The New Zealand Peering Ecosystem

Abstract

This is a document in the style of "The Evolution of the U.S. Internet Peering Ecosystem" and "The Asia Pacific Internet Peering Ecosystem" (both by William B Norton <wbn@equinix.com>) but dealing specifically with the New Zealand situation.

The need for such document arises from the observation that "There are vast differences between each region of the Internet Peering Ecosystem" and certainly, New Zealand with its location, size, culture and Internet evolution is going to have it's own quirks and optimisations which need recognition before political, commercial and economic decisions are imposed.

The document will briefly describe the evolution of the Internet in New Zealand and what is known of peering and transit services. The effect of a unique connection resource and its use as an open public Internet exchange point will be discussed, along with the second public Internet exchange.

There will be an examination of some trends in technology and policy regarding this sector and what such trends may deliver in the future NZ Internet.

NB. Throughout this document certain terms which have distracting connotations or ambiguous meanings are used to describe very precise relationships and implications. The critical terms are PEERING, TRANSIT and CONNECTION.

see the glossary for more information.

The Ecology

The term "Regional" refers to the "Internet Region" described by Norton, and best equates to "National," New Zealand being small geographically by comparison to Australia or the US. Thus "regional" in that context is replaced by "national." Where "regional" is used, it refers to the regions of NZ outside the metropolitan areas.

The Environment

New Zealand is a small country (about the size of Colorado) with a population of four million people spread over two relatively long thin islands. About 80% of the population lives in cities.

New Zealanders are regarded as cost sensitive and more willing than most to adopt new technologies where greater efficiency results.

Brief Internet History

The New Zealand telecommunications market was once a state monopoly, but over the period 1 October 1987 - 1 April 1989 the supply of telecommunications customer premises equipment was progressively deregulated. From 1 April 1989 all legal restrictions on entry into telecommunications service markets in New Zealand were removed.

In 1989, roughly speaking, NZ gained its first connection to the global Internet. The connection terminated at Waikato University and was named NZGATE.

For a number of years afterward, this was the source of International transit and national peering. It seems likely private peering relationships existed, but no documentary evidence is available.

This posting to the NZ Network Operators Group mailing list provides most of this historical background.

Waikato University then outsourced NZGATE to Netway (a 50% subsidiary of TCNZ) around the beginning of 1996 and Waikato's involvement gradually declined while Netway formed the NZIX, an Intenet Exchange that offered peering, and international transit from a number of vendors.

The geographical location of NZGATE, far from the largest markets, reduced its appeal and the current IXs formed in Wellington and Auckland.

The Wellington Internet Exchange (WIX) was established in 1997, and the Auckland Peering Exchange (APE) installed in the Sky Tower in 1999.

Carriers

This term is used to describe facilities based carriage providers who may or may not provide additional services over the infrastructure they control.

Telecom NZ is the major incumbent carrier with national reach and a full-service orientation. They provide both wired and wireless services and own infrastructure to support both.

TelstraClear NZ is the other national full-service wired carrier. They have some infrastructure, mostly in CBDs and wholesale services from TCNZ to resell. They provide cellular services via an arrangement with Vodafone NZ, though with Vodafone's move into corporate cellular services, this agreement may lapse and TCL has given indications it may build a cellular network of its own, New Zealand's third.

There are also a growing number of regional and metropolitan carriers who tend toward an open access, low cost model with a variety of service providers delivering their wares over the network.

A very early example of this type of near-serviceless network is CityLink formed in late 1995

Other operators have entered this market, including Wired Country, Vector Communications, and others have expressed intentions to do so.

Internet Exchanges

Currently there are two neutral, open access, public Internet exchanges in New Zealand, WIX and APE.

These IX are hosted by CityLink and have slightly different models in Wellington and Auckland.

Participation in each exchange is free with a connection to the local PublicLAN environment and requires compliance with small number of essential protocol requirements.

In Auckland it is possible to select a third-party carrier to access a switch located in the SkyTower, connected to PublicLAN and thus participate in APE.

There are now two "room"-style adjunct locations to the WIX.

Tier 1 ISPs

Using the definition that a Tier 1 ISP pays no regional transit (only peers and is paid transit by others), there are a number who fit the criterion. TelstraClear and Xtra a service of Telecom NZ.

There are not the exclusive facilities based Tier 1 ISPs such as the big four in Australia or the US Tier 1s.

The NZ Internet region was established by smaller players who used carrier infrastructure. Once the Telco carriers entered the market (TCNZ by developing Xtra, TCL by acquisition of Netlink, Paradise and ClearNet) large "Tier 1"-style market dominating players have begun to emerge.

Particularly active in the formation of a more conventional Tier 1 environment in NZ is TCL who is by far the largest of the "Gang of Four" in the Australian Internet region.

TCL has withdrawn from peering and requires other network operators (private or public) to pay for a domestic combined peering & transit service and connect with them at a private transit point.

At this time it is difficult to definitively identify the motivation. It could be intended to increase revenue, inconvenience competitors, or that the significant risk to the proprietary Central Office model posed by public, neutral Internet Exchanges has been recognised, or all of the above, or something else.

Tier 2 ISPs

If the carrier strategy succeeds and the two giants form a Tier 1 club, the rest of the ISPs (and a growing number of significant corporates and content supppliers) will form a defacto Tier 2.

It remains to see where the balance of customers or traffic will lie after such a shake out. It may result in a reversal of the aspirations of those currently offering domestic transit, rather than the paid, they may become payers.

Customers

At the present stage of development of the NZ Internet, customers fall into four basic categories:

1. "Tier 2" ISPs
   
2. Corporates
   
3. Net Savvy Content Providers
   
4. SME/Residential

While this section is titled customers, the NZ Internet, like many sectors, is benefiting from lower costs of entry into IP service operation (cheap commodity routers based on Open Source products, cheap commodity bandwidth offered by metropolitan LAN operators) to begin to supply their own needs.

This group shares the desire to reduce their transit costs, and for all of these entities, this can be achieved by aggressive peering.

Trends

A description of the present day Internet ecology is bound by the history and legacy of the region in which it has grown. Since the beginning of the Internet there have been significant changes in some of the drivers that lead to the situation observed today.

Basing current decisions on the results of conditions that no longer exist may lead to non-optimal outcomes.

Later the trends identified here will be extrapolated, as best we can, to cover the likely future ecology of the NZ Internet, including some theoretical options which, though desirable, may not necessarily arise.

Blurring of Hierarchical Participant Distinctions

“We may say that the movement of the progressive societies has hitherto been a movement from Status to Contract”. (Henry Maine: Ancient Law 1861)

Increasing Customer Competence

As the importance of Internet activity grows for customers, more attention is being paid to the supply, availability and cost of Internet information exchange.

ISPs who are customers of other ISPs are not new, but customers without an ISP are a growing proportion of the market, and definitely of the traffic on the NZ Internet. This change has been driven by access to peering and the development of pure transit-play providers at the retail level.

Large customers and content providers (some with their own national networks) are beginning to take control of their Internet activities, peering at Internet Exchanges, arranging their own IP addresses, AS numbers and international transit arrangements.

Such customers are not intimidated by the Internet and need less support from an ISP. They are experiencing the benefits of greater independence and control of their own Internet destiny, rather than being at the mercy of their up stream provider.

Net-Savvy Content Providers

As noted in other Internet regions, content providers are becoming more aware of the impediment metered transit causes to their ambitions, and are taking steps to use their clout to obtain internetworking arrangements.

As the ability for incumbents to "control the customer" declines, and as DIY means that customers become competitors, the trend will give content providers more of a relationship with the customer than the carrier.

Broadband

The increasing speeds of connection will make metering problematic. Firstly as the task of counting and accounting for every byte in multi-gigabit streams becomes less feasible. Secondly, as the ability for a customer to pay for the potential volume of traffic that might occur lessens, particularly for content providers.

For example, a recent event in New Zealand, from one server (among three) was sending traffic at at rate of 750Mbs. Using a media over-cap per megabyte figure of $0.15, this would cost the content provider around $40,000 an hour to serve via a transit arrangement. This is a reasonable estimate of the revenue declined by the non-peering oper

For this reason the service was limited to peers at the Internet Exchanges. Those peers who were retail ISPs could be expected to bill a similar amount to their paying customers.

The lesson here is that at higher bandwidths, "double dipping," being paid by the sender and the receiver is unreasonable and likely to choke the development of such services.

Wireless

The rapid development and innovation in the unlicensed spectrum sector promises to bring the development of local auto-configuring mesh wireless systems where the infrastructure is owned by the customer.

Local groups of customers will peer naturally, and may then seek to use a wired or wireless backhaul to an exchange to peer with a wider area of users. Transit services would form a much smaller proportion of the traffic than at present under such an arrangement.

Pioneering examples of such meshes exist. MIT’s Roofnet, a project to create a self-organizing wireless network in which an amorphous, unmanaged collection of cheap Linux computers equipped with Wi-Fi cards collaborate to efficiently route data packets TR290803.

MIT Media Laboratory continues to research "core-less" viral communications which promise significant disruption to the provider/consumer centralised model of telecommunications.

The military use a product by 3e Technologies International which features robust security and auto-configuration of backhaul mesh connectivity.

Early research has shown that such networks have difficulty scaling, "Unfortunately, congestion as the system scales is unavoidable if the traffi is not limited to some degree of locality." Transfer to a wired highspeed network after a few hops can address this.

Such meshes can provide an environment for aggregating neighbourhood and suburban traffic. Peering can occur within such meshes, providing unmetered local traffic, as a consequence of purchasing the hardware and participating.

These aggregations can then join with wider networks to reach an Internet Exchange, where a range of services (including transit) and a wider peering population can be found.

Summary of Trends

The collective effect of the trends above is toward customer-owned networks, driven by the edge device, rather than a central service model of the legacy proprietary service network.

Both providers and consumers, and those distinctions can be expected to continue blurring, will communicate at speeds, established at very low cost, that will not make metering feasible or worthwhile.

In such circumstances, the market for carriage will be guaranteed, but for services will be extremely competitive, prone to large returns for success and heavy losses for failure.

Thus peering amongst content participants is likely to grow, while the control exerted by legacy service providers wanes.

The Internet Exchange is thus likely to become the core of the new network, but it will not, nor ever be, other than open and neutral. Any other model will be replaced by a more efficient Exchange, and open and neutral could be argued to be the optimal form of such an exchange.