Publisher's Letter: Ten Reasons Why IP Will Not Replace Circuit Switching Anytime Soon

The IP Hypemeisters are beginning to concede the point that international and long distance IP telephony services are cheaper than circuit-switched telephony only because of international settlements and access charge avoidance. Now they are refocusing the hype, claiming IP technology itself is cheaper and can create more value through customize services, and, thus will begin to replace circuit-switched technology within a few years. This too is hype. Circuit switching technology for voice services will dominate for a decade or longer, and here are 10 reasons why:

1. It's All About Incremental Costs
The IP Hypemeisters pump out an equipment cost message that goes something like this: a router (e.g. an IP packet switch) costs a fraction of what a time division multiplexer or a circuit switch does. This is an apples-to-oranges comparison, and it's an irrelevant statement when applied to today's cost of carrier service provisioning.

For example, say someone moved into a house or apartment and called the local telco to order phone service. All that's done to initialize service from a network perspective, in almost every case, is that a telco customer service representative pushes a key on a computer. No hardware is added or touched, no memory is added or computer code written. The circuit-switched network is there, and the incremental cost to add subscribers, excluding local loop plant, is near zero. This assumes, of course that the local loop plant is in place.

OK, what about new markets or new services? Not counting the cost of installing more copper loops in the network, everything is economically governed by small incremental costs from the point where the copper or fiber enters the local telco's switch center building, including all long distance network facilities. Yes, it's expensive to install a circuit switch and dig the streets to bury fiber. But once installed, it's highly scalable and incrementally low in cost.

Consider there are roughly 20,000 local switches in the United States today, with an average of 9,000 lines per switch. Roughly 90 percent of these switches are digital, with most having growth capacity to 100,000 lines each. Also, the actual number of local exchange switches over the last decade has decreased. Why? It's cheaper to put in a remote switching module (e.g. slave switch) and link it via fiber to a master switch miles away than it is to put in a new full blown switch.

Even AT&T has stopped putting in circuit toll switches. My understanding is that they have around 150 today, up from 132 four years ago. Why the lack of growth? These switches, as with the other established IXC's switches, are generally under-utilized capacity wise!

The bottom line is that when you compare the cost of an IP packet switch to the cost of a circuit switch in the public network, the incremental circuit switch costs will be relatively speaking near zero for years to come.

2. IP Voice Quality Stinks
In my "Understanding Telecommunications Technologies" seminars, I begin the Internet Telephony section by asking who's aware that the Internet can be used for telephony? Almost everyone raises a hand. Second question: How many people have actually placed a phone call over the Internet? Surprisingly, only a few hands generally go up. Then the IP telephony zinger: How many people who have placed IP telephony calls would use the Internet to call a potential client? No hands go up, but plenty of chuckles can be heard.

Voice Over the Internet (VoIP)

The only people making any money or profit doing VoIP, from a carrier service perspective, are those using the public Internet with access via the circuit-switched or Public Switched Telephone Network (PSTN) to the IP carrier's telephony gateway switch. These IP gateway switches convert telephony signals on the originating end into IP packets and then reverse the process at the terminating IP gateway switch back to PSTN or circuit-switch signals.

Under this configuration, the quality of voice over the Internet (VoN) is poor compared to conventionally completed PSTN calls, but the price to the consumer is lower. To put IP and PSTN telephony on an equal footing, consider eight service quality-to-cost issues, as summarized on page 8.

VoN or PSTN? Cost vs. Quality

A. Transmission: The multiplexing technique of VoN is based on the User Datagram Protocol (UDP). Packets containing digitized voice samples have roughly a 5 percent chance of getting dropped or delivered out of order over the public Internet during the peak usage period (9 a.m. to 11 p.m.). This problem is unavoidable on the public Internet. Also, variable delay occurs, which requires buffering, which in turn causes a "satellite" round-trip delay effect. There are no solutions to these quality problems when using the public Internet for transport. Lower cost, but lower quality. The alternative, Time Division Multiplexing as used in the PSTN for switching and transport, causes almost no significant information loss and no perceptible latent delay unless satellite transmission is used-which is rare for international calls between developed nations or domestic long distance calls these days.

B. CODEC: To economically benefit from VoN, voice signals have to be compressed by using CODECs with 6-8 KBPS sampling rate versus 64 KBPS over the PSTN. Yes, this produces 87 percent or more cost savings, but it also produces a loss in the clarity of the call. Worse yet, coupling lower coder rate quality with packet loss in transmission as identified above, eliminates the capability to send fax or in-band data (e.g. dial up modems), as well as eliminating touchtone signaling for integrated voice response units after call set-up with today's IP gateway switch products.

C. Silence Suppression: VoN uses silence suppression, which prevents voice samples being sent during silent intervals in a conversation or when the other person is talking. This again reduces use of transport resources by roughly 50 percent, but it also reduces voice quality due to clipping of syllables as well as the requirement to generate "silence" (i.e., low-level background noise) at the terminating end, which can sound annoying to the listener.

D. Dial Tones: With today's VoN, a caller has to deal with two dial tones. The first is heard when a caller is connected to the gateway switch. The second is to prompt the caller for authentication ID (or prepaid calling card number) and the terminating party's phone number. This is unnecessary with the PSTN because the originating PSTN carrier, without prompt, sends Automatic Number Identification (ANI) to the long distance carrier for authentication and billing, eliminating the second dial tone.

E. Call Set-Up Delay: A domestic long distance call today has post-dialing delay of 2-3 seconds after the last digit is dialed. This is mainly due to the ringing circuit, which gives 1-2 seconds of ring to the called party before the caller can hear the in-band audible ring. IP calls placed via VoN typically expect a 20 second or more call set-up time. First, a local or toll free number is dialed, then a second dial tone is sent in-band and the caller enters more digits (ID and called party's number). Then, if the IP gateway uses the H.323 standard (ISDN Q931), multiple IP packet exchanges over the Internet are used to establish a session between gateways. Also, gatekeeper access for authentication between ISPs contributes to the delay as well. Once again, VoN is cheap but the user suffers the lack of service quality - the high post-dialing delay.

F. IN Features: A no-brainer here, PSTN has calling cards, 800 numbers, voice VPNs and other features, but at a price. VoN gateway switch IN features are not available today. VoN IN features are free because they don't exist.

G. Access: Today when an IP gateway switch is accessed through the PSTN for long distance service, a 2 cents to 2.5 cents per minute access long distance charge is avoided. VoN access is free to the ISP, other than a local business line charge ($35 per month per line for unlimited terminating calls). Additionally, the call transmission quality is that of a line, not a trunk. Trunk-side transmission is engineered with less distortion than line-side transmission.

H. Termination: The same business case applies here as with originating access: VoN avoids the 2 cents to 2.5 cents per minute access charges, as well as the international termination call settlement fees which can run nearly 50 cents or higher in Asia. Also, it's a line side vs. a trunk connection, so again inferior transmission quality.

Fix the IP Quality and Destroy the Economics

IP zealots at this point in the telephony quality argument generally say, "Yes, the above is true today, but all that will be fixed over time as the technology is developed." I agree, to a point, regarding quality, but improving the quality destroys the cost savings. Yes, IP quality can be improved by adding processing power or by throwing more bandwidth at the problem. Here are three fixes, but they are very expensive:
A. SS7: Deploying an SS7 interface to the PSTN can solve the second dial tone issue, decrease call set up time, permit IN features and permit trunk interfaces at the originating and terminating ends of a call. But this quality is gained at a high price. A $20,000 router becomes a $500,000 SS7-capable device and by the way, when the carriers interface to the PSTN through SS7-capable devices, access charge and settlement avoidance go out the window. Only common carriers may use the SS7 network; common carrier status requires payment of access charges and international settlements.

B. More powerful DSPs: A powerful digital signal processor (DSP) card can improve coder quality, reduce clipping from silence suppression, detect modem signals and DTMF/touchtone signals for processing, and more. Improved DSPs add hundreds of dollars to the interface, and it's not clear if a carrier could ever achieve the low cost-per-port achieved with circuit switching. Alternatively, companies such as Qwest and other IP-oriented IXCs can just forget about low-bit rate coders and silence suppressers and go with 64 KPBS coders. Of course, this scenario achieves no transport savings and no cost avoidance, when compared to circuit switching.

C. More Bandwidth: Even after an IP carrier has thrown large amounts of money at the above IP telephony quality issues, IP still may not be able to match PSTN quality. There remains the insurmountable issue of lost or delayed IP packets. Any time packets travel multiple paths between routers to get from point A to point B, packets are going to get lost or delayed. Additional transmission bandwidth and a decrease in the number of routers and links solves this problem to a degree, however, it results in a network that is not as efficient as the circuit-switched/TDM networks. A TDM-based system can run at 90 percent transmission utilization efficiency for voice connections between switches during busy hours. To approach circuit-switched/TDM utilization efficiency, packet multiplexing networks actually require more bandwidth. Note that a typical Ethernet LAN carrying packets has a bandwidth allocation of 10 MBPS but an effective throughput of less than 3 MBPS. This 70 percent overhead is allocated to reduce packet loss. The same ratio holds true for WAN IP multiservice networks. Yes, a carrier can add priorities for voice packets, but that potentially adds more router delay. Alternatively, a carrier could segment the IP network for voice only and achieve TDM efficiencies, but then what would be the point of IP telephony in the first place?

Bottom line: When a carrier tries to fix the IP telephony quality issues, it ends up with a significantly more expensive network compared with circuit switching - and IP telephony quality will always remain inferior to circuit-switched.

3. Stability

Some people are not swayed by the economics/quality issues of IP and remain dazzled by the IP visionary's statement, "Anything can be sent over IP!" This vision is 180 degrees from the 1980s vision that "IP can be sent over anything!" Members of the Internet community then, and to a great degree even now, view networks (wireless, fiber, satellite, PSTN and others) as the transport means to get IP packets from point A to point B, regardless of network quality in terms of packet loss.

The new view is that an IP network can be created to carry any form of information, including voice, data and video. This sounds good but will remain a fantasy unless the quality of IP networks can be managed. In a nutshell, IP technology is too dynamic and unstable for the common carrier community to bet the farm on. To replace circuit switches with IP, the new network architecture would have to be frozen in time for years ahead. Here are three reasons why IP is not yet considered a stable architecture.

A. Standards: There are three options regarding IP telephony standards. First, go to the International Telecommunications Union (ITU). This route would take many years and, besides, the ITU is dominated by Europeans and emerging nations that are even more circuit-switch focused than most U.S. carriers. Second, a carrier could look for stability from pseudo standards created by the Internet Engineering Task Force (IETF). The problem is that the IETF is still dominated by university types more interested in brilliant academic solutions than in practical business solutions. Finally, a carrier could go with the IP network technology leader, Cisco. Take the position that whatever Cisco and its legions of technical experts think is the way to go, and just choose that for standards. That hasn't worked to date in the IP world, and it's not likely to work anytime in the near future.

B. QoS: Managing a network requires the establishment of a quality of service (QoS). That's why circuit-switch carriers love ATM/SONET network technology. With IP telephony, however, there's no consensus today on what coder rates, packet size, delay/buffer tolerance, packet loss or silence suppression threshold should be, let alone a way to measure QoS. Now take this QoS/measurement/management issue to the real world. Picture a call, as IP zealots do, originating on an IP telephone, passing through a LAN-based PBX, then through an originating ISP network, to a WAN ISP network, through a terminating ISP PSTN network, and finally, to the called party's PSTN or IP phone. This call passes through multiple networks, each with multiple traffic types being processed, and each with its own individual industry definition of QoS and network stability. Defining managing and measuring end-to-end IP telephony QoS is going to take years before a stable IP network can be realized.

C. Multiple IP Carriers: The circuit-switched networks are stable today because the old Bell System defined what the PSTN should be, and that was that. The regulators didn't argue about it, the non-Bell System equipment vendors didn't argue, and neither did MCI and the other Bell System competition that popped up in the late 70s and 80s.

Today literally thousands of carriers view IP as the key differentiator of their service offering. To make IP networks even more unstable, no one network provider will be providing end-to-end IP service; instead, multiple originating, WAN and terminating IP carriers will be involved. Point 3 bottom line: Unless a stable IP network architecture is developed that can be managed over multiple carriers, circuit switching will live on well into the 21st Century.

4. Where's the IP OSSs?
OK, a potential IP carrier could plan to forego the IP telephony cost savings, and could assume that its customers can live with sub-circuit-switched-voice quality. And, of course, each potential IP carrier assumes IP marketplace dominance, and so can forget the IP instability argument due to multiple carriers. What's next? Virtually no off-the-shelf OSS products to support IP telephony! Few products exist at a quality level of circuit switching, let alone products to enable the creation of a viable telephony carrier business beyond pre-paid calling card calls.

Let's examine a few IP OSS problem areas:
A. Billing: An IP router generates hundreds of messages per second, identifying events that may or may not be billable. Before sending to (or flooding) a billing system, these events will have to be filtered and enhanced with other information, then merged with other data.

B. Service provisioning: In the IP world, service provisioning means provisioning over multiple devices simultaneously, without shutting down a network. Circuit-switched networks are on 24 hours a day, 7 days a week, and provisioning events transpire both day and night with no network shut down.

C. ILEC Interfacing: IP telephony networks cannot remain islands surrounded by ILEC circuit switching if IP is to prevail. For many years to come, IP carriers will be dependent on ILEC unbundled network elements (loops, SS7 databases, operator service, directory services and more).

Yes, there are IP OSS products under development and, in some cases, in beta testing (see Dr. Matthew Lucas and Dave Labuda's article "Batch System for IP Billing? Think Real Time!" p.38). But for years to come, these products will be targeted toward hybrid circuit-switched IP product/service offerings, with circuit-switch oriented OSSs doing the bulk of the work.

5. Phones Must Be Ubiquitous and Cheap:
For a new technology to replace an old one either the performance has to improve dramatically or the cost has to come down dramatically. ISDN phones didn't and won't replace touchtone phones because they were too expensive for the marginal improvements in performance. Also token ring LANs provided better performance than Ethernet LANs, but their much higher cost didn't justify wholesale replacement of the simple, cheap Ethernets.

On the other hand, planes replaced trains, buses and ships as the business travel technology of choice because of performance. The shorter travel time justified the higher cost. Bottom line: If IP technology is to surround circuit-switched technology and ultimately replace it, it will have to win the equipment battle on the customer premises equipment. That means replacing today's phones or local loop PSTN interfaces with IP! Finally, what can consumers do today with an IP phone that they can't do with a PSTN phone with a PC attachment? The answer is nothing!

6. How about wireless IP?
At the last CTIA show, the cellular/PCS CEOs in unison sent the message: "wireless data revenues will exceed voice revenues in 5 years (see Billing World's "Publisher's Letter", April 1998). If so, there must be an IP opportunity to replace today's wireless circuit switches. Sorry IP zealots - not any time soon. Here are three reasons why:
A. CDMA: The multiplexing choice for many wireless carriers today and for practically all wireless carriers in a couple of years is code division multiple access (CDMA). IP is basically a multiplexing technology, too (see Billing World's "Publisher's Letter", October 1998), except packets and labels are used to separate one user's digits from another, rather than codes. IP telephony over mobile wireless buys nothing except service degeneration (lost or delayed packets) and unnecessary overhead (less spectrum efficiency).

B. Coder/Suppression: Today's wireless phones already use 8 KBPS coders, some use silence suppression, and all new phones in a few years will use both efficiency techniques. Wireless IP would add nothing except service degradation due to lost and delayed digits.

C. Data Revenues? The market projections of massive new wireless data revenues within a few years are all but dead promises. For wireless data to take off as a revenue winner, wireless industry CEOs now realize consumers will need new wireless data terminals, carriers will need a packet switching metering infrastructure, and enterprises will probably need to change their websites. To accommodate low function wireless data terminals, the CEOs' view today is to build up circuit-switched minutes of voice use by introducing wireless data features, including integrated messaging, e-mail and voice mail. New IP switches or access methods aren't needed to do this. IP technology will be an add-on to wireless circuit switches for years to come.

7. Cable TV and IP:
Will IP technology, in general, revolutionize cable TV, and will providing IP telephony be important? Absolutely yes. Will this result in circuit-switched technology becoming obsolete? Absolutely not! Here are three more reasons why circuit switching will live on and on!

A. It's IP for Multiplexing: The reason IP telephony makes sense for a cable TV system is that it's a great multiplexing technology for integrated access. Cable TV systems have a large broadband pipe that can be managed regarding quality. As soon as those IP digitized voice bits get to the cable head-end, it's off to a LEC's or IXC's circuit switch. Just like a wireless call, bits come in over various multiple access approaches, but they all are reformatted for circuit-switched processing.

B. Cable IP Replacing Telco Loops?: The cable industry has zero interest in assuming the universal service role of today's ILEC. Why would a cable operator want to take on 911 emergency calling responsibility anyway? By leaving at least one circuit-switched PSTN line in service, the cable operator is free of any requirement to provide lifeline service that requires powering the phone in the home via the network when commercial power goes out. This reason alone will preserve circuit switching for decades.

C. Who Wants Local Calls Anyway?: Why would a cable operator want to use its bandwidth for local calls? The consumer rate would probably have to be flat, with unlimited use to be competitive. Besides, cable operators would be competing with a subsidized ILEC service. Alternatively, the cable operator will target outgoing long distance and value-added voice, which will be priced per minute of use for a long time to come. Bottom line: Cable companies need circuit-switch technology as much as the ILECs. IP is revolutionary for the cable industry because of integrated access, not switching.

8. The xDSLs and IP
The xDSL technologies are hot because of IP. But that's because consumers and small businesses want faster Internet access or more enhanced corporate LAN access from home at low cost. Nobody's saying, "Forget data, we want the xDSLs for cheaper voice service!"

To see where the xDSLs and IP fit, look at the carriers pushing xDSL deployment:
A. ILECs: The ILEC vision is Asymmetrical Digital Subscriber Line (ADSL) to the home for high speed Internet access via their DSLAM (modem/router) and if luck will have it, to their ISP service. Nowhere do their plans call for circuit-switch replacement. If a voice call came in over the ADSL connection, it would just be routed to a conventional circuit switch.

B. IXC CLECs: The IXC CLECs (AT&T, MCI WorldCom and others) see High Bit Rate DSL (HDSL) as a cheap T-1 circuit that can be divided, perhaps to support circuit-switched voice, Frame Relay or Native IP access and more. Again, no circuit-switch displacement is seen here.

C. Data CLECs: Finally, the data CLECs today serve ISPs, corporate users and others who want access to xDSL technology but can't get it from the ILECs because they are not certified carriers. Data CLECs are drawn by the need for better access to the Internet and corporate LANs.

The xDSLs will not drive circuit switch replacement. But as with Cable TV, xDSL and IP will be revolutionary because of better and/or integrated access to the Internet or corporate intranets.

9. FCC as a Champion?
The FCC would be out of their minds in supporting policies that would see IP replacing circuit switching. Here are three, no brainer, if it ain't broke, why fix it, reasons. First, a mechanism is in place to gather money from ILECs and IXCs to support the Universal Service Fund for high cost-service areas. This revenue is generated by circuit-switched technology and can be audited to the penny! Second, circuit-switched infrastructure is depreciated on a 12-to-15 year schedule. Very few digital circuit switches are fully depreciated. Policies to accelerate the depreciation of circuit-switch voice technology will lead to higher consumer prices. Finally, to implement a policy to replace circuit switches would put the FCC smack up against the State Public Utility Commissions. Chances of this happening are zero, zip, nada. There are no reasons for the FCC to start a circuit-switch replacement campaign, but dozens of reasons why not.

10. ILECs on Circuit Switch Replacement:
If ever there was an opportunity to spout the classic telecom industry zinger, "the only people who are interested in standards are people who don't have market share," this is it. Except in this case, replace the word "standard" with "IP telephony". The ILECs have absolutely no incentive to see their local switches replaced with IP telephony switches, and every incentive to see that it doesn't happen.

First, replacing circuit switching would require the phasing out of a $100-billion plus investment. Second, the State PUCs wouldn't approve accelerated depreciation, anyway. Finally, circuit-switched infrastructure generates nearly 100 percent of revenues and profits. Replacing circuit switched technology with IP would generate not more revenue but, certainly, more competition.

If the ILECs have succeeded in stonewalling OSS integration of their networks with CLEC networks for the last three years, even though it was mandated by the Telecom Act of 1996, then ILECs will also stonewall the integration of IP telephony and their network for as long as possible. The law says that the ILECs must interconnect their OSSs with the CLECs. Considering the CLECs use the same circuit switch technology as the ILECs, what makes the IP telephony people think that they are going to see seamless integration with the PSTN anytime soon?

These are my 10 reasons why circuit switching will not be replaced by IP for years. However, do I think aggressively implementing IP technology in general and IP telephony in specific is crucial for carrier success in the year 2000 and beyond? Absolutely yes!

In next month's issue of Billing World, I'll give my analysis of IP winning strategies for those developing billing, customer care and OSS products and services.