Previous | Table of Contents | Next |
David Curley
Integration of voice and LAN networks will be an essential IT strategy for many businesses in the next three to five years. Consolidating the long-separate voice and data networks has implications not only for the network infrastructure, but also for the PC, the telephone set, the PBX, and the IT organization itself. This chapter is a road map to guide organizations in making the right voiceLAN-related investment decisions.
VoiceLAN is the transmission of voice traffic over a LAN infrastructure. VoiceLAN enables server-based telephony architecture for voice switches, terminals/phone sets, and applications.
Today, voice traffic is transmitted across a separate circuit-switched infrastructure, with a PBX or key system (for smaller offices) serving as a centralized switch. Under a voiceLAN scheme, both data and voice traffic are interleaved and switched as frames or cells over the same data network.
Organizations should consider running their voice traffic over the LAN infrastructure for several reasons:
Migration to voiceLAN is likely to encompass a number of smaller elements or activities. Migration cannot happen overnight, but is an evolutionary process that includes beneficial steps along the way. Over time, organizations can focus on improving elements of their network infrastructure, their desktop workstations, and their organizations, in addition to their telephone systems.
A first step in deploying voiceLAN is to upgrade the present LAN infrastructure to support the demands of voice traffic without affecting the flow of existing data traffic. Infrastructure refers to the cabling plant and the local networking equipment used to carry traffic from end station to end station (i.e., hub, bridge, router, switches, and network adapters). The PBX is not considered part of the infrastructure in a voiceLAN environment; rather the PBX will evolve into a call server that can be considered another type of end station on the LAN.
Voice bandwidth is not usually of much concern when using LANs for transmission. An uncompressed high-quality voice conversation needs only 64K-bps, and compression or packetization reduces bandwidth requirements further. This represents only a small fraction of a dedicated 10M-bps Ethernet LAN segment.
More important, voice is a delay-sensitive application that demands minimal latency (or minimal variations in latency, otherwise known as jitter) in communications. The vast majority of LANs today are based on shared-bandwidth media. With Ethernet LANs, all users contend for bandwidth on a first-come, first-served basis. Token Ring LANs are somewhat more deterministic, since each end station transmits only when that end station holds the token, which passes from end station to end station, at more or less regular time intervals. However, under both of these shared-bandwidth schemes, significant transmission delays, as well as variations in transmission delay, occur severely disrupting a real-time voice conversation between end stations.
Part of the solution to this problem is to provide dedicated bandwidth to each user end station through desktop LAN switching. In a fully switched network, end stations do not contend (as in Ethernet) or wait (as in Token Ring) for bandwidth with other users; instead, each user workstation gets its own dedicated LAN segment for connectivity into the network. Migrating to a fully switched network (i.e., a single workstation or server per dedicated switch port) entails replacing existing shared-media LAN hubs with LAN switches.
Dedicated LAN switching has become affordable. Commodity Ethernet switches currently sell for less than $200 per port ($US), and ATM25 switches can be obtained for less than $400 per port ($US).
LAN switching only addresses bandwidth contention to the desktop. Links between desktop switches, or from desktop switches to building/campus switches, must also provide predictable, minimal delays for voice communications.
In most enterprise networks, routers are used to calculate paths and forward packets between LAN segments at layer 3 of the OSI model. These routing algorithms introduce significant delay and usually add noticeable latency to voice communications. By contrast, switching involves a much simpler and faster process. Segmenting the network at OSI layer 2 through switching, rather than at layer 3 through routing, increases the capacity of the network to support delay-sensitive applications such as voice.
Although routing will continue to be necessary, especially in larger enterprise environments, implementation of voiceLAN requires minimizing routing in favor of switching. If deployed properly, switching removes the delay-inducing routing process from the path of most network traffic.
In many cases, this migration step entails replacing a collapsed backbone router with a backbone switch. The routing function can either be centralized through a one-armed router (or route server model) or distributed in switches providing desktop or departmental connectivity. In either case, traffic is typically switched through the network and only passes through a routing function when absolutely necessary.
Previous | Table of Contents | Next |