Traffic Types and Growth in Backbone Networks Alexandre Gerber, Robert Doverspike AT&T Labs Research
Outline Overview of a US carrier inter-city backbone optical network: Services running on ROADMs Breakdown of IP services by business This is not an industry-wide view, rather a single US ISP data point IP Services drill down: Application Mix Growth over the last decade Implications on content distribution efficiency: CDNs Migration of TV content and Multicast
What s on the top of fibers in backbone? What do we mean by traffic type? Web Video OSI 7 layer model Application Presentation Session Transport Network Data Link Physical Reality HTTP TCP RTP IP GTP UDP IP MPLS VPN MPLS Ethernet VLAN SONET/SDH/OTN Ethernet PHY DWDM 3 Fiber
Private Line vs. IP Services Simplified depiction of core-segment network layers Private Line Services (low rate) Private Line Services (high rate) 60% Wideband DCS Layer IP Services IOS Layer IP Layer SONET /SDH Ring Layer 40% Layer with automatic restoration Layer without automatic restoration Private Line Services (line rate) DWDM = Dense Wavelength Division Multiplexing Broadband DCS Layer 15% 42% 43% ROADM / Pt-to-pt DWDM Layer ROADM = Reconfigurable Optical Add/Drop Multiplexer Fiber Layer DCS = Digital Cross-Connect System IOS = Intelligent Optical Switch 4 Source: US ISP, Summer2010.
What are these IP Services? Wireline Broadband Services of ISP 29% Wireless Services of ISP (Data Center Backhaul) 45% Wireless Services of ISP (Data Center To Internet) MPLS VPN (inc. VoIP) 17% 4% 5% Business Internet Access and Hosting (inc. foreign and other US Wireline & Wireless Braodband Services) Takeaway: Wireline Broadband Services is the largest source of traffic. Business traffic accounts for a large share, too. Wireless is not significant yet. Source: US ISP, Average traffic. Summer 2010. 5
7/31/2004 10/31/2004 1/31/2005 4/30/2005 7/31/2005 10/31/2005 1/31/2006 4/30/2006 7/31/2006 10/31/2006 1/31/2007 4/30/2007 7/31/2007 10/31/2007 1/31/2008 4/30/2008 7/31/2008 10/31/2008 1/31/2009 4/30/2009 7/31/2009 10/31/2009 1/31/2010 4/30/2010 Normalized Peak Downstream Traffic per subscriber 2/17/2001 6/17/2001 10/17/2001 2/17/2002 6/17/2002 10/17/2002 2/17/2003 6/17/2003 10/17/2003 2/17/2004 6/17/2004 10/17/2004 2/17/2005 6/17/2005 10/17/2005 2/17/2006 6/17/2006 10/17/2006 2/17/2007 6/17/2007 10/17/2007 2/17/2008 6/17/2008 10/17/2008 2/17/2009 6/17/2009 10/17/2009 2/17/2010 6/17/2010 10/17/2010 Normalized Average Downstream Traffic per subscriber 2 components of IP Services growth A. Organic growth per Broadband subscriber Takeaway: stable organic growth over the last couple of years (32% CAGR). Slight acceleration over the last 2 years. Normalized Average Traffic per sub 2001-2010 16 14 12 6 Normalized Peak Traffic per sub 2004-mid2010 10 8 6 5 4 4 2 0 3 2 1 0 Takeaway: this is a good data point to predict the long term growth rate of the Internet. Source: US DSL Downstream traffic per subscriber. 6
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Broadband Penetration rate (Year over year growth rate) 2 components of IP Services growth B. Growth of Broadband penetration rate 100 90 80 70 60 50 40 30 Takeaway: Average growth rate of 31%/year from 2000 to 2009 in the US. Decreasing from 48% to 4% as we get closer to market saturation Japan USA OECD United Kingdom 20 10 0 Source: OECD. 7
Jan-98 Jun-98 Nov-98 Apr-99 Sep-99 Feb-00 Jul-00 Dec-00 May-01 Oct-01 Mar-02 Aug-02 Jan-03 Jun-03 Nov-03 Apr-04 Sep-04 Feb-05 Jul-05 Dec-05 May-06 Oct-06 Mar-07 Aug-07 Jan-08 Jun-08 Nov-08 Apr-09 Sep-09 Feb-10 Jul-10 Dec-10 Normalized Average Traffic (log scale) IP Services growth A US ISP view over the last 12 years 10000 Takeaway: combining the 2 types of growth, we see a growth of almost 4 orders of magnitude on that backbone over the last 12 years (blue) with a clear saturation over time. 1000 100 10 1 Takeaway: Cellular traffic follows a similar pattern approx 8 years later (red) 8 Source: US ISP. Monthly average backbone traffic.
4Q2002 2Q2003 4Q2003 2Q2004 4Q2004 2Q2005 4Q2005 2Q2006 4Q2006 2Q2007 4Q2007 2Q2008 4Q2008 2Q2009 4Q2009 2Q2010 What is that traffic? Layer 4 Protocol breakdown 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Web/HTTP (inc. Multimedia) P2P Other NetNews Non HTTP Multimedia Mail Games Business Takeaway: HTTP is back! It is now the workhorse protocol of many applications and it accounts for 60% of the traffic. Source: US ISP Backbone. Netflow. 9
Feb-09 Apr-09 Jun-09 Aug-09 Oct-09 Dec-09 Feb-10 Apr-10 Jun-10 Aug-10 Normalized Traffic Volumes What is that traffic? Layer 7 Protocol breakdown: the rise of video 6.0 5.0 4.0 3.0 2.0 1.0 0.0 Other P2P Non HTTP Multimedia HTTP Multimedia Web/HTTP (excluding Multimedia) Source: US DSL Downstream Traffic per subscriber during the busy hour. 10 Takeaway: Video traffic is running on the top of HTTP. Lately it has been growing at an annualized growth rate of 83%. How will it evolve?
Implications of video growth Rise of the Content Distribution Networks (CDN) 100% 90% 80% 70% 60% 50% 40% 30% CDN Other 20% 10% 0% Backbone Wireline Broadband Services Wireless Takeaway: A few CDNs account for 39% to 55% of the traffic during the busy hour. Source: US ISP, August 2010. CDN traffic identified in Netflow records based on IP addresses. 11
Normalized On Net Air Miles Are CDNs doing a good job? Yes! 3.5 3 2.5 2 1.5 1 0.5 0 CDN Top 5 Content Providers All Web traffic (incl. CDN) All P2P traffic Takeaway: The distance traversed by CDN traffic on the backbone is 3 times shorter than the distance traversed by other content providers. Source: US ISP, August 2010. OnNet traffic only (no hot potato peering traffic). Average air miles defined as the distance between the ingress access router and the egress access router weighted by the amount of traffic. 12
Downstream traffic per household during the busy hour (kbps) But TV traffic is delivered much more efficiently today than Over the Top Videos 1200 1000 800 600 Takeaway: Internet traffic is still much smaller than TV content consumed today. That TV content is distributed today much more efficiently via Broadcast or Multicast (IPTV). 400 200 0 Internet traffic TV Takeaway: The growth rate of backbone traffic will be determined by how much of that TV traffic will be migrated onto the Internet, how fast and how it will be distributed (e.g. Multicast vs. Unicast)? Source: Internet: Cisco VNI 2010: 14.9 GB/month/sub. Peak/average ratio 1.72. Video: assumes 2Mbps video stream to 50% of the households during the peak. 13
Summary What is running on a US carrier backbone optical network? The largest single source of traffic is video traffic for Wireline broadband subscribers over HTTP over TCP over MPLS over IP over ROADM. But there are many more traffic types: Private Lines, MPLS VPN, etc. How fast is the traffic growing? Combination of a stable organic growth of 32% per year with a deceleration of the growth of number of Wireline Broadband users. Strong growth of video traffic recently; overall organic growth recently above average What will happen to that video traffic? In the hands of a couple of CDNs that deliver traffic more efficiently than other content providers. But there is much more video traffic out there. If it migrates at a fast pace from Broadcast/Multicast to Unicast streams on the Internet, it will significantly impact backbones. 14
Thank You! Alexandre Gerber gerber@research.att.com AT&T presence at OFC: http://www.research.att.com/conferences/ofc 15