310 likes | 430 Views
This paper, presented by Keith Mitchell, Executive Chairman of the London Internet Exchange (LINX), addresses the significant changes in the Internet peering environment as of December 1998. It explores key trends like the consolidation of the ISP market, the growth in Internet Exchange Points (IXPs) and their impact on competition, as well as the dynamics of peering agreements. The insights cover the importance of regulation in fostering fair competition and the role of settlement models in shaping the peering landscape.
E N D
The ChangingPeering Environment Keith Mitchell keith@linx.net Executive Chairman, London Internet Exchange Internet Peering 3rd December 1998
Overview • Trends • Is the market consolidating ? • Peering & Settlement • Experiences at LINX • Competition
Some Trends • Growth in IXP numbers: • handful in 1993 • over 100 end 1998 • distribution typical (~50% in US) • more “local” exchanges “IXP” = Internet eXchange Point
Telephony Tariff Trends • Much retail voice telephony traffic is becoming “too cheap to meter” • e.g. • inclusive minutes • Internet telephony • national “L” rate 0845 in UK • free local call regime standing up to Internet dial-up demands in US
Tariff Convergence • Trend is not simply of voice settlement model being adopted by Internet • Evidence that as bandwidth demands increase, it is not always economic or practical to count every last bit
Has the ISP market Consolidated ? • A few very big playersmerging mergingmerging • Medium players getting larger • Total number of ISPs is staying constant • So there must be new entrants
Is the ISP market Consolidating ? • New players are seeing market opportunities: • due to new technologies • in particular market segments • Is this different from other industries ?
Will the ISP market Consolidate ? • Biggest difference from other industries is issue of “capture” • Key emergent resources and technologies need protection from vested interest control • Stifling competition stifles innovation • Key is strong regulation of potential capturing monopolies • e.g. Microsoft, NSI
Settlement & Competition • Non-settlement exchange can inhibit competition by preventing peering where there is no other basis for agreement • Settlement exchange can inhibit competition by encouraging revenue flow from small to large players
IXPs & Competition • Exchange point andco-location facilities promote fair competition: • reduce supplier lock-in • lower barriers to entry • increase transparency
UK State of Play • Most peering via LINX • majority of members do peer (~80%) • disputes unusual • Some smaller players via2 regional exchanges • Limited (<10) private bi-lateral peerings • Lots of settlement-based bi-lateral wholesale/transit
LINX Background • LINX is UK nationalInternet Exchange Point • Represents 60 largest UK++ ISPs • Tries to encourage open peering and competition between ISPs • Promotes self-regulation (e.g IWF), but is not “regulator”
LINX Peering Environment • Published & well-defined membership criteria • Minimum of interference in member peering autonomy • Peering agreements private matter between members • Incentives to peer
LINX Peering Practice (1) • Members must peer with at least: • one other member • LINX routers • In order to: • to acquire voting rights • to remain member after 3 months
LINX Peering Practice (2) • Members must: • publish peering contacts • respond to peering requestswithin 2 days • Peering matrix on web page helps end-users put pressure on • Independent staff can intervene • Template “standard” peering agreement bring worked upon
Good Peering Practice • “Self-regulatory” measures • Peering policies should be: • registered • in public domain • consistently & fairly implemented • stable
Settlement at the LINX • Recently removed restriction on settlement-based peering • Historical rule set by founder members • Was inhibiting some large players from joining • Potential regulatory concern area • Not consistent with principle of non-intervention in peerings
Alternatives to Settlement • Tiered peering used by some and works well: • exchange of subset of customer routes/territory • multiple ASes/routing policies • or bandwidth limited • Fixed fee peering
QoS and Settlement • Quality of service commonly advocated reason for need for settlement-based peering • It is a red herring • Existing complex inter-provider arrangement are at manageability limit already • This will be a minority premium service, not the norm
Quality of Service Issues • Inter-provider routing on Internet is policy based, uses BGP • BGP does not support QoS • QoS routing (e.g. RSVP) cannot do inter-provider policy routing • So only practical inter-provider QoS mechanism is link rate limiting
QoS in Practice • 3 ways to do link rate-limiting: • private point-to-point links • ATM switches • new generation of LAN (ethernet) switches
Where do IXPs fit in ? • Within EU, “national” IXPs should be needed less as telecoms market opens • Small number of pan-European IXPs emerging e.g. AMS-IX, DGIX, DE-CIX, LINX, VIX all have 25-35% non-domestic members
IXP Growth • Many of the new IXPs are local/regional e.g. MaNAP, LoNAP, Scot-IX in UK; many in US • Motivated by strong IXP/co-lo synergy (e.g. PAIX) • Often intent is to stimulate local Internet economy rather than simply efficient traffic exchange
Growth in Co-location • Major boom industry at present • e.g. Telehouse, Compaq, Co-lomotion • TeleCity in UK major venture-capital funded start-up • No less than 8 responders to LNX 2nd site tender less than 10km apart ! • PAIX is research project that makes profit !
Growth in Co-location • Remains much unmet demand forco-lo space outside UK & US • ISPs want quality space • Web hosting ISPs want lots of quality space • Co-lo customers want choice of ISPs & carriers in facilities • Existing facilities filling up
Co-Lo Growth & Peering • Private bi-lateral peering is not simple or cheap across multipleco-lo sites • Cheap dark fibre may be one solution e.g Stockholm, Palo Alto,almost Docklands • Is scaling this future role for IXPs ?
Conclusions • Open peering can promote competition • Closed bi-lateral exchange can inhibit it • Open peering arbiter can facilitate competition: • as L1/L2 exchange • as organisational environment
Summary • Market will determine: • when • whether • where settlement or not is best • There is room for diversity • Attempts to buck market will probably fail