A planner`s guide for fiber-optic broadband networks
Understanding the technology issues associated with moving to high-speed optical broadband networks should hel¥planners structure an orderly evolution
t3plus networking inc.
To ensure fiber-optic broadband networks merit business value, planners must evaluate available services, select those that meet the application and specify devices that guarantee flexibility and investment protection.
Today, many voice, video and data applications consume less than 128 kilobits per second per application, which is well below the threshold capacity of megabit-to-gigabit fiber-optic broadband networks. For these applications to cost-justify broadband networking, they must be combined onto a common facility--a fiber-optic trunk or a carrier service, for instance.
The cost of fiber-optic trunks, not including multiplexing equipment, is approximately the same for digital carrier system, level 1 (1.544-Mbit/sec), T3 (45-Mbit/sec), synchronous optical network optical carrier, level 1 (51.84-Mbit/sec) and Sonet OC-3 (155.52-Mbit/sec) networks.
In all cases, total bandwidth is the key network parameter. To determine this value, network planners must audit company private lines and tally the switched public network calls.
The options for communications connections include two broad classes:
Local interfaces, normally device- or application-specific, are used when a particular kind of information is generated and consumed in the same facility. An example of a local interface is the RS-232 serial data interface or the high-speed serial interface.
Carrier interfaces, defined to allow user connection to a carrier-provided communications service, are frequently application- and device-independent. An example of a carrier interface is the T1 interface.
Despite the fact that carrier interfaces define only carrier-to-user boundaries, many companies extend these interfaces within the enterprise to voice, video or data devices. For example, network planners might add a T1 interface to a router; use a local T1 multiplexer to combine voice, video or other data traffic onto a single fiber-optic trunk; and combine interfaces and devices with other T1 sources and connect them to a broadband service. This setu¥is beneficial where T1 is employed in private networking applications.
The practice of using carrier service interfaces on individual devices can limit broadband opportunities, however. For example, an asynchronous transfer mode broadband multiplexer could apply ATM Adaptation Layer 5 to provide Class C services to the data application. When the data application is idle, Class C services would permit the bandwidth to be assigned to other applications. But if the data is "buried" in one digital signal, level 0 of a T1 channel, the broadband multiplexer might not be able to separate that data out for special handling. An ATM multiplexer might assign the entire T1 channel to ATM Adaptation Layer 1 Class A service handling, which would consume 1.9 Mbits/sec of bandwidth for each 1.544-Mbit/sec signal transmitted, regardless of the active or inactive state of any DS-0 channel on that T1 connection.
To preserve service flexibility, a first rule in network planning is to avoid the use of direct carrier service interfaces for non-voice devices. This approach means using high-speed serial, V.35 or RS-232 interfaces on data devices, and avoiding the use of low-level multiplexers to feed broadband multiplexers.
The best broadband product that network planners can buy is one that will connect to a variety of voice, video and data sources using local rather than carrier interfaces, and that will support carrier services from T1 to the broadband range. The broadband multiplexer or node is the critical element in a successful broadband network plan.
A second rule for broadband networking is to consider multimegabit inverse multiplexing when multiple T1s serve the same destination. A network linked by multiple T1s can be considered a fractional T3 network. Treating it as such, with non-channelized inverse-multiplexed trunks between sites, makes the network flexible. It can support applications that need more than a T1`s level of bandwidth.
If a line failure occurs, critical applications can be quickly switched to whatever T1 lines are operating, because the combination of lines appears as a pool of capacity. This concept is crucial to ATM technology, whose efficiency depends on being able to use the entire trunk as a single circuit, not as a parallel series of channels.
Multimegabit inverse multiplexing should also eliminate some hurdles typical with attempts to justify broadband networks. Some network planners might find it difficult to jum¥from a T1 network directly to a T3 network. Multimegabit inverse multiplexing allows migration to occur in multiples of T1 to the point where T3 access becomes less expensive. User applications "see" a communications channel with a 1.5-, 3-, 6- or 9-Mbit/sec capacity, and not a disconnected series of 1.5-Mbit/sec channels. Given a critical local area network internetworking application, a single connection of 2 Mbits/sec is no longer a problem.
The third rule in network planning is to eliminate multiple parallel data networks based on time-division multiplexing facilities. Many companies have locations where LAN internetworks and systems network architecture terminal networks exist in parallel, each supported with at least one DS-0 on a private T1 backbone. But data traffic is likely to waste bandwidth because it usually contains a large number of idle periods.
Many network planners expect to migrate to an enterprise LAN internetwork structure to transport both data-center and LAN traffic. This approach would combine all data traffic onto a single connection--one or more DS-0s on a private backbone. A router would then connect this enterprise network at each site to the broadband multiplexer.
An orderly way
The broadband multiplexer must support T3, Sonet and ATM. It must also provide direct user serial interface support and the key interfaces associated with other information forms. The multiplexer would thus allow network planners to move quickly from one broadband service to another or to employ several services in the same network.
A broadband strategy should be based on:
Multiple sources of information of various types, multiplexed onto T1 or T3 trunks by existing equipment
Data equipment with T1 interfaces rather than serial interfaces
Data equipment with serial interfaces, particularly V.35 or high-speed
Voice services from analog or digital (T-carrier) interface sources
Generally, preliminary broadband networks would have most traffic generated from existing equipment multiplexing multiple sources of information onto T1 and T3 trunks and data equipment with T1 interfaces. The T1 multiplexers and native T1 interfaces might have been installed to support earlier T1 private networking, and these investments may not yet be written off.
In this case, network planners should consider the following actions:
Use multimegabit inverse multiplexing of T1 private transmission or carrier services through a broadband multiplexer as a first step. Retain the current T-carrier channelized structure in the network for now. Without significant native data interface support, ATM cannot be efficiently utilized.
Cease buying T1 multiplexers and T1 interfaces for data devices; terminate new applications on the broadband multiplexer directly.
If data applications are expected to grow to the point where they may require peak bandwidths over T1 speed, ensure the router or other data equipment used in these applications support the ATM data exchange interface.
As a company grows toward full T3 usage, more of its existing T-carrier equipment and interfaces should be displaced by direct connections to the broadband multiplexer. Network planners can then assess the broadband strategy on which to base the next ste¥in evolution. The key issue is whether to continue to grow within the basic T-carrier structure (T3 or Sonet) or to make an early commitment to ATM.
The main benefit of ATM is bandwidth utilization efficiency. ATM is most useful where the user has applications that have radically variable demands for network capacity. Today, only data can generate this kind of variability. Consequently, network planners must make sure data is isolated, in pure data interface form, at the point of broadband concentration. Data encapsulated in T1 is lost to ATM in most cases. The only exception may be a T1 connection that is dedicated to non-channelized data.
For data to justify the implementation of ATM, the data peaks must have bandwidth above the T1 level. If many lower-level data connections are combined to the broadband level, individual data peaks and valleys of demand tend to even out, resulting in a flow of information that lacks the kind of variability that ATM supports best. Individual applications that generally burst to greater than T1 bandwidth are likely to be examples of LAN internetworking or channel extension.
In moving to the next evolutionary broadband phase, network planners should consider the migration to ATM. The move is based on the T3 interface (or, if possible, a number of T1 signals), if at least 20% of the traffic is data that can be expected to show peaks above the T1 level. If this level of data activity is not expected, handling data as an element of a T3 or Sonet strategy, via frame relay, is probably a better solution. Data devices should be connected to a broadband multiplexer via a data exchange interface, frame relay or user network interface to minimize the adaptation load on the multiplexer and to extend ATM management visibility to the source device.
Consider ATM in Sonet OC-3 form only if the pricing of OC-3 access is favorable. There will be a capital equipment cost increment moving to OC-3; at this early stage of migration, it is doubtful that this much bandwidth can be consumed.
Note that ATM can be offered either as a private network service based on multiple Tl, T3 or Sonet public carrier trunks, or consumed as a public carrier service. The decision on which one to use must be based on cost, availability and the possibility of using T3 or Sonet to carry a mixture of ATM and time-division multiplexing communications.
Migrate to Sonet at the OC-1 threshold at the sites where it is available. Virtual tributaries can be used to support both pre-existing T1 applications and voice, video or data applications connected directly to the broadband multiplexer.
Use of ATM
The use of ATM on the premises would introduce the third phase of broadband evolution. At this point, the network planner is committed to consumption of ATM (or the decision to employ it at the premises probably should not have been made) at the carrier level.
The following planning points would then apply:
Network trunking should be at the Sonet OC-3 rate over fiber-optic cables, either alone or with ATM.
Some portion of the fiber trunks may be allocated to virtual tributary transport, provided the broadband multiplexer and the carrier support this level of integration.
Data devices of all types, but particularly routers and hubs, should be adapted to consume ATM services via a data exchange interface, frame relay or user-to-network interface.
The broadband multiplexer should be equipped with video interfaces whose ATM adaptation capabilities are compatible with the video standard used for deskto¥video. This will ensure video compatibility throughout the enterprise. u
Paul Zalloua is product marketing manager at T3plus Networking Inc. in Santa Clara, CA.