Cover V09, I06
Figure 1
Figure 2
Figure 3


Convergence Economics

Gilbert Held

In two previous articles, I examined the rationale for convergence and techniques for tuning your network to make the technology work. This last article in this series covering convergence of voice and data transmission onto a common network infrastructure focuses upon economics. Although one of the primary goals of convergence is to reduce the cost of communications, not all methods associated with the transmission of voice over data networks result in economic savings. As shown in the two examples presented in this article, in certain situations it is more economical to simply pick up the telephone and make a call over the existing Public Switched Telephone Network (PSTN) than attempt to integrate new equipment into your network infrastructure. As you might expect, there are other network situations where the economics associated with convergence can be quite appealing and justify the effort involved in making it a reality.

Making the Dime-a-Minute Lady Blush

When I teach courses covering the convergence of voice and data, I sometimes mention Candace Bergen, who appeared in several Sprint Corporation advertisements for Sprint's 10-cents a minute long-distance plan. During those advertisements, Ms. Bergen became known as the Sprint Dime-a-Minute Lady. Even though Sprint now offers “nickel nights” during which the cost of a long-distance call is 5 cents per minute, the convergence of voice and data under certain situations can result in long-distance costing less than a penny a minute. As I like to remind folks, at this rate for making a long-distance call, even the Dime-a-Minute Lady would blush!


Illustrating the economics associated with the convergence of voice and data over a common network infrastructure requires a network, the use of certain types of networking equipment, and a set of assumptions concerning the use of the network and networking equipment. Because voice over Frame Relay is currently a more practical technology than Voice over IP, I will focus my economic models on the use of a Frame Relay network.

There is currently no standardized and implemented method to provide quality of service (QoS) over the Internet. Although the Resource ReServation Protocol (RSVP) has been standardized for several years as a mechanism to allocate bandwidth, it does not scale well for large use and is currently restricted for use on private TCP/IP networks. Although several relatively new technologies (including Differentiated Services (DiffServ) and Multi-Protocol Label Switching (MPLS)) were developed to enhance the flow of traffic through the Internet, numerous issues remain unresolved. One of these issues is how different Internet Service Providers (ISPs) bill one another for third-party expedited traffic flow through their network. In comparison, in a Frame Relay networking environment, various service level agreements (SLAs) are available that guarantee end-to-end latency through the networks. When coupled with the ability of an end-user to select a Committed Information Rate (CIR) for which the service provider guarantees the ability of frames to flow end-to-end without being dropped, it becomes possible to configure a QoS for voice being transported via a Frame Relay network. Now that I've explained why voice over Frame Relay is a reality, while voice over IP is a hit or miss proposition, I'll return to the focus of this section and develop a network infrastructure as a foundation for performing an economic analysis.

For the first networking scenario, I will assume the organization does not currently use Frame Relay, but was told it would be a good idea to transmit voice over that network since it could result in significant economic savings. Figure 1 illustrates the use of two Frame Relay Access Devices (FRADs) installed at two different company offices connected together via the transmission facilities of a vendor's Frame Relay network. Because the use of almost all Frame Relay networks are distance insensitive, it doesn't matter if the two locations are New York City and Newark, or Bangor and San Diego.

Assume that each FRAD is acquired to support up to four voice calls via a connection of the FRAD to four PBX E&M ports. Also assume that each FRAD is connected to the point-of-presence (POP) of the Frame Relay provider via a T1 access line for which we are running at a fractional T1 speed of 128 Kbps and for which the cost is $625/month.

In addition to billing for the access line, the Frame Relay network provider will also bill us for the port on their switch at the POP and for any private virtual circuits (PVCs) configured between the two locations. The monthly cost associated with the use of a switch port is based upon the operating rate of the port. Because a T1 line terminates at each switch port, we would be billed a monthly fee based upon the operating rate of the T1 line even though the actual transmission will be a maximum of 128 Kbps. Assume that the monthly cost of each switch port is $415.

Because I want to allow up to four simultaneous calls, I would ask the service provider to configure four PVCs between locations. Assuming that I ordered four PVCs, each with a 32-Kbps CIR, I would incur an additional $20/month cost for each PVC that would be added to the organization's monthly bill. Last, but not least, I need to consider the cost of each FRAD.

If the basic cost of each FRAD is $6000 and if the life of the FRADs are four years prior to obsolescence, then divide their cost by 48 to obtain an approximate monthly cost. Since I will perform a similar computation in a second economic example, my consistency in not being precise will result in a valid comparison of two networking scenarios.

Based upon the prior cost assumptions, the monthly cost of each FRAD becomes 6000 divided by 48, or $125, while the port cost at each end of the network is $415/month. The monthly cost of the PVCs at each end is $20 times 4, or $80, while the monthly cost of the access line is $625. Thus, the total cost for each end of this converged network becomes 2 times (125 + 80 + 415 + 625), or $2490 per month.

Computing the Cost Per Minute

Because long distance is normally billed on a per-minute basis, the conversion of the previously computed monthly cost to this familiar metric provides a method of comparison to easily work with. To do so, I need to make a few more assumptions. Since I am attempting to convert a monthly cost into a cost per-minute, I need to project the usage of the PVCs that will be used to transport voice calls over the Frame Relay network. I first need to determine how many days per month the configuration shown in Figure 1 will be used.

In a business environment, a good rule of thumb is to assume there are 22 business days per month. This is a reasonable metric to use since the usage of FRAD to transport voice over Frame Relay through corporate PBXs would primarily occur Monday through Friday. I can start by assuming each PVC is used one hour per business day and that a business day represents a normal 8-hour period. Thus, I will assume that the voice ports on the FRADs are used 12.5 percent of the business day. Because there are four ports on each FRAD, then the network configuration shown in Figure 1 can support 4 hours of traffic per day, or 88 hours per month when each port is used 12.5 percent of the business day.

Based on the previous assumptions, I can convert the 88 hours per month of projected voice activity into call minutes. Doing so, I obtain 88 times 60, or 5280 call minutes. Thus, if I divide the monthly cost by the usable call minutes, I can project the cost per minute for using the configuration shown in Figure 1. I obtain the following:

$2490/month = 47.2 cents/minute
5280 call minutes
If you consider the results of the previous computation, it is not very appealing. In fact, unless you are considering locating a FRAD in Ulan Bator, Mongolia, or another exotic location, it is probably more economical to simply pick up the telephone and make a call via the PSTN. However, before I jump to a conclusion concerning the viability of using a Frame Relay network to strictly transport voice, I'll examine the effect of an increased level of call volume.

Figure 2 illustrates the effect of increasing the occupancy of the four FRAD ports from 12.5 percent to a full occupancy of 100 percent. In examining the cost per minute computations shown in Figure 2, note that as the occupancy of the FRAD ports increase, the cost per minute decreases. In fact, if I can keep the ports half occupied during the business day, the per minute cost will decrease to slightly less than 12 cents, while at full occupancy the cost is less than 6 cents per minute.

While any daytime business cost less than 7 cents per minute may represent a good deal in today's cost environment, at the speed that long distance rates are falling, you may want to consider the viability of this cost several years down the road. Additionally, to obtain this rate, you must keep the voice ports relatively fully occupied during the business day. Before jumping to a conclusion about the cost associated with the networking configuration illustrated in Figure 1, I'll examine a slightly different scenario.

Examining the Incremental Cost of Voice

For the second networking scenario, assume that the cost associated with obtaining a pair of FRADs, a pair of access lines, and a pair of switch ports was incurred due to an existing use of Frame Relay to transmit data between two locations. If the FRADs are modular devices that are upgradeable, I might be able to simply purchase four-port voice cards for each FRAD. If I assume that each four-port voice card costs $1500, then the equipment expense is reduced to $3000. If I again divide the cost by 48 months, I obtain a monthly cost of $3000/48, or $62.50. Since I still need an additional PVC for each PBX to PBX transmission that will flow over the Frame Relay network, I must add $20 per PVC or $80 for four PVCs to the $62.50 equipment cost. This results in a total monthly cost associated with the incremental addition of voice to an existing Frame Relay network of $142.50.

I can now perform a similar series of computations to determine the cost per minute as the occupancy of the four ports on the FRAD increases from 12.5 percent to 100 percent. Figure 3 summarizes the economics of increased call volume when the incremental cost associated with adding a voice capability to the two FRADs is considered.

In examining the cost per minute associated with adding a voice capability to FRADs, note that it starts off at 2.7 cents per minute and decreases as the occupancy of the ports increase to a fraction of a penny per minute. This illustrates an important point that is, the cost per minute is significantly less when I add voice to a network infrastructure currently used for data transmission than when I establish a network infrastructure to transmit voice over a data network (compare Figures 2 and 3). Although actual results will vary based upon the cost of equipment, the cost of network facilities, and the occupancy level of voice ports, I can expect the incremental cost associated with adding voice to an existing data network to be less than a solo voice over data network solution. While I may not expect to obtain a cost of a third of a cent per minute, from my experience, the actual cost per minute obtained via the use of voice capable FRADs in several network situations was under a tenth of a penny per minute.

To accomplish this economic marvel, one organization used two four-port voice boards in a pair of FRADs and kept the FRAD ports occupied approximately 75 percent of the time. Because the average occupancy was 75 percent, there were periods when the occupancy level reached 100 percent. The organization simply programmed their PBX to use the trunks to the FRAD as a rotary access group when all ports were occupied. Then, when the rotary overflowed due to all trunks in use, the overflow was directed to another group of trunks that provided a connection to the PSTN. Thus, the PSTN became both the backup for the voice over Frame Relay network infrastructure as well as the overflow mechanism to support continuity of operation.

In concluding this series on convergence of voice and data networks, it is important to note that we are in the infancy of a new technological revolution. Although the use of packet networks to transport voice conversations presently represents a very small fraction of all voice calls, we can expect this percentage to significantly increase in the future. Although most current products on the market do not do a very good job at being scalable, they provide the opportunity to become familiar with the evolving technology as well as to achieve a most reasonable level of economic savings under certain network conditions. Due to this, I will close this series by adopting the old Boy Scout adage to the focus of this series -- that is, be prepared for convergence.

About the Author

Gilbert Held is an award-winning lecturer and author. Gil is the author of over 40 books and 250 technical articles. Some of Gil's most recent publications include Voice and Data Internetworking 2ed. and Cisco Security Architecture (co-authored with Kent Hundley), both published by McGraw Hill Book Company, and LAN Performance 3ed. and Internetworking LANs and WANs 2ed., published by John Wiley & Sons. Gil can be reached via email at: