services, Lucent, and Callisma.
