Cisco Packetized Voice and Data Integration

Cisco Packetized Voice and Data Integration

by Robert Caputo

Cisco Packetized Voice and Data Integration gives you the knowledge and insight to quickly get up to speed on planning, designing, and implementing Cisco-based voice and data networks. Through the use of real-world examples, case studies, and detailed configuration analyses, you will obtain the skills necessary to successfully deploy voice and data networks. See more details below


Cisco Packetized Voice and Data Integration gives you the knowledge and insight to quickly get up to speed on planning, designing, and implementing Cisco-based voice and data networks. Through the use of real-world examples, case studies, and detailed configuration analyses, you will obtain the skills necessary to successfully deploy voice and data networks.

Product Details

McGraw-Hill Professional
Publication date:
McGraw-Hill Technical Expert Series
Product dimensions:
7.36(w) x 9.15(h) x 0.98(d)

Read an Excerpt

Chapter 2: Packetized Voice Overview


Compressed Real-Time Protocol

Compressed Real-Time Protocol, or CRTP, is a mechanism that operates on the RTP headers used to encapsulate digitized voice segments. CRTP draws heavily from Van Jacobson's TCP header compression algorithm and manages to reduce the standard IP/UDP/RTP header from 40 bytes to a mere 2-5 bytes. This is significant, as it reduces both bandwidth consumption and the time required for a voice packet to traverse a low-speed network link. The efficiencies gained by CRTP help offset some of the overhead associated with packetized voice traffic. Tables 2-2 through 2-5 demonstrate both packetization overhead and the efficiency gains achieved using CRTP as well as some of the efficiency losses incurred by packetizing voice traffic. A 6-byte data-link header is assumed as this size is consistent with frame relay, PPP, and HDLC header sizes. Low-bit-rate encoders, such as MP-MLQ and CS-ACELP, benefit the most from CRTP because of their smaller payload relative to the IP/UDP/RTP header size. Their gains are significant when considering the transmission of multiple voice calls over low-speed links. Without CRTP, only two MP-MLQ or CSACELP calls could be supported on a 56-kbps link; with CRTP, four calls could be supported over the same link. Further, lower end-to-end delays will be incurred for each call since the serialization delay for each packet will be 100 to 135 percent lower.

Weighted Fair Queuing

Weighted fair queuing, or WFQ was first introduced in IOS v11.0 and has since become the default queuing mechanism for WAN interfaces running at less than 2 Mbps. WFQ classifies traffic streams, queues them individually, and ensures fair access to the exit interface by intelligently scheduling queue transmissions using a weighted algorithm. Traffic is classified by network and transport-layer information as well as traffic rates and frame sizes. Low-bandwidth sessions are given higher priority to prevent them from being starved by the higher bandwidth sessions. This behavior results in more consistent and predictable transmission delays for both high-bandwidth sessions such as file transfers and low-bandwidth interactive sessions such as Telnet or chat. Weighted fair queuing's greatest virtues are that it operates automatically without requiring the user to configure complex queue structures and that it seamlessly supports all protocols. WFQ is important in voice environments because it automatically prioritizes voice traffic, especially when QoS parameters are set, and provides a consistent queuing delay through the network.

Weighted Random Early Detection

Weighted random early detection, or WRED, is a lightweight congestion avoidance technique that enables higher overall link utilization and goodput. Random early detection monitors link congestion and traffic flows and randomly drops packets from individual traffic flows, allowing their upper-layer protocols to adapt to the congestion condition. This technique optimizes the transmission rates of the individual flows and prevents both congestion collapse and synchronization issues. Weighted Random Early Detection provides a means for supporting multiple priority levels, each with different discard thresholds. WRED helps voice over IP networks by providing preferential service to voice traffic while intelligently controlling nonvoice traffic and optimizing link efficiency.

Generic Cell Rate Algorithm

Generic cell rate algorithm, or GCRA, was originally known as the dual leaky-bucket algorithm. As the newer name implies, GCRA provides a means for controlling the rate at which traffic flows into the network. The dual leaky-bucket algorithm uses common traffic control parameters such as minimum cell rate and burst size along with a complicated queuing and scheduling algorithm to police and shape traffic flows through the network. GCRA is directly applicable to ATM traffic, but variants of the leaky-bucket algorithm are applicable to variable-length packet networks as well. GCRA can be used in voice over IP networks to guarantee minimum bandwidth levels to voice traffic and to shape nonvoice traffic and prevent it from introducing undue delay to voice traffic.

Committed Access Rate

Committed access rate, or CAR, is a value-added feature from Cisco that runs on the 7x00 class routers. CAR provides a means of controlling the bandwidth received or transmitted through an interface to a set of destinations. The set of destinations can be a discrete number of sites, specific TCP or UDP ports, or simply all traffic through the interface. CAR measures traffic flow using a token-bucket algorithm and can tag, drop, or simply forward traffic based upon decision criteria for meeting or exceeding traffic rates. CAR is helpful in engineering networks to support differentiated service levels as multiple CAR policies can be applied to a single interface to control access to network bandwidth. It is useful in voice environments to help ensure that voice traffic receives the appropriate amount of bandwidth across the backbone.

Token Bucket-Generic Traffic Shaping

Cisco's generic traffic shaping uses a token bucket algorithm to monitor and shape traffic flow through a given interface. A token bucket algorithm differs from a leaky-bucket algorithm in two important ways. First, it transmits traffic at media rate instead of a specified rate, and, second, it provides greater adaptability for bursty traffic. Generic traffic shaping allows for excess traffic to be transmitted, but does so at a lower priority. The interface's queuing mechanism then filters out noncompliant traffic. Generic traffic shaping is of value in voice over IP networks because it provides a lower cost (processing-wise) way of controlling traffic flows and allows for both voice over IP traffic to be prioritized and nonvoice over IP traffic's effects to be limited.

Frame-Relay Traffic Shaping

Cisco's implementation of frame-relay traffic shaping uses a single leakybucket algorithm to solve some of the QoS problems associated with frame-relay interaction. Frame-relay traffic shaping provides a means for limiting output rates on an individual PVC basis. The strict output rates enforced by the leaky-bucket algorithm prevent the router from transmitting traffic in excess of the individual committed rates for each PVC. This overcomes a common problem where "hub" sites with highspeed access lines overwhelm "spoke" sites with lower-speed access lines, Traffic shaping minimizes packet loss and helps ensure that critical traffic gets through, which is of significant importance to voice traffic.

IP Precedence

The IP precedence bits are special bits reserved within the headers of IP datagrams. The developers of the Internet Protocol envisioned that they would one day be used to prioritize traffic, but until recently, they have been ignored by most systems. Cisco uses the IP precedence bits to provide network-wide classes of service by integrating precedence bits into queuing and classification algorithms. CAR, WRED, and WFQ all utilize the IP precedence bits to help prioritize traffic. In voice environments, IP precedence bits are important as they are frequently used to ensure that voice traffic receives high priority when transiting the network.

IP to ATM Class of Service

Cisco has introduced a method for providing differentiated classes of service for IP traffic transiting an ATM network. Within the scheme, traffic is queued on a per-ATM virtual circuit basis and IP QoS mechanisms are performed on the individual queues. This feature relies on ATM's ability to guarantee cell delivery rates. IP to ATM class of service (CoS) is in its infancy, but will grow to support complex methods of forwarding IP traffic over ATM networks, including flexible selection of output VCs based upon traffic priority. Thus, in the future, the router will be able to transmit delay sensitive voice traffic over a CBR or rt-VBR circuit, while transmitting normal data traffic over an ABR circuit. This feature is of significant value to voice over IP networks with ATM backbones in that it helps to marry the QoS mechanisms from both technologies to form a more cohesive network that can provide consistent voice quality...

Read More

Customer Reviews

Average Review:

Write a Review

and post it to your social network


Most Helpful Customer Reviews

See all customer reviews >