DSL: Does it make sense for small telephone companies

DSL for Small Telephone Companies

Joe Callahan

Manager, Data Network Engineering

Engineering Associates, Inc.

Does it make sense for a small telephone company to offer high-speed data services using DSL technology?

Naturally, the answer to this question will depend on a combination of factors that are unique for each company. Here we will present critical factors to be considered in answering the question for your organization. Following a brief background on DSL and how it works, we consider reasons you might or might not want to add it to your offerings. Then we provide pointers for using DSL to compete with cable modems and CLEC offerings, and finish with a list of questions that need to be considered as you decide whether DSL is right for you.

Characteristics of DSL

Digital Subscriber Loop (DSL) service offerings are available in many "flavors"; some standards-based and some proprietary. This whole collection of possibilities is sometimes referred to as xDSL.

The various implementations offer speeds from 128 Kbps to 52 Mbps on voice-grade copper cable pairs.

Some provide symmetric data rates (same bit rate in both directions) and some are asymmetric, with much higher speeds downstream (toward the end user location) than upstream. Nearly all provide full-duplex service (simultaneous traffic in both directions), but at least one proprietary product is half-duplex while another advertises a data rate that, in fact, is adaptively split between downstream and upstream traffic.

Some versions are "rate-adaptive", testing to see what data rate the line will support and running at that speed. Customers served with identical equipment configurations will not necessarily get the same performance.

The data connection is always "ON". There are no call set up delays, and the data traffic does not use any of the switching capacity of the voice network.

Components of a DSL Installation

CPE

The active CPE is a modem (See diagram), which connects to the phone line on one side and to an Ethernet port (not the serial port that dial-up modems use) on the customer equipment. That customer equipment can be a single computer or a network device that is part of a customer LAN. Some CPE vendors offer modems integrated with routers that have security and management features particularly appropriate for connecting small LANs to the Internet. Some include multiple Ethernet ports so that the modem can serve as a hub for a small LAN at the customer site

DSLAM

DSL itself only works over limited distances on a copper pair. Everything upstream of the end of that copper that is in or closest to the CO must be carried on some other digital technology – DS-1, DS-3, Fast Ethernet, etc. The active equipment at the upstream end of the copper pair consists of (1) a group of modems that correspond to the modems used on the customer sites, and (2) a multiplexer that aggregates the data traffic from these modems onto the high-speed uplink. This combination is called a DSL Access Multiplexer (DSLAM).

Initially, the industry concentrated on developing DSLAMs for high-density requirements in COs. Now it is possible to find a few DSLAM products that are designed for the harsh environments, small spaces and limited power available in many remote equipment shelters. These small units are sometimes referred to as mini-RAMs (Remote Access Multiplexers).

Providing DSL to customers served out of remote sites requires a connection from the remote site to the CO that is separate from the one that carries voice, and might have higher bandwidth. The bandwidth provided here is one of several factors that have a major influence on the performance perceived by the customer. The DSLAM is effectively a statistical multiplexer. It takes advantage of the fact that only a few users are likely to be moving traffic at any given instant, making it possible to handle all their traffic on an uplink with much less capacity than the sum of their individual data speeds. But this concept has limitations. You might be able to use a single DS-1 to provide completely satisfactory performance to 24 users of a service advertised as "1.5 Mbps downstream". But some versions of DSL permits downstream speeds on the copper pair of up to 8 Mbps, and if you offer such a service you obviously cannot get by with a single DS-1 to deliver it. The uplink capacity requirements of DSL can come as a shock to people accustomed to supporting voice at remote sites. The availability and cost of uplink capacity from remotes to central sites has to be carefully considered when choosing the speeds that will be offered and the prices that will be charged.

Central Site Concentrator

At some central telephone company site, the uplinks from all the DSLAMs converge on what can be called a DSL concentrator. Often it is a larger DSLAM, which aggregates all the traffic onto a still higher speed uplink directly to a single Information Service Provider (ISP). Much more sophisticated arrangements are possible, and are necessary if the customers are to have the option of connecting to more than one destination. The most advanced concentrators include high performance routers and integration with back office and network management systems.

Effect of Standards

It is worth mentioning an effect that the development of standards-based technology has had on the market for DSL products. While several vendors still do offer proprietary product lines, and some offer standards-based lines that include all the elements of a DSL network, there are several important vendors who provide only the Customer Premise Equipment (CPE), or only the equipment for telco sites. One might expect that any standards-compliant CPE should interoperate with any standards-compliant telco site equipment, but there is enough looseness in the standards that products can be fully compliant and still have noteworthy differences. As a result, some combinations of standards-based products interoperate better than others, and the manufacturers list the counterpart products with which their own products are known to work well.

Connection to the ISP

The uplink from the DSL concentrator to the ISP is another place where the bandwidth needed to meet customer expectations can be surprising. However, skimping can quickly undermine the reputation of the service. It is also a place where you can provide different service levels for different prices, such as residential grade and business grade uplinks.

It is important to be sure that the ISP understands what is necessary here, too. DSL user expectations probably cannot be satisfied with the same ISP technology and links to the Internet that have been sufficient for dial-in subscribers. DSL users are very different from dial-in users, not only in the speed of their connections, but also they are always connected. To a considerable extent the maximum load from dial-in users is limited by the number of modems. If too many dial-in users try to connect at once, some get busy signals and are effectively blocked from generating any load on the ISP’s connection to the Internet. With DSL, there is no shared modem pool to limit the number of simultaneous users. Every DSL customer is able to generate Internet traffic all the time that his or her computer is turned on. Peak demand will be a function of the total number of DSL subscribers, not some artificially limited subset of them.

Several product lines enable users to turn on and turn off connections to multiple upstream destinations. For example, a telecommuter could simultaneously connect to a corporate LAN and to a commercial ISP. Some carriers offer this as a premium service, and normally just "nail up" a connection from each user to a single destination the user selects when the service is ordered. This multi-destination capability is more likely to be valuable in metropolitan settings with many ISPs and some large corporate installations.

Many of the DSL product lines use ATM all the way to the CPE modem. One or more ATM Permanent Virtual Circuits (PVCs) carry traffic from the customer site, through the DSLAM to the concentrator. That concentrator might be an ATM switch, and the ATM PVC often goes all the way to the ISP access router. These PVCs, or L2TP tunnels to which they are sometimes mapped, are among the possible mechanisms to insure the privacy of traffic on the shared bandwidth, and to permit a user to have simultaneous connections to multiple upstream destinations.

Which Version of DSL?

There is a mix of standards-based and proprietary product lines on the market. Although the proprietary versions have some very attractive capabilities, home computer manufacturers are starting to produce systems with built-in DSL modems that are standards-based. This could push telephone companies using proprietary versions of DSL to parallel or replace them with standards-based products to meet customer demands.

The most talked about standards-based versions (ADSL and G.Lite) can run on the same pair that carries existing voice service. Some vendors offer versions that can operate on the same pair that already carries ISDN. To achieve this shared use of the voice line, the voice and data signals are separated at each end of the copper pair by filter, called a "splitter" in this market. The cost of the truck-roll to install this filter on the customer premises has been a major concern to telephone companies, and has led to new approaches that will be covered below.

ADSL

ADSL is the more mature of these. Most of the products using this standard are rate-adaptive, and one often sees the acronym RADSL use to highlight this capability. Vendors advertise speeds up to 8Mbps downstream and 640 Kbps upstream. Under ideal conditions it can deliver those speeds out to about 12 kilofeet (KF) on non-loaded, 24 gauge copper. That distance is reduced if the copper is 26 gauge. The speed drops off to as low as 300 Kbps downstream and 64 Kbps upstream when the length of the loop gets to 18 KF. The emerging versions of ADSL are supposed to be largely immune to the presence of bridged tap on the copper pair, but it can be problem in some cases. This wide range of performance under various conditions explains why the phrase "up to" is so heavily used in the product descriptions. It probably also tells you that you will not be able to offer ADSL service to some of your customers because the plant delivering their voice service will not support it.

G.Lite

G.Lite is a newer standard for customer premise equipment (CPE) that runs at a maximum of 1.5 Mbps downstream and 512 Kbps upstream. Furthermore, it does this in a manner that makes it interoperable with ADSL equipment at the upstream end of the copper pair. The design goal was to eliminate the need to install a splitter at the customer site by offering a version of DSL that was less sensitive to the problems that the splitter was meant to prevent. If everything worked as planned, this would make it practical for consumers to buy DSL modems on the open market and do the installation themselves, with only a call to the telephone company to activate the service. The unfortunate, belated learning by the industry is that many of the cheap telephones found in customers’ homes have such undesirable electronic characteristics that even G.Lite cannot work well with them directly connected to the voice line. To alleviate this, G.Lite proponents recommend the use of a customer installable "microfilter" on each phone being used at the customer site. They insist that this still qualifies as "splitterless" because it does not involve technician going to the customer site to install a splitter. A recent study of customer self-installation of G.Lite modems by a relatively technically knowledgeable user community reportedly found that over 30% still required on-site assistance to get the service working properly. The percentage will probably be higher for installations by technically naive customers.

All the effort to avoid truck-rolls has ignored the fact that, if the customer gets voice service from a remote site, there still will have to be a technician dispatched to cross-connect the customer’s current voice service into the DSLAM.

A second goal of the G.Lite design was to enable telephone companies to support both ADSL and G.Lite customers with just ADSL modems in the DSLAMs, and at this they have been successful..

The home computer industry has been enthusiastic about the G.Lite approach, and it is the standard to which they are building their built-in DSL modems. This could produce a large number of disappointed computer buyers, given the technical limitations already described and the tough business choices telephone companies have to make about where and when to make DSL service available.

Motivations

What pulls or pushes a telephone company to consider DSL, or to avoid it? There are several categories.

Opportunities

Power users and network game players want high performance.
Telecommuters and small office/home office (SOHO) workers see it s a productivity booster.
Business travelers working from hotels and resorts are willing to pay extra for the vastly faster connections to home office networks.
Real estate developers and apartment or business property owners want to offer high-speed networking as a business differentiator.
Or you might want to operate as a data only CLEC in another ILEC’s territory.

Drivers

Dial-up data by heavy users is overloading the CO switch.
The local CATV operator is about to offer cable modems.
A CLEC has targeted your territory.
The power company is planning to sell power line based data communications services.
The municipal government has installed fiber for its own use and now wants to get into the network services business.

Concerns

The market is untested. There might not be the critical mass of customers to make it worthwhile.
You will not be able to provide the same service to all customers.
Offering it at prices that compete with cable modems might cannibalize ISDN and T1 revenue -- but competitors might have that effect if DSL doesn’t.
There is much more cost to consider than just the hardware. There are order administration and engineering processes to develop and implement; administrative and technical staff training to do, and new customer support requirements, to name a few.

Competitive Considerations

You probably have an advantage in terms of knowing and being known by your customers. But competitors might be more experienced with rules and techniques of competition than you are. What advantages can you exploit?

Cable Modems

In addition to the poor reputation that some CATV operators have for reliability and time to repair, they have some serious technical hurdles to overcome.

CATV service often is non-existent in the business buildings where many customers for high-speed data services are found. Adding them to the cable network would be very expensive. The wiring for DSL is already in place.
Where the cable system has not yet had the expensive upgrade to two-way operation, the cable modem user’s upstream traffic continues to use a dial-up modem. This gives a low speed uplink, it still has the inconvenience of connection set-up delay, and it still ties up the phone line whenever the cable modem is in use. DSL is faster, always connected, and leaves the voice line free.
The performance of cable modem systems will vary, depending on how many users are attempting to move traffic at the same time. The larger the group of customers served on a single head-end modem, the worse this problem is. Reducing the number of users on a head-end modem takes additional fiber feeds from the head-end out to the fiber/coax interfaces (optical network units, ONUs), and additional expensive ONUs. Delivering high quality cable modem service takes more than upgrading the existing cable plant to two-way operation. This is different from the need for adequate bandwidth from the DSLAM to the concentrator, mentioned above, because it can involve setting up entirely new ONU locations to make it possible to subdivide the coax part of the cable plant in the manner necessary to support a high performance installation. DSL already uses dedicated pairs to each subscriber in this part of its physical plant, providing dedicated bandwidth in this part of the network. Upstream of the cable system head end, the network congestion and performance characteristics of a cable modem operation is very similar to that of a DSL service upstream of the concentrator.
The cable modem industry is just beginning to deliver products that keep a user’s traffic private. The nature of the technology makes the traffic to and from one user available to all his neighbors systems, too, if they take the trouble to listen. In some poorly configured systems, even the files on a user’s PC were accessible to other cable modem users. New standards will incorporate encryption to eliminate this exposure, but these are just beginning to come on the market and the problem will persist wherever pre-standard modems are used. DSL uses dedicated copper pairs and ATM (or IP tunneling) to insure privacy.
Much of the cable industry has adopted a business strategy that forces every cable modem subscriber in a given territory to use one particular ISP service that has preferential business ties to the cable system operator. Many users resent this and would be attracted to a service offering that let them choose among several ISPs. The large telephone companies have been barred from such restrictive behavior, so most new DSL products include the ability to provide connections to several different destinations. (Some cable modem product lines already include a similar ability. This advantage could disappear if it starts to hurt cable modem sales, or if regulators force cable operators to abandon their current practice.)

CLEC

Wireline

A wireline CLEC coming in to your territory has some obstacles to overcome. Hottest among these are the regulatory issues.

There is the question of having to provide Unbundled Network Elements (UNEs) to CLECs. Are you exempt, and can you keep that exemption to prevent incursions?
If you are not exempt, can the CLEC demand shared use of current voice pair, or can they be forced to use and pay for an additional pair? This issue is a regulatory hot potato at the time this paper is being written.
You might have a practical advantage in terms of being able to replace line cards in existing chassis, while the CLEC would have to install complete electronic packages.
They will be dependent on you for co-location space in COs and remote shelters, plus power within and uplink bandwidth from those remotes. If you happen to not have any spare capacity in any of those areas, you are not required to build them just for CLEC use.

Wireless

DSL has some distinctive advantages over wireless solutions.

Essentially all of the wireless high-speed data technologies require line-of-sight between antennas in order to operate. DSL is not limited to line-of-sight opportunities, and does not involve any antenna installation at the customer site.
DSL is not affected by weather conditions.
DSL does not require you to have purchased rights to use the necessary frequencies in your market, or to go through the licensing process that is necessary for some wireless frequencies.

Question

These considerations produce an extensive set of questions to be addressed as you assess the opportunity and develop your business plan.

What services do likely customers really want -- speeds, symmetry, single or multiple destinations? Do the users want to have more than one computer connected at a site?
Will individual customers order the service from you, or from the ISP as a reseller of what you offer? Are there corporations that would want small offices or individual employees to have DSL connections to the company LAN? Are there hotels, or conference centers, or office buildings ,or apartment complexes that would offer the service to set themselves apart in their markets? Who will bill whom for what? If something is not working correctly, who does the user call?
Does the service need to be standards-based to (1) support DSL modems built in to newer personal computers, and (2) enable you to offer the service without having to be the sole source of modems?
Is user demand strong enough to make them willing to pay enough to make it profitable? How does the price you will need to charge compare to competitive offerings? Do you offer distinctive benefits, such as access to the customer’s choice of ISP or better security, that will it possible for you to charge higher prices than your competitors?
Where are likely customers relative to the technically feasible service areas? How many customers will it take to warrant offering the service out of a typical remote site?
Do you have spare space and power in your remote sites? Does your current DLC vendor offer an easy solution, or do you have to add a new chassis in each remote shelter? Will the DLC vendor’s solution satisfy the technical requirements?
Do you have the spare bandwidth for uplinks from your remote sites to the concentrator location?
Do you already have or plan to use ATM? If not, are you willing to start using it as part of establishing DSL service?
What uplink technologies and line speeds will the ISP(s) support – PPP on DS-1, ATM on OC-3, …? How will user authentication and address allocation be handled?
How fast will the uplink to the ISP need to be? Can/will the ISP provide sufficient performance on his uplink to the Internet? What will your ISP uplink line and access arrangement cost?
How much change and training will be needed in administrative processes, in engineering and operations, and in installations? Is the CPE really user installable?

Summary

Deciding whether and how to begin to offer DSL services to your customers is more complex than estimating the cost of the new hardware.

There are significant motivating factors that either attract or push you to offer high-speed data services: customer demand, dial-up data load on the CO, and imminent competition.

At the same time, there are many technical factors that govern your ability to deliver DSL services, and financial ones that can challenge your ability to make a profit on services that customers are willing to buy.

Return to EA Home Page