Interconnection in an IP-Based NGN Environment

1. Interconnection in anIP-Based NGN Environment J. Scott Marcus, Senior Consultant ITU Workshop: What rules for IP-enabled NGNs? Geneva, March 23-24, 2006

2. Interconnection in anIP-Based NGN Environment • Introduction • The economics of interconnection • Fixed, mobile, Internet • Retail and wholesale arrangements • Quality of Service • Market power and interconnection • Interconnection and universal service • Billing and accounting challenges • A hypothetical scenario • Summary

3. Introduction • IP-based NGNs represent the “marriage” of the Public Switched Telephone Network (PSTN) with the world of the Internet • Very different interconnection arrangements prevail in these two worlds. • Different technology. • Different regulatory history. • Different industry structure. • What should happen “when worlds collide?”

4. Introduction • Why do we regulate? • Market failures: Market power • Market failures: Desirable capabilities that would not deploy without help (some of which constitute “public goods”) • Manage limited resources (spectrum, numbers)

5. Introduction • What role for regulation in the world of the IP-based NGN? • Where service providers possess Significant Market Power (SMP), they will tend to have both the ability and the incentive to exploit that market power, to the detriment of consumers. In the absence of regulation, interconnection often serves as a locus for the exploitation of SMP. • “Coase Theorem” (1959) – private parties can often negotiate arrangements more efficiently than government regulators, provided that necessary preconditions have been met. • In markets where competition is fully effective (no SMP exists), competitive forces will generally make regulation unnecessary. “Things should be as simple as they can be, but no simpler.” - EInstein

6. Introduction • NGN access versus NGN core (source: ECTA) • NGN access: “the deployment of fibre into the local loop, either to the incumbent’s street cabinet … or the deployment of fibre all the way to customer premises (typically apartment blocks rather than individual houses). • NGN access: “the replacement of legacy transmission and switching equipment by IP technology in the core, or backbone, network. This involves changing telephony switches and installing routers and Voice over IP equipment.” • Significantly different regulatory implications. • My primary focus in this talk is on the NGN core, but broadband deployment generally and NGN access in particular interact with these issues.

7. Introduction • My history • Senior Consultant, WIK-Consult (Germany) • Senior Advisor for Internet Technology, FCC (USA) • Chief Technology Officer, GTE Internetworking (USA) • Engineer by training • My approach to these interconnection issues is primarily through economics rather than engineering.

8. The economics of interconnection – retail • Calling party pays (CPP): the party that initiates the call pays for the call, usually based on the duration of the call; generally, the party that receives (terminates) the call pays nothing. • Receiving party pays (RPP) or Mobile Party Pays (MPP): the originating and terminating parties each pay a share for the call. In North America, where this system historically has been used, mobile receiving parties paid but fixed receiving parties did not. • Flat rate: the consumer pays a fixed (monthly) fee for unlimited domestic calls. • The “buckets of minutes” plan: the consumer pays a fixed (monthly) fee for some number of minutes of domestic calls, but pays a per-minute fee for minutes in excess of those in the “bucket”.

9. The economics of interconnection – retail • CPP arrangements reflected the historical perception that the caller is the primary beneficiary of the call, and also the main cost-causer. • This concept has been challenged in recent years • Clearly, the receiver also benefits. • If the receiver saw no merit in the call, he or she could simply hang up; thus, after the first minute, caller and called party can be viewed as (equal) partners in the call. (Cf. Jeon et. al.) • In the world of the IP-based NGN, origination and termination are likely to become less relevant over time. (Cf. de Graba)

10. The economics of interconnection – retail • Consumers tend to grealy prefer flat rate (or “buckets”) plans over usage-based plans (Cf. Odlyzko) • AT&T Wireless’s offer of Digital One Rate (1998) • America Online’s flat rate Internet access (1995) • In the United States, flat rate / bucket plans are increasingly prevalent at all levels • Mobile services • Fixed services, including long distance • Internet access

11. Payment Originating Operator Terminating Operator The economics of interconnection – wholesale • Calling Party’s Network Pays (CPNP): the calling party’s network (the originating operator) makes a wholesale payment to the receiving party’s network (the terminating operator). • “Bill and Keep”: a U.S. term of art denoting the absence of any regulatory obligation for payments between the networks.

12. The economics of interconnection – wholesale • In an unregulated CPNP system, carriers will tend to establish very high termination charge levels. Normal economic forces provide an inadequate “brake” on this practice, because the terminating operator is imposing the charges indirectly on another carrier’s customer. The terminating operator does not bear the full burden of suppressing demand through a price that is arguably too high. • These high prices impact consumer welfare in a number of ways. This problem is general referred to as the termination monopoly. • Paradoxically, small operators will be motivated to set termination charges to even higher levels than will larger operators. (Cf. Laffont and Tirole (2001); Haucap and Dewenter) • Regulatory asymmetries – for example, between regulated fixed operators and unregulated mobile operators – can exacerbate this problem.

13. The economics of interconnection • Termination charges at the wholesale level interact with retail pricing arrangements. • The termination fee generally sets a floor on the retail price. • Where termination fees are high, they generally prevent flat rate or “buckets” plans from emerging. • This is true even where payments between the operators are in rough balance, such that little money changes hands. • Each operator will tend to view the termination charge as a component of its marginal cost. (Cf. Laffont and Tirole) • If an operator chooses to ignore this wholesale cost in the hope that the payments will balance anyway, that operator risks attracting customers who place disproportionately many calls to customers of other providers (“adverse selection”).

14. The economics of interconnection • Mobile operators that implement CPP/CPNP tend to have the following characteristics at the retail level: • Low or zero initial cost • Low or zero monthly cost • High usage (per minute) cost • Mobile operators (U.S.) that implement “buckets” plans and Bill and Keep tend to have the following characteristics at the retail level: • Higher initial cost • Higher monthly cost • Low or zero effective usage (per minute) cost • These differences tend to lead to faster adoption of the mobile service in CPP/CPNP systems, but much lower rates of utilization.

15. The economics of interconnection Source of data: U.S. FCC, 10th CMRS Report, July 2005, Table 10, based on Glen Campbell et al., Global Wireless Matrix 4Q04, Merrill Lynch, Apr. 13, 2005.

16. The economics of interconnection • In the U.S., the FCC has been attempting for years to migrate their interconnection arrangements to a Bill and Keep basis for all services. (Cf. de Graba (2000), Atkinson and Barnekov (2000)). • The FCC’s sense has been that: • Bill and Keep simplifies regulatory rate-setting or avoids it altogether. • Bill and Keep already works well in many settings in the U.S. • Bill and Keep will be easier to apply to the IP-based networks of the future. • To date, the U.S. has been unable to forge a political consensus to move forward on this issue.

17. Peering – economic models • An extensive economics literature exists about interconnection in the traditional PSTN world. • An emerging literature deals with interconnection in the world of the Internet. • We are in the early stages of understanding the relationships between the two.

18. Peering and Transit • “Peering is an agreement between ISPs to carry traffic for each other and for their respective customers. Peering does not include the obligation to carry traffic to third parties. Peering is usually a bilateral business and technical arrangement, where two providers agree to accept traffic from one another, and from one another’s customers (and thus from their customers’ customers). … • Transit is an agreement where an ISP agrees to carry traffic on behalf of another ISP or end user. In most cases transit will include an obligation to carry traffic to third parties. Transit is usually a bilateral business and technical arrangement, where one provider (the transit provider) agrees to carry traffic to third parties on behalf of another provider or an end user (the customer). In most cases, the transit provider carries traffic to and from its other customers, and to and from every destination on the Internet, as part of the transit arrangement. In a transit agreement, the ISP often also provides ancillary services, such as Service Level Agreements, installation support, local telecom provisioning, and Network Operations Center (NOC) support. • Peering thus offers a provider access only to a single provider’s customers. Transit, by contrast, usually provides access at a predictable price to the entire Internet. • Historically, peering has often been done on a bill-and-keep basis, without cash payments. Peering where there is no explicit exchange of money between parties, and where each party supports part of the cost of the interconnect, … is typically used where both parties perceive a roughly equal exchange of value. Peering therefore is fundamentally a barter relationship.” - NRIC V (advisory council to FCC)

19. Peering and Transit Transit Connection Larger ISP or Backbone Regional or Local ISP Many remote locations connect to a regional or local ISP with individual, low bandwidth connections Concentration to a larger ISP or backbone provider with global connectivity by means of a concentrated, high bandwidth connection

20. Peering and Transit Transit Connection Larger ISP or Backbone Regional or Local ISP This peering connection will tend to exist if the cost of the connection to each ISP is less than the money each saves due to reduced transit traffic. Transit Connection Larger ISP or Backbone Regional or Local ISP

21. Peering and Transit • In general, money flows upstream, while obligations flow downstream. • Transit agreements are vastly simpler than peering agreements. • In general, peering is a bilateral technical and commercial arrangement. Cf. Lixin Gao (2000)

22. Peering – economic models • Define: co as cost of origination ct as cost of termination a as an access charge levied on the sender • Due to shortest exit, ct > co • Then cost for the originating network is co + a cost for the terminating network is ct – a The model extends in a straightforward way to accommodate multiple levels of quality of service (QoS). Network i Network j Source: Laffont et. al., “Internet Interconnection and the Off-Net-Cost Pricing Principle”

23. Peering – economic models “A key difference with this telecommunications literature is that in the latter there is a missing price: receivers do not pay for receiving calls; … The missing price has … important implications: … The operators’ optimal usage price reflects their perceived marginal cost. But when operators do not charge their customers … for the traffic they receive, operator i ’s perceived marginal cost of outgoing … traffic is … the unit cost of traffic is the on-net cost c, augmented by the expected off-net “markup”. … Comparing the two perceived marginal costs of outgoing traffic with and without receiver charge, for given access charge and market shares, the price for sending traffic is higher (lower) than in the presence of reception charges if and only if there is a termination discount (markup). … In sum, the missing payment affects the backbones’ perceived costs, and it reallocates costs between origination and reception.” Source: Laffont et. al., “Internet Interconnection and the Off-Net-Cost Pricing Principle”

24. Market power and interconnection • Regulators continue to find it necessary to intervene where an operator has Significant Market Power (SMP). • The migration to NGN will not necessarily eliminate SMP. Notably, market power associated with last mile bottlenecks will continue to be a significant regulatory concern for the foreseeable future. • A new market power challenge has appeared, primarily in the U.S.: the network neutrality issue.

25. Market power and interconnection • Network neutrality means different things to different people: • The possibility that an integrated ISP might offer better performance to some Internet sites than to others; • The possibility that an integrated ISP might assess a surcharge where a customer wants better-than-standard performance to certain Internet sites; • The fear that the integrated ISP might permit access only to affiliated sites, and block access to unaffiliated sites; • The fear that the integrated ISP might assess surcharges for the use of certain applications, or of certain devices; • The fear that the integrated ISP might disallow outright the use of certain applications, or of certain devices, especially where those applications or devices compete with services that the integrated ISP offers and for which it charges; and • The fear that the integrated ISP might erect “tollgates” in order to collect unwarranted charges from unaffiliated content providers who need to reach the integrated ISP’s customers.

26. Market power and interconnection “The chief executive of AT&T, Edward Whitacre, told Business Week last year that his company (then called SBC Communications) wanted some way to charge major Internet concerns like Google and Vonage for the bandwidth they use. "What they would like to do is use my pipes free, but I ain't going to let them do that because we have spent this capital and we have to have a return on it," he said.” NY Times, March 8, 2006

27. Market power and interconnection • Many of the concerns that have been raised in regard to network neutrality relate to behaviors that, in the absence of market power, would tend to enhance consumer welfare. • Some would appear to represent legitimate price discrimination. • Others enforce the economic property of excludability (the ability to prevent someone from using a service that he did not pay for) in support of price discrimination. • The form of market power that could potentially be exploited in anticompetitive ways in connection with network neutrality relates to network externalities (where the value of a service depends on the number of users of the service). (Cf. Katz and Shapiro (1985)). • The degree to which this issue has heated up recently in the U.S. probably reflects increasing concentration in the relevant markets.

28. Market power and interconnection • Trying to address these network neutrality challenges through regulation ex ante (in advance) is likely to prove extremely difficult. • A first line of defense for regulatory authorities should instead be to maintain the competitiveness of the underlying markets, especially as regards broadband Internet access and as regards high capacity Internet transit. Service-based competition (rather than facilities-based competition) would be sufficient for this purpose. • In countries where competition law provides an ex post (after the fact) complement to regulation, it might be most appropriate to deal with occasional or sporadic problems related to network neutrality through the exercise of competition law.

29. Differentiated Quality of Service (QoS) • A series of initiatives to implement inter-provider QoS on an inter-provider basis have generated scant results. • Early Nineties – RSVP • Late Nineties - DiffServ • Backbone peering attempts – especially circa 2000 - 2001

30. Differentiated Quality of Service (QoS) • Early Nineties – Integrated Services Architecture (RSVP) • Comprehensive system of prioritized delivery. • Required significant “soft state” in the routers. • Perceived as too difficult to deploy. • Not entirely true – BBN (my former employer) had working RSVP-capable production networks from the mid-Nineties. • Technical success. • Market failure. • No customer willingness to pay a significant premium for on-net differentiated service.

31. Differentiated Quality of Service (QoS) • Traffic is classified on entry to a network • Metered • Marked • Policed • Shaped • Services implemented based on defined Per-Hop Behaviors (PHBs) • Queue processing (prioritization) • Queue management (drops)

32. Differentiated Quality of Service (QoS) • At a technical level, QoS is not fundamentally hard. • DiffServ is technically trivial. • MPLS in a single network is technically trivial. • Cross-provider MPLS is only marginally harder. • Even RSVP is not that hard. My company, BBN, had working production RSVP-compliant networks in 1995! • In terms of the basic economics, QoS is not fundamentally hard. • Nonetheless, there is no significant cross-provider roll-out to date. • WHY NOT?

33. Differentiated Quality of Service (QoS) M/G/1 queueing analysis of link performance (with clocking delay of 50 μsecs (284 byte packets) and a 155 Mbps link)

34. Differentiated Quality of Service (QoS) • For real time services such as voice telephony traffic, it is important that mean delay and variability of delay be held to low values. • Delay in excess of about 150 milliseconds causes „collisions“. • Buffering can address variability as long as the mean and variance are not too great. • The buffer then represents a fixed increment to the propagation delay. • For circuit speeds of T-3 and up, queuing delays in a properly designed network will generally be less than 1 millisecond per hop under normal operating conditions. • Propagation delay (speed of light) will tend to dominate any variable queuing delays under normal operating conditions.

35. Differentiated Quality of Service (QoS) • IMPLICATION: Most of the time, and under normal conditions, variable delay in the core of the network(s) is unlikely to be perceptible to the VoIP user. • FURTHER IMPLICATION: Consumers will not willingly pay a large premium for a performance difference that they cannot perceive. • Packet delay is more likely to be an issue: • For slower circuits at the edge of the network • For shared circuits (e.g. cable modem services) • When one or more circuits are saturated • When one or more components have failed • When a force majeure incident has occurred

36. Differentiated Quality of Service (QoS) • Technical challenges, or economic challenges? • Revenues • Limited customer willingness to pay a substantial premium. • Limited benefits until widely deployed (network effects). • Costs • Agreements needed with many peering partners. • Economic transaction costs to negotiate each agreement. • Measurement, management and dispute resolution challenges. • The business case is difficult to “prove in”. • Implies difficulties in getting past the initial adoption hump.

37. Interconnection and universal service • Most countries seek to ensure that all can afford basic communication services. This often implies some degree of subsidy support to areas that are remote or that have low teledensity. • There is some economic basis for this, grounded again in the theory of network externalities. The communications network is worth more to everyone if nearly everyone is reachable through it. • At the same time, economists worry about the economic distortions implicit in any system of universal service. Universal service implies that service will be deployed to areas where rational economic judgment alone would not lead to deployment. • In many countries, interconnection charges are used to provide implicit or explicit funding support for universal service programs.

38. Interconnection and universal service • Some developing countries (and also some developed countries, including the U.S.) permit rural operators to assess higher termination fees than urban operators, presumably reflecting higher unit costs (due to lower density of consumers, and possibly to more challenging geography). This effectively subsidizes the rural operator. Organizations such as the World Bank have tended to be supportive of such practices. • International termination charges have also tended to generate subsidies from developed countries to developing countries, even if the rates are symmetric. Far more calls are placed from developed countries to developing countries than vice versa. This asymmetry generates an effective subsidy to the developing world, which could be used to support universal service programs.

39. Interconnection and universal service • The migration to IP-based NGNs will tend to put pressure on any interconnection arrangements that are widely at variance with cost. It will, moreover, tend to expand opportunities to bypass inefficient interconnection arrangements. • There are many, obvious problems associated with using termination charges to subsidize universal service. Nonetheless, the concern that must be raised is that it is not clear what developing countries might do if these subsidies were to disappear over the next few years.

40. Billing and accounting • Billing/accounting systems must enable a service or network provider to reliably charge for the service. • Both parties must ultimately agree on the correctness of the charge. To the extent that charges are not predetermined, this implies: • Clear documentation of the basis for the bill. • Data capture (accounting) to create and substantiate the bill. • Dispute resolution / reconciliation procedures.

41. Billing and accounting – requirements • In a DiffServ environment, billing and accounting systems would likely need to: • Enable the sending provider to mark the DS codepoint required for a particular IP datagram. • Enable the receiving provider to accept or decline the request, remarking the IP datagram if necessary. • Authorize one provider to apply a surcharge to the other's traffic in exchange for respecting its request for differentiated handling. • Enable both providers to verify that the service that was delivered conformed to their service level specification commitments to one another. • Facilitate trouble-shooting across multiple provider networks. • The providers could, instead, choose to waive payments (Bill and Keep).

42. Billing and accounting – challenges • In an NGN world, the network service provider (the ISP) will not necessarily be the application service provider. A VoIP service or an IPTV provider will not necessarily be a network provider. • The network provider will have only limited visibility into third party applications running over its network (and the user could further reduce visibility by encrypting the data). • The unaffiliated application provider may have extensive visibility into the application that it provides, but only limited visibility into the use of network resources. • Usage-based billing will be possible only to the extent that the usage can be rigorously and unambiguously measured.

43. Billing and accounting – challenges • How will providers and customers ensure that service commitments are met? Whose statistics will govern? • Competitive providers are reluctant to share statistics about their respective networks with one another, and peering agreements typically restrict the ability of the providers to disclose information about one another‘s networks to third parties. Can sufficient information be disclosed to customers? • How will responsibility be allocated if a customer’s traffic fails to achieve its committed service level specification? Traffic data can legitimately be interpreted in more than one way. Will it be possible to administer payments and penalties rigorously and fairly? • How can providers prevent fraud? How can they distinguish between fraud and legitimate use?

44. A Hypothetical Scenario • Consider a migration to an IP-based NGN in a European country. • Prior to migration, the country has: • an incumbent wired and wireless operator that had previously been the country’s PTT (BigCo), and that still has substantial market share and market power; • various wired and wireless competitive operators; • various independent providers of broadband Internet services, some facilities-based, some providing service competition based on procompetitive regulation (LLU, bitstream, and shared access); • several independent providers of VoIP; and • a number of local providers of Internet content, both web and video.

45. A Hypothetical Scenario • BigCo chooses to migrate • To achieve economies of scope through improved integration; • To achieve faster time-to-market for new services. • Competitors may feel pressure to migrate to maintain competitiveness with BigCo. • BigCo would likely continue to be subject to procompetitive regulatory obligations for competitive access to last mile facilities (e.g. local loop unbundling [LLU], shared access and bitstream access), until and unless three or more effective competitors were to emerge for the last mile. • Interconnection obligations would likely remain in place as long as traditional interconnection is offered.

46. A Hypothetical Scenario • Would it be necessary to impose interconnection obligations for IP-based interconnection? Assume arguendo that obligations are not imposed. • Extrapolating from today’s experience, BigCo would likely be motivated to implemented (settlement-free) IP-based peering arrangements with large international operators, and probably with a small number of its largest domestic competitors. • If BigCo were not motivated to peer with any domestic competitors, regulatory action might possibly be appropriate. (Cf. Telstra in Australia) • Smaller competitors could purchase transit from BigCo, or from one or more of its competitors.

47. BigCo MiddleCo ISP ISP ISP ISP ISP ISP ISP ISP SmallCo A Hypothetical Scenario ForeignCo

48. A Hypothetical Scenario • Assume that key underlying markets (including broadband access, and leased lines) are subject either to (1) effective competition, or (2) effective regulation. • Under these various assumptions, BigCo’s competitors should be able to reach BigCo customers with unit costs that are not greatly different from those of BigCo itself. This would appear to imply that a competitive market should emerge, with prices not greatly in excess of marginal cost. • ForeignCo would be able to reach BigCo customers with unit costs that exceed those of BigCo primarily by the incremental costs imposed by ForeignCo operating from outside the national territory. If BigCo were to price excessively, doing so might encourage a ForeignCo to attempt entry. This implicit threat also serves to constrain BigCo’s behavior.

49. A Hypothetical Scenario • BigCo should be able to charge a premium for better-than-best efforts service under these assumptions, but its ability to charge a premium will be limited to the willingness of consumers to pay, given that they have meaningful competitive alternatives. • BigCo should be able to sell application services (VoIP, IPTV and more) to its broadband customers, and there is no obvious reason why this should not be profitable; however, their price will tend to be constrained by the ability of independent service providers (e.g. Skype) to offer services directly to BigCo’s broadband customers. • Again, all of this seems to imply the likely emergence of an appropriately competitive market, as long as underlying facilities (broadband access, leased lines and so on) are subject either to effective competition or to effective regulation.

50. A Hypothetical Scenario • This scenario did not assume a regulatory obligation to interconnect at the IP level. • The apparent implication is that a Coasian solution – letting the providers determine their own interconnection arrangements through private arrangements – should continue to function well in an NGN world (much as it does today), as long as regulators are careful to prevent exploitation of market power in regard to underlying facilities.