STEM newsletter

The economic case for next-generation voice

30 October 2003

In the STEM newsletter for October 2002, we announced that we were going to examine the economic case for the shift to VoIP in the public switched-telephony network, with the help of a model which would provide a detailed breakdown of the required investment and its impact on existing capital and operating expenditure. Through a collaboration with Mahindra BT and Siemens, the framework for this model is now complete, and its structure has been presented at both the STEM User Group Meeting in September and the ITU Telecom World Forum in Geneva in October.

In a continuation of the original newsletter article, we examine the technological and economic context which promotes the possible transformation of the core network, make a high-level comparison of technologies and define a plausible transition roadmap. We then decide on a suitable geographical scale and topology as the basis for an econometric model.

Time for change

Time division multiplexing (TDM) circuit-switched voice networks represent a mature and polished technology. Many incumbent digital voice networks have been in operation with little fundamental change for ten years. While some vendors still maintain and continue to deliver these products, others now focus exclusively on promoting IP-based solutions, with increasing maintenance costs for local exchanges as spares become scarce.

Although the modern PSTN is a well-oiled product in its present form, ‘value-added’ extensions to the stand-alone voice service are costly and slow to develop, since any related data must be transported over parallel networks. Furthermore, demand for converged services is growing, especially for call-centre applications.

VoIP technology is maturing

IP telephony over the public Internet is a ‘best-efforts’ offering and suffers from routeing delays and occasional blocking. However, private networks running IPv6 technology support priority classes and quality of service (QoS), and multi-protocol label switching (MPLS) reduces routeing delays. VoIP is becoming recognised as a proven technology on corporate LANs and WANs. The use of dense wavelength division multiplexing (DWDM) equipment in the core network means that bandwidth prices are falling, and the emergence of very fast core routers removes the necessity for processing overheads (such as silence compression) which would otherwise introduce delays. Although an uncompressed 64kbit/s voice channel may incur an overhead of between 100% and 200%, depending on the core IP or ATM network encapsulation, the fact is that ‘brute force’ network dimensioning is becoming a reality, and removes many of the barriers to high performance.

Next-generation voice offers hope

Fixed-line revenues are falling as a result of regulation and mobile cannibalisation. Operators hope that an integrated IP platform may support ‘joined-up’ voice services, such as ‘click-to-call’ from the PC, with the potential to generate new revenue streams.

Dynamic routeing with VoIP offers the possibility to have a single number that connects to your current call location, be it at home, in the office, or on the move. Residential broadband could enable the introduction of fixed handsets with features that are characteristic of mobile systems, or even a roaming ‘personal handset’ configuration that could include a portable address book.

Making the case for change

All of these value statements are great for advertisements and for motivating research and development, but not for justifying billion-dollar investments! A detailed business case must justify investment in next-generation networks (NGN), demonstrating the promised reductions in opex and revenue from new services, in order to convince investors wary from the lessons of the crash in hi-tech stocks. Note: technical viability is an assumption for the business case!

Focus on the core network

A significant proportion of most incumbents’ customers and revenues come from rural areas, requiring the local network intelligence of digital local exchanges to be distributed over a very large number of sites. VoIP offers a greatly simplified switching architecture: media gateways at the edge of the network ‘talk’ over a homogeneous IP platform with the aid of a few core soft switches and related servers; the vast majority of switching sites become commodity routeing or aggregation points in a rationalised core network.

Even if residential broadband offers home gateway devices with features which will finally render conventional handsets obsolete, operators cannot control the pace of replacement of residential customer premises equipment (CPEs). However, conventional services can still be delivered at the edge of the network, with media gateways at the remote concentrator (‘edge-of-fibre’ network) in such a way that the core voice transport is opaque to the user. Our model will therefore focus on the migration to VoIP in the core network, with the majority of lines still offered as voice ‘circuits’; rather than IP pipes.

Opex spending should be reduced

A central objective for the operator is to reduce the number of parallel networks which must be maintained at local exchanges by pushing service delivery to the edge in the form of so-called ‘multi-service access platforms’. Smaller capacity (and therefore better utilised) units may be located at remote concentrators where required, or deployed as enterprise CPEs. This move away from the so-called ‘stove-pipe’ model should reduce equipment maintenance overheads and facilitate more efficient service configuration and delivery. For example, BT’s 21c network project has target opex savings of around GBP270 million per year for the UK.

Comparison of technologies

The Exhibit below illustrates the main cost components of the conventional TDM voice architecture. The copper local loop connects a subscriber’s handset to a line card at a remote concentrator unit (RCU), i.e. a small building housing the distribution frame, line-card equipment (possibly an aggregator for other data services), then typically an STM-1 interface. The RCU has a fibre connection to the digital local exchange (DLE), which may in fact serve customers from 5–20 or more RCUs, depending on the area.

Simplified view of conventional TDM voice architecture

The DLE in turn has a fibre uplink to a digital main switch unit (DMSU). Depending on the number of subscribers, this may require an STM-16 or STM-64 interface, and in fact most (though not all) DLEs will be ‘dual-homed’. In other words, parallel uplinks are implemented to two separate DMSUs for maximum network resilience.

International gateways and other points of interconnection will typically be located at DMSU sites and the DMSUs are served by some kind of meshed core network. (Since the dimensioning of this core network is dependent on both voice and data services, and probably not significantly affected by a migration to VoIP, it is not included in the model.)

Essentials of VoIP core overlay

A media gateway is a device which is designed to interface between a conventional voice circuit and packet-switched voice traffic on an IP network. The simplest upgrade path is to install a media gateway which plugs into the TDM interface from a rack of conventional line cards, and to connect this to a Gigabit Ethernet network running IPv6 and MPLS.

VoIP core overlay

Number–address translation is performed by a central ENUM server(s), and a ‘soft switch’ handles the call or session set-up (for billing and so on). A number of soft switches may be distributed around an incumbent’s network, but only at main switch sites. These soft switches are servers dimensioned in term of ports or concurrent sessions. Since the actual voice packets are routed directly to the call destination, not ‘through’ a soft switch, their actual location is not that critical.

An important point is that, once all RCUs for a given DLE are migrated to VoIP, the complex DLE is replaced by simple bandwidth aggregation. This effect is repeated at very many sites as the migration is rolled-out over the network.

A PSTN gateway is essential during the migration, acting as the interface between migrated customers and the remaining PSTN, and will remain necessary for external legacy networks for years to come.

Technology comparison

The model compares two main scenarios, the first in which the operator retains all voice services on TDM, and the second exploring a roadmap for the introduction of NGN services and the migration of all core voice transport to IP. Additional scenarios will examine technical alternatives for the edge of the network.

Geographical scale

This vendor-neutral model will describe the complete network of an idealised incumbent operator in a small Scandinavian country with:

  • 2.5 million residential lines
  • 500 digital local exchange (DLE) sites
  • 25 digital main switch unit (DMSU) sites.

This provides a more manageable data set than many larger countries, but retains sufficient detail to invite comparison with any incumbent operator’s network. The demonstrator model developed to date uses five geo-types for local exchange costs, but an individual operator may choose to model at the individual local exchange level for maximum detail.

Roadmap for transition

Phase 1
Edge CPE devices are introduced to deliver integrated multi-service access interface for corporate customers, some with PABX traffic routed over a virtual E1 to the PSTN, but a growing proportion running VoIP on the corporate LAN, and transported over Gigabit Ethernet (providing required QoS) to a couple of VoIP/PSTN gateways (redundancy). Soft-switch capacity is gradually increased, with capex falling from GBP25 to GBP10 per port over five years.
Phase 2
Mix of SDSL or FTTP for SMEs with smaller edge access devices, again with a growing proportion with IP phones; up to five VoIP/PSTN gateways.
Phase 3
Assumption that an appropriate feature-set for carrier-grade voice will have been developed through experience with business sector. Individual local exchanges are replaced (according to an explicit replication template) as obsolescence determines. Converged DSLAM edge devices deliver TDM voice or ADSL (remote configuration) to 98% of subscribers, such that individual local concentrators may be ‘peeled off’ PSTN network according to a manageable work plan.

So it works, but does it make money?

The ‘do-nothing’ scenario should show strong erosion of both price and market share for a fixed-line voice product, as mobile call charges fall and 3G addresses 2G spectrum limits.

Different incumbents have varying obsolescence issues regarding TDM, but one conclusion may be that an incumbent should make the transition (to IP) while it can still justify investment in its voice network.

We expect opex costs to rise over time for TDM, in contrast to those for VoIP, which we expect to fall as global economies of scale are realised.

The model which has been developed so far demonstrates the feasibility of linking these complex technological and topological factors into a business case, but the variables (including age of existing TDM network, and even resilience of copper to local climate) are such that conclusive quantitative results will only emerge from critical studies on behalf of individual operators.

Sections of this article are © 2003 International Telecommunication Union (ITU) and were first published in the Forum Proceedings of the ITU Telecom World 2003 Forum. No endorsement by the ITU of the opinions expressed in this article is implied.

© Implied Logic Limited