rvices offered by telecom carriers--is undergoing a radical change.
For years, most corporate networks ran on a telecommunications infrastructure that was fairly stable. But in the last four years that all changed: The widespread deployment and acceptance of ISDN, and the growing acceptance of other switched services, includi
ng frame relay, Switched Multimegabit Data Service (SMDS), and even asynchronous transfer mode (ATM), are turning corporate WANs into a kind of telecommunications Loma Prieta.
All this at a time when the organizational structure of corporations is shifting from a setting where most employees work in large central offices to one where employees increasingly work in many smaller, geographically dispersed offices.
The result is a radical shift in the architecture of corporate networks. Four years ago, you'd likely see a private backbone network of dedicated, leased T1 or T3 lines carrying voice and data traffic between, at most, a handful of large regional offices. And you'd see low-speed (9.6- or 19.2-Kbps) dedicated leased lines carrying transaction-oriented data between dumb terminals in smaller offices (a bank branch office, for example) and mainframes in large data centers.
Today it's more likely that this backbone architecture is supplemented with higher-speed links carrying LAN traffi
c from branch offices, dial-up and switched-circuit connections between the various branch offices, and dial-up access to the backbone for telecommuters and employees on the road.
The shift is dramatic. It
signals a change
from the use of private corporate networks where most, if not all, of the traffic ran over leased lines managed by the corporation to a system where most traffic runs over the public switched telecommunications infrastructure.
The Perfect Match
The challenge most organizations face today is choosing the right service for their connectivity needs. Determining what exactly is the most appropriate service is often a juggling act. First, you have to consider performance issues. Does the service provide enough bandwidth for my applications? Does the service let me set priority levels so that my most time-sensitive traffic gets through even when the network bogs down?
Second, you have to consider financial aspects. Pricing for the sam
e service can vary greatly between regional Bell companies. ISDN is the perfect example of this. Users in Southern California have seen fairly low monthly fees and usage charges because of Pacific Bell's aggressive ISDN campaign. Whereas users in the New York/New England region pay a much higher rate for the same service. (For a more detailed breakdown of WAN service costs, see the article
"The Price of WAN Connectivity"
.)
Third, there's the issue of availability. Sure, ISDN might offer the best performance characteristics and be reasonably priced, but if it isn't available in the town your accounting department's located in, you need to reevaluate your choice.
Finally, to select a WAN service you must have an understanding of the traffic that will be running over a link. Basically, you need to predict the traffic volume for each link, the mix of protocols that will run over the link, and whether any of the traffic is time-sensitive.
The Backbone
There are bas
ically four connectivity scenarios you will need to analyze: the backbone network, LAN-to-backbone connectivity, LAN-to-LAN connectivity, and user-to-LAN connectivity. Each scenario has characteristics that will determine the best service to choose.
Backbone networks link large regional centers of an organization and typically carry both voice and data between the sites. Backbone networks have traditionally been built around dedicated leased lines that operate at either 1.544 Mbps (T1 speeds) or 44.736 Mbps (T3 speeds).
The backbone is a private network managed by the company using it. A manager can allocate up to 24 64-Kbps channels for each T1 line and up to 673 64-Kbps channels for a single T3 line (the equivalent of 28 T1 links). The way you allocate these channels depends on the needs of your company. For example, within a single T1 connection you might set aside four 64-Kbps channels for voice, 768 Kbps (12 64-Kbps channels) for videoconferencing, and 512 Kbps (eight 64-Kbps channels) for
a data connection between mainframes in different sites.
To get traffic onto the network, you use a time-division multiplexer in each site. It aggregates numerous voice and data channels into one stream of traffic that goes out over the link. A device called a channel service unit (CSU) connects the multiplexer to the actual phone line.
T1 and T3 circuits are still the most commonly used circuits connecting large regional centers to form a core backbone network. But there are alternatives to consider. Some companies use frame relay or SMDS to meet their backbone connectivity needs. Currently, frame relay is one of the hotter technologies.
With frame relay, data travels between two sites over a logical link called a virtual circuit. There are two types of virtual circuits. With a permanent virtual circuit (PVC), you define the path and the end points of a link once and they always remain the same. With a switched virtual circuit (SVC), as the name implies, the frame-relay hardware assemble
s and tears down the logical link between two points for each transmission.
You can use public or private frame-relay circuits to build your backbone network. You order service by specifying the minimum guaranteed throughput you would like over a particular PVC. This minimum figure is called the
committed information rate
(CIR).
One of the appealing factors of using frame relay is its flexibility when it comes to designing a network. You can select a CIR ranging from 64 Kbps up to T1 rates; this range of speeds makes sure you have some room to maneuver when trying to match the actual performance needs of a connection with a telco service offering.
Another appealing thing about frame relay is that it accommodates bursty traffic, such as the traffic generated by your typical LAN applications. Even though you order frame-relay service based on the CIR, traffic loads above this amount can be accommodated (at no extra charge) if the telco's network has the capacity to handle the traff
ic. (How often and how long you can exceed the CIR is something that must be negotiated with the carrier before the service is installed.)
For the most part, frame-relay networks support data only. However, a number of companies run their voice traffic over their private frame-relay networks. And it is even possible to carry voice traffic over a public frame-relay network (see
"Voice Gets Framed,"
February BYTE).
Similarly, companies have traditionally used SMDS for data. It's a public high-speed packet-switched service aimed at interconnecting LANs at T3 rates. But it, too, can carry voice if you have the proper equipment. For example, some multiplexers allow you to combine voice, data, and video traffic so that it can be carried over a single SMDS link.
One of the great attractions of SMDS is that it's the only high-bandwidth public net available today (at least until ATM becomes more widespread). That makes it attractive for companies that have to send large files (for example
, an automaker zapping CAD files to a parts manufacturer). If both companies use SMDS, they can always communicate through the public network without the hassle of setting up private lines.
ATM is another possible service to consider for a backbone technology. But today, it's availability is fairly limited. That could change in the next few years.
Branching Out
Often the choice of a backbone WAN service dictates the service choices for connecting branch offices. For instance, if you're using frame relay for the core network, you'll likely use frame relay to bring in the traffic from your branch offices. But there is a degree of flexibility.
In a typical LAN-to-backbone scenario, you will have to support multiple networking protocols, handle time-sensitive traffic to and from legacy systems, such as IBM hosts, and support a high volume of LAN traffic.
Many private networks use fractional T1 service between sites. As the name implies, this service is offered in fra
ctional units of a T1's link--usually in increments of 64, 128, 512, or 768 Kbps. Typically, you would select the bandwidth based on your peak traffic loads between sites. That's the safe way to do it. But it can also be uneconomical. You might exceed the peak traffic only once a week (when everyone logs into the system at the same time Monday morning, for example).
The ideal solution would be to select your service based on average loads and somehow get some extra bandwidth when it's needed. Frame relay does that to some extent, allowing you to exceed your CIR when you have a burst in traffic load.
But if you are using leased lines between sites, there's another way to design your networks to save money:
Use switched
services to supplement leased lines. Suppose your average traffic loads are below a T1 line's capacity and your peak loads are somewhat above that. Using traditional communications options, you might opt for a second line just to be safe.
A better choic
e would be to use a T1 link that satisfies your average traffic loads and then simply add bandwidth on an as-needed basis using dial-up ISDN lines. It's simple: Once it determines that a leased line link is saturated, a router in a branch office simply brings up an ISDN connection to the main office. This process is commonly called bandwidth on demand. (Most major router vendors have built bandwidth-on-demand features into their router management software.)
Using ISDN like this gives a link some elasticity to accommodate peak traffic. Most routers can add a single 64-Kbps B channel. To add more bandwidth requires an inverse multiplexer, which allows you to aggregate the bandwidth of additional B channels into a single higher-bandwidth link. For example, with an inverse multiplexer you could easily merge the two B channels of a Basic Rate Interface (BRI) ISDN connection to form what appears to be a single 128-Kbps pipe to carry videoconferencing traffic between two sites.
Another way you can use
ISDN connections with existing leased lines is as a backup link in case the main link fails--a feature commonly called
dial backup
. This, too, can save money on WAN service charges. For example, many companies install redundant leased lines between sites in case the primary one fails. Typically, the secondary, or backup, line remains idle unless there's a failure. All the while, you're paying monthly charges for this additional line. The way dial backup works is that once a branch-office router detects a failed link, it brings up an ISDN connection to the main office to carry traffic.
Additionally, you can use ISDN in LAN-to-backbone connectivity as the sole connection. Basically, an ISDN router attached to a LAN in a remote site dials up a connection when it receives traffic destined for a LAN in another location.
If you want to connect a number of smaller offices to a regional office using ISDN, you can equip the regional office in one of two ways. You can have multiple ISDN BRI lines
coming into the facility. Or you can choose Primary Rate Interface (PRI) service for the central site. This way the carrier multiplexes the signals coming from the various sites onto a single access line into your facility.
Keep in Touch
Connectivity needs are different when linking branch offices to each other. They're so different that you should pay special attention when selecting the service you'll use. For example, it is usually impractical to have dedicated links from each office to all other offices. Such meshed networks were common when you needed to link only a handful of large offices. But today it's quite common for a large organization to have hundreds of smaller offices. Instead, the more typical pattern now is a star structure where smaller offices feed into larger data centers. In this scenario, a person in Boston might send a file to a coworker in Pittsburgh by way of the company's headquarters in New York. But there are often times where it makes sense to let two
branch sites connect directly.
Determining whether the link between two smaller offices should be dial-up or dedicated is not a simple matter. You pay a fixed monthly rate for a leased line. You can fill the link to capacity for every second of the month or not send a single byte of data over it and still pay the same amount. With dial-up ISDN, you typically have a small set monthly fee and a usage fee based on the time the connection is up.
There are no hard and fast rules of when to use a dedicated line versus when to use an on-demand dial-up connection--it largely depends on the pricing of the service. A general rule of thumb is that if you need more than three to four hours of connectivity per business day, you should use a dedicated line. Anything below that and you can get by with a dial-up line. Most routers support a function called dial-on-demand for this type of LAN-to-LAN connectivity between smaller offices.
For many years Switched 56 service was a big hit. It's a dial-up serv
ice that lets two sites exchange data at 56 Kbps, but it uses a channel service unit/data service unit (CSU/DSU) instead of a modem to place the call. With Switched 56, you pay a monthly fee plus a usage fee that is based on the time the circuit is up. That makes it economically practical for casual connectivity, where one site simply needs to open a link to another site for a short period each day. Switched 56 is well-suited to applications such as videoconferencing and telemedicine where large image files must be transferred.
But the use of Switched 56 is likely to drop off as ISDN becomes more available. ISDN offers more bandwidth (two 64-Kbps channels in a PRI circuit vs. 56 Kbps for a Switched 56 link). And many carriers are pricing ISDN so that it makes better economical sense to use it rather than Switched 56 services.
ISDN is also starting to gain ground in the single user-to-LAN connectivity area. Today most telecommuters and mobile workers still rely on modems and analog phone lines. H
owever, as ISDN BRI service becomes easier to get, and the price for ISDN terminal adapters drops, more telecommuters will switch to ISDN because of its higher speeds.
Get with the Program
The nature of connectivity has changed dramatically in the past year. Corporations are much more decentralized now, and that makes the old paradigm of private networks obsolete. Everyone wants more bandwidth, more performance, and more sites connected, all while paying less for the telecom services.
To accommodate such connectivity needs is no mean feat. Many businesses are using a mix of public telecom services that best meets their performance and economical needs.
The good news is that there are many services to choose from. The bad news is that some of the services are not ubiquitous, or their pricing structure makes them too expensive.
BACKBONE GUIDE
Traffic
Volu
me Type Of Traffic Protocol Mix Recommended Service
Heavy Voice, data, and time- Multiple protocols Dedicated T1 or T3
sensitive data, such such as TCP/IP, leased lines (ATM
as SNA and IPX, NetBIOS, in the near future)
transaction processing SNA
LAN-TO-BACKBONE GUIDE
Traffic
Volume Type Of Traffic Protocol Mix Recommended Service
Moderate LAN data traffic and IP, IPX, some AppleTalk, Frame relay,
some time-sensitive and possibly a small fractional T1,
traffic, such as SNA amount of SNA ISDN, Switched 56
session traffic
LAN-TO-LAN GUIDE
Traffic
Volume Type Of Traffic Protocol Mix Recommended Service
Moderate LAN data from E-mail IP, IPX, possibly Frame relay, ISDN,
to Low exchanges and some AppleTalk analog
occasional file transfers
SINGLE USER-TO-LAN GUIDE
Traffic
Volume Type Of Traffic Protocol Mix Recommended Service
Low to Access to server and IP, IPX (usually ISDN or analog
Moderate host applications, E-mail, only one)
file transfers
illustration_link (12 Kbytes)
High-speed dial-up lines and switched cir
cuits are commonly used to carry LAN traffic from branch offices into the corporate backbone. And analog and ISDN dial-up lines are used to connect mobile users and telecommuters to the corporate network.
illustration_link (7 Kbytes)
Companies can reduce the number of leased lines in their corporate networks (and thus save money) by taking advantage of on-demand dial-up switched services, which allow you to establish and retain a connection only when you need additional bandwidth.
Salvatore Salamone is a BYTE news editor and author of Reducing the Cost of LAN Ownership (Van Nostrand Reinhold, 1995
). You can reach him at
ssalamone@bix.com
.
The New WAN
