Mr Marcus Wiko Consult Peering Qos Price and Quality

433 views

Published on

Published in: Technology, Business
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
433
On SlideShare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
0
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Mr Marcus Wiko Consult Peering Qos Price and Quality

  1. 1. Peering, QoS, and Price andQuality Differentiation J. Scott Marcus, Director IDATE, Montpellier, 16 November 2011 0
  2. 2. Peering, QoS, and Price and Quality Differentiation • Peering, transit, Internet access • Quality of Service (QoS) • The economics of discrimination • Traffic, costs, prices • Concluding observations 1 Peering, QoS, and Price and Quality Differentiation: IDATE, Montpellier, 16 November 2011
  3. 3. Peering, transit, and Internet access (1)• Transit - The customer pays the transit provider to provide connectivity to substantially all of the Internet. - Essentially the same service is provided to consumers, enterprises, ISPs, content providers, and application service providers.• Peering - Two ISPs exchange traffic of their customers (and customers of their customers). - Often, but not always, done without charge.• Variants of both exist. 2 Peering, QoS, and Price and Quality Differentiation: IDATE, Montpellier, 16 November 2011
  4. 4. Peering, transit, and Internet access (2) Peering Peering ISP A ISP B ISP C 3 ISP A2 ISP B1 ISP C2 ISP A1 ISP C1ISP A1a ISP C1b ISP C1a ISP A1b ISP A2 3 Peering, QoS, and Price and Quality Differentiation: IDATE, Montpellier, 16 November 2011
  5. 5. Peering, transit, and Internet access (3)• Meanwhile, unit prices for global transit are declining rapidly.• This decline reflects not only equipment costs but also circuits (over land and under water).• Labour and other OPEX elements play only a small role, since they depend mostly on the number of subscribers. Source: Telegeography (2011), WIK calculations. 4 Peering, QoS, and Price and Quality Differentiation: IDATE, Montpellier, 16 November 2011
  6. 6. Peering, transit, and Internet access (4) Shortest exit routing Backbone ISP New YorkPeering Chicago Peering Point San Francisco Washington DC Point Los Angeles Atlanta Dallas Web Site Backbone ISP Copyright (c) 1999 by Addison Wesley Longman, Inc. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior consent of the publisher. 5 Peering, QoS, and Price and Quality Differentiation: IDATE, Montpellier, 16 November 2011
  7. 7. Quality of Service (QoS) (1) General Applications Gaming Applications Stringent Latency RequirementsStringent Latency Requirements high Video high VoIP Telephony sports, racing, shooter Channel Surfing Audio role-playing IRC Streaming Video cards, Streaming Browsing, board games strategy Shopping, low Banking Email low FTP low high low high Bandwidth Requirements Bandwidth Requirements • Real time bidirectional audio or video: stringent requirements • Email: liberal requirements • Streamed audio and video: liberal requirements (channel surfing?) 6 Peering, QoS, and Price and Quality Differentiation: IDATE, Montpellier, 16 November 2011
  8. 8. Quality of Service (QoS) (2)• Voice drives revenue, but is a declining fraction of traffic.• Concerns have been voiced in recent years over the explosion of video traffic over the Internet, and its implications for network cost. Source: Cisco (2011). 7 Peering, QoS, and Price and Quality Differentiation: IDATE, Montpellier, 16 November 2011
  9. 9. Quality of Service (QoS) (3) M/G/1 Queuing Delay (155 Mbps Link) 350.00 300.00 Coefficient of Variation 250.00 0.00Wait Time (microseconds) 0.50 200.00 1.00 1.10 1.20 150.00 1.50 2.00 100.00 50.00 - 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 Utilization (rho) M/G/1 queueing analysis of the performance of a single link (with clocking delay of 50 μsecs (284 byte packets) and a 155 Mbps link) 8 Peering, QoS, and Price and Quality Differentiation: IDATE, Montpellier, 16 November 2011
  10. 10. Quality of Service (QoS) (4)• Per-hop delay, even at 90% load, is about 1,000 times less than the 150 millisecond delay “budget” for real-time bidirectional voice.• IMPLICATION: Most of the time, and under normal conditions, variable delay in the core of the network is unlikely to be perceptible to the users of VoIP or other delay-sensitive applications.• 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 9 Peering, QoS, and Price and Quality Differentiation: IDATE, Montpellier, 16 November 2011
  11. 11. Quality of Service (QoS) (5)• Although the technology is reasonably straightforward, little practical experience in enforcing QoS across IP-based networks.• It is not due to a lack of standards – there are too many standards, not too few.• Classic problem of introducing change into a technological environment: - Network effects – no value until enough of the market has switched. - Long, complex value chains. - Costs and complexity of transition.• Analogous problems have slowed IPv6 and DNSSEC. 10 Peering, QoS, and Price and Quality Differentiation: IDATE, Montpellier, 16 November 2011
  12. 12. Quality of Service (QoS) (6)• Efforts to extend Quality of Service (QoS) across network operators have failed to catch fire for many reasons: - Scale: Bilateral peering arrangements will tend to be acceptable to both network operators only when the networks are of similar scale, or more precisely when both networks can be expected to be subject to similar cost drivers for carrying their respective traffic. - Traffic balance: Where traffic is significantly asymmetric, cost drivers are likely to also be asymmetric. - Monitoring and management: There are many practical challenges in determining whether each network operator has in fact delivered the QoS that it committed to deliver. - Financial arrangements: There has been no agreement as to how financial arrangements should work. In particular, there has been enormous reluctance on the part of network operators to accept financial penalties for failing to meet quality standards. 11 Peering, QoS, and Price and Quality Differentiation: IDATE, Montpellier, 16 November 2011
  13. 13. Quality of Service (QoS) (7)• Many efforts over the years to define inter-provider QoS standards.• One of the best and most practical was organised by MIT, with substantial industry participation.• The following values from the MIT white paper would appear to be resonable for IP interconnection suitable for real time bidirectional voice: Delay: 100 msec Delay Variance: 50 msec IPPM Loss Ratio: 1 x 10-3 (One Way Packet Loss)• The MIT WG white paper also explains how to measure these, and how to allocate end-to-end requirements to multiple networks. IPPM probes could be suitable.• A challenge: No network operator will want another to operate probes within its network. 12 Peering, QoS, and Price and Quality Differentiation: IDATE, Montpellier, 16 November 2011
  14. 14. Quality of Service (QoS) (8)• As part of the functional/operational separation of Telecom New Zealand, there were commitments - To interconnect with competitors using IP - To support a suitable QoS for VoIP in those interconnections• The first of these is in place.• For the second, Telecom New Zealand made a quite interesting proposal, based on their methodology for the first of these. 13 Peering, QoS, and Price and Quality Differentiation: IDATE, Montpellier, 16 November 2011
  15. 15. Quality differentiation and network neutrality • Quality differentiation • Economic foreclosure • Two-sided (or multi-sided) markets 14
  16. 16. Quality differentiation• Quality differentiation and price differentiation are well understood practices.• In the absence of anticompetitive discrimination, differentiation generally benefits both producers and consumers.• We typically do not consider it problematic if an airline or rail service offers us a choice between first class and second class seats. 15
  17. 17. Economic foreclosure• When a producer with market power in one market segment attempts to project that market Google power into upstream or downstream segments that User would otherwise be competitive, Yahoo Commercial Broadband ISP ISP that constitutes economic foreclosure. Bing• Foreclosure harms consumers, and imposes an overall socio- economic deadweight loss on society. 16
  18. 18. Two-sided markets• The Internet can be thought of as a two-sided market, with network operators serving as a Google platform connecting providers of content (e.g. web sites) with User consumers. Yahoo Commercial Broadband ISP ISP• Under this view, some disputes Bing are simply about how costs and profits should be divided between the network operators and the two (or more) sides of the market. 17
  19. 19. Traffic, costs, prices, and profitability (1)• A.T. Kearney (2010): “Internet traffic is exploding in an unprecedented way due to increasing use of video. Costs for network operators are sky- rocketing, even under existing technology and even without considering the huge investments needed for fibre-based Next Generation Access. Due to market defects, there is no way to make consumers shoulder the cost of the increased bandwidth; thus, it will soon become necessary for firms that provide content to pay for the network for the first time, much as content and advertising typically pay for over-the-air broadcast television.”• Intuitive? Satisfying? Plausible? 18 Peering, QoS, and Price and Quality Differentiation: IDATE, Montpellier, 16 November 2011
  20. 20. Traffic, costs, prices, and profitability (2)• Traffic growth is largely a function of: - an increase in the number of subscribers, and - an increase in traffic per subscriber.• Some costs are largely driven by the number of subscribers, and are largely independent of usage per subscriber.• Unit costs for network equipment in the core and concentration networks (including routers and optoelectronics), where costs are usage-dependent, are declining at a rate comparable to that of Internet traffic increase per user in the fixed network. This can be viewed as an example of Moore‟s Law.• Cost per customer and revenue per customer in the fixed network remain in balance, despite the increase in traffic. 19 Peering, QoS, and Price and Quality Differentiation: IDATE, Montpellier, 16 November 2011
  21. 21. Traffic, costs, prices, and profitability (3)• The core network is about 7% of total cost, the concentration network about 6%.• Both benefit from these technological enhancements. Monthly cost of a bundled DSL broadband/voice service (BNetzA 2009) Wholesale price ULL Provision 6% Monthly rental ULL 21% Cost of DSLAM Cost of splitter 31% Transport in concentration network 9% Transport IP backbone network Collocation 3% 7% Usage dependent cost telephony 6% 14% Customer acquisition, maintenance, common cost, billing, bad debt 3% Source: German BNetzA (2009). 20 Peering, QoS, and Price and Quality Differentiation: IDATE, Montpellier, 16 November 2011
  22. 22. Traffic, costs, prices, and profitability (4) Fixed Data Traffic 40000 Traffic is indeed increasing in both 35000 the fixed and the mobile networks. 30000Data traffic/month (PB) Middle East and Africa 25000 Latin America Japan 20000 Asia Pacific 15000 North America Central Eastern Europe 10000 Western Europe 5000 0 2006 2007 2008 2009 2010 2011* 2012* Mobile Data Traffic 1400 1200 Source: Cisco (2011), WIK calculations. Data traffic/month (PB) 1000 Middle East and Africa Latin America 800 Japan Asia Pacific 600 North America 400 Central Eastern Europe Western Europe 200 0 2006 2007 2008 2009 2010 2011* 2012* 21 Peering, QoS, and Price and Quality Differentiation: IDATE, Montpellier, 16 November 2011
  23. 23. Traffic, costs, prices, and profitability (5)• However, the rate of growth in percentage terms is declining over time. Source: Cisco (2011), WIK calculations. 22 Peering, QoS, and Price and Quality Differentiation: IDATE, Montpellier, 16 November 2011
  24. 24. Traffic, costs, prices, and profitability (6)20,000 $25.0018,00016,000 $20.00 Here we have the shipment14,00012,000 $15.00 quantities in Mbps and the price10,000 8,000 $10.00 per Mbps (USD) for high end 6,000 4,000 $5.00 routers and for long haul DWDM 2,000 0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 $0.00 optoelectronic equipment. Worldwide high end router capacity shipped (Mbps) Price per Mbps (USD) These are among the key cost drivers for Internet core and 7,000 $7.00 aggregation networks. 6,000 $6.00 5,000 $5.00 The growth in shipments generally 4,000 $4.00 tracks the Cisco projections. 3,000 $3.00 2,000 $2.00 The growth in shipment volume 1,000 $1.00 does not equate to a growth in 0 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 $0.00 costs, because the decline in unit Worldwide long haul DWDM capacity shipped (Mbps) Price per Mbps (USD) costs is nearly in balance with it. Source: Dell‟Oro (2011), WIK calculations. 23 Peering, QoS, and Price and Quality Differentiation: IDATE, Montpellier, 16 November 2011
  25. 25. Traffic, costs, prices, and profitability (7)• The trend in underlying equipment costs (and many other costs) tracks subscribership and revenue, not with the volume of traffic. 2,000 1,800 1,600 1,400 1,200 1,000 800 600 400 200 - 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 IP traffic Fixed BB subscribers SP Edge total revenue SP Core total revenue DWDM Long Haul total revenue WDM Metro total revenue Source: Dell„Oro (2011), Cisco (2011), WIK calculations. 24 Peering, QoS, and Price and Quality Differentiation: IDATE, Montpellier, 16 November 2011
  26. 26. Concluding observations• In competitive markets, quality differentiation typically benefits both suppliers and producers.• Beware quick fixes! If a solution seems too good to be true, it probably is.• Market mechanisms often reach better solutions than well intentioned policymakers. 25 Peering, QoS, and Price and Quality Differentiation: IDATE, Montpellier, 16 November 2011
  27. 27. wik-Consult GmbHPostfach 200053588 Bad HonnefTel 02224-9225-0Fax 02224-9225-68eMail info@wik-consult.comwww.wik-consult.com

×