About this Journal Submit a Manuscript Table of Contents
International Journal of Digital Multimedia Broadcasting
Volume 2010 (2010), Article ID 562832, 10 pages
http://dx.doi.org/10.1155/2010/562832
Research Article

Bandwidth Reduction via Localized Peer-to-Peer (P2P) Video

1Telcordia Technologies, One Telcordia Drive, Piscataway, NJ 08854-4151, USA
2Huawei, USA

Received 1 June 2009; Accepted 14 October 2009

Academic Editor: Stan Moyer

Copyright © 2010 Ken Kerpez et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

This paper presents recent research into P2P distribution of video that can be highly localized, preferably sharing content among users on the same access network and Central Office (CO). Models of video demand and localized P2P serving areas are presented. Detailed simulations of passive optical networks (PON) are run, and these generate statistics of P2P video localization. Next-Generation PON (NG-PON) is shown to fully enable P2P video localization, but the lower rates of Gigabit-PON (GPON) restrict performance. Results here show that nearly all of the traffic volume of unicast video could be delivered via localized P2P. Strong growth in video delivery via localized P2P could lower overall future aggregation and core network bandwidth of IP video traffic by 58.2%, and total consumer Internet traffic by 43.5%. This assumes aggressive adoption of technologies and business practices that enable highly localized P2P video.

1. Introduction

IPTV, IP video on demand (VOD), over-the-top streaming, video downloads, and user-generated content are all becoming increasingly popular with the proliferation of broadband. With the exception of multicast IPTV, these video streams are sent unicast (with a dedicated stream for each source-destination pair) through core and aggregation networks, and together they can consume tremendous amounts of bandwidth. Worse, video streams cannot benefit from statistical multiplexing concentration the way that most Internet data does, because of the need to support simultaneous constant streams at the prime-time peak.

Another trend is the continued popularity of peer-to-peer (P2P). In many cases, P2P traffic traverses long distances, across core networks and multiple Internet Service Provider (ISP) networks, even though the content could have been retrieved from a much closer location. There has been a spate of recent effort to localize P2P traffic, as discussed in Section 2.2 here. Localized P2P traffic may only traverse a few hops instead of ten or twenty, allowing for a vast decrease in core network bandwidth.

This paper expands and extends previous work examining bandwidth implications of localized P2P [1]. This paper focuses on the maximum localization that is generally possible to localize traffic within access networks and central-office (CO) serving areas.

High P2P video traffic volumes can be accommodated across optical access networks; and it is shown here that localized P2P video is readily enabled by the Next-Generation Passive Optical Network (NG-PON) as envisioned by the Full-Services Access Networks (FSAN) Group and others.

Results here show that nearly all of the traffic volume of unicast video could be delivered via localized P2P. Strong growth in video delivery via localized P2P could lower overall future aggregation and core network bandwidth of IP video traffic by 58.2% and total consumer Internet traffic by 43.5%. This assumes aggressive adoption of technologies and business practices that enable highly localized P2P video.

2. Driving Trends and Enabling Technologies

2.1. Traffic Growth: P2P and Video

Video and peer-to-peer contents are increasing Internet bandwidth demands. Reports from the Distributed Computing Industry Association (DCIA) [2] predict an “exaflood” from advances in rich media content delivery, particularly via peer-to-peer (P2P) technologies, along with user-generated content (UGC) and video. Recent data [3] shows both global and US Internet bandwidth growing at approximately 40% to 60% per year.

P2P traffic occupies much Internet bandwidth. It was found that 62% of the total volume of Internet traffic was used by P2P in Japan [46]. A small segment of users dictate the overall behavior; 4% of heavy-hitters account for 75% of the inbound traffic volume.

Ellacoya Networks studied European Internet traffic in 2005 [7], and found that usage by the top most active 5% of subscribers represented approximately 56% of total bandwidth, while the top 20% of active subscribers consumed more than 97% of total bandwidth. P2P was by far the largest consumer of bandwidth with 65.5% of traffic on the network being P2P applications.

CacheLogic conducted direct packet monitoring of Internet backbones and ISPs data streams via Layer 7 packet analysis [8] in 2005. This study found 61.4% of peer-to-peer traffic to be video, 11.4% audio, and 27.2% other traffic.

Ipoque [9] analyzed the Internet traffic in five regions of the world between August and September 2007. The results for these different regions vary considerably. P2P produced, on average, between 49% and 83% of all Internet traffic. From 47% to 79% of the exchanged P2P traffic was video. BitTorrent and eDonkey were by far the most commonly used P2P systems. Multimedia Intelligence [10] predicts P2P traffic will grow almost 400% over the next 5 years.

Some recent reports suggest that the percentage of Internet traffic which is P2P may have dropped somewhat. DSL Prime [11] reported that P2P dropped from about 40% of Internet traffic in 2007 to between 20% and 26% in early 2008. Streaming, largely YouTube, appears to have taken up the slack. Overall Internet traffic growth was reported at 30–40% percent per year.

On the other hand, some reports show P2P traffic share increasing. Sandvine [12] presented recent statistics of Internet traffic in North America in 2008. In aggregate, consumer broadband traffic was 44% P2P, a slight increase from 2007 when 40.5% of traffic was P2P. This increase could perhaps be attributed to improved DPI recognition of P2P. In aggregate, web browsing was 27% and streaming was 15% of North American consumer broadband traffic. P2P, web browsing, and streaming constituted much of the downstream traffic, while P2P dominated upstream with 75% of the upstream traffic.

Cisco [13] has presented a detailed forecast with numbers similar to the other studies but more comprehensive. Cisco projects 49% annual growth from 2007 to 2012 in overall Internet traffic, with only 31% annual growth for P2P but with 68% annual growth in Internet video plus IPTV traffic. Cisco estimated that 60% to 70% of P2P traffic is actually video. As seen in Figure 2, Internet video alone is expected to account for 50% of all consumer Internet traffic in 2012, in addition to P2P video and IPTV.

562832.fig.001
Figure 1: Localized P2P.
562832.fig.002
Figure 2: Cisco [13] forecasts of Global Consumer Internet Growth. One ExaByte =  Bytes.

Forecasts show strong growth in Internet video. In-Stat [14] has recently forecasted that 160 billion user-generated served videos will be served in 2012. User-generated video revenue is projected at $1.2 billion by 2012. IPTV growth projections are also robust; Gartner projects that in one year (2008) there will be a 64.1% increase in global IPTV subscribers and 93.5% increase in IPTV revenue to 4.5 billion. Recent forecasts from Strategy Analytics predict that IPTV service revenues in the USA will grow from 694 million in 2007 to nearly 14 billion in 2012 [15].

The underlying trend is continued growth in P2P Internet traffic, and strongly increasing growth in video traffic. Much video is now being streamed from servers, from YouTube, or from content owners such as CBS, ABC, NBC, Netflix, and so forth. If more video shifts to P2P, then we can expect very large increases in P2P traffic.

2.2. Localizing and Legitimizing P2P

Localized P2P can save network bandwidth by delivering titles from a source that is closer than currently used peers, and perhaps even closer than the nearest network-owned video server [1]. P2P systems that prefer to get content from the most local sources should have traffic that traverses fewer links. An article [16] showed that 99.5% of current P2P traffic (using “eDonkey” in France) traversed national or international networks. It further showed that 41% to 42% of this long distance traffic could be made local if a preference for local content was built into the protocol.

Methods for localizing P2P have been discussed. It is not uncommon to use the IP hop count or TTL value to localize somewhat when choosing peering sources. Current routing schemes in P2P networks such as Chord [17] work by correcting a certain number of bits at each routing step. Reference [18] used the IP number to localize, and found that the first octet in the IP number provided localization to roughly a national level, an improvement over global delivery. Reference [19] accesses hosts in the same domain of the DNS namespace, for example, in the same country, which are often also geographically closer than hosts in other domains or countries. Oversi has a product called NetEnhancer that automatically routes P2P traffic to local peers, and away from interconnect peers and overloaded access networks.

P2P may work in concert with traditional content network delivery. References [20, 21] both focus on methods that allow partial control of P2P traffic by network providers, with [20] focusing on P2P. Pando Networks Inc. and Kontiki are companies now offering peer-assisted content delivery. A recent popular method of network-provider control for P2P is called “Proactive network Provider Participation for P2P (P4P).” P4P [21, 22] provides network information (called “itrackers”) to distributed P2P systems which then optimize their traffic through the network. Pando Networks claims that P4P tests at Verizon, Comcast, and AT&T sped delivery and increased the percentage of data routed internally. The percentage of data delivered within each ISP increased from fourteen percent (14%) with normal P2P delivery to as much as eighty-nine percent (89%) with P4P delivery.

The DCIA and Pando Networks have formed the P4P Working Group (P4PWG) to help specify P4P. The IETF P2PSIP Working Group is working to standardize SIP signaling for P2P and user and resource location data in an SIP environment with no or minimal centralized servers. The IEEE Next Generation Service Overlay Network (NGSON) Working Group may help enable carrier-grade services delivery over P2P. NGSON considers a control-plane “network overlay” for controlling P2P systems. These P2P overlays can be broadly categorized as three types: Structured (based on network connectivity and number of hops), Hierarchical (with multiple autonomous groups), and Semantic (using hashed-based loop-up routing).

Recent reports say that P2P will be integrated into traditional content distribution [23]. P2P Content Distribution Network (CDN) leader Akamai has recently bought the company Redswoosh and now has their P2P distribution system. Abacast is a hybrid CDN, combining the advantages of P2P, both for live streaming and on-demand downloads.

Digital rights management (DRM), licensing, and copy control systems are emerging for P2P content [24, 25]. New PCs have powerful processors that can perform strong encryption and help enable distributed content management. Reference [26] defines a distributed P2P content delivery capability with authentication, personalization, and payment functions that can increase carrier and content owner revenues.

It’s predicted that the bulk of P2P will become content that is legitimate and licensed [2]. P2P is predicted to become a legitimate mainstream distribution method, and licensed P2P content is expected to grow 10 times as fast as unauthorized P2P traffic [10]. Some content filtering engines are said to be able to detect and interdict illegal content distribution at any point in a network. MediaDefender and Cloudshield can be used for content protection on P2P. Brilliant Digital Entertainment has developed an application for Internet service provider (ISP) networks that is said to automatically and rapidly identify illegal files.

2.3. P2P Video

Much current P2P traffic is video, usually file downloads. In the future, P2P may take a more central role in delivering Internet video and IPTV [27]. P2P can be employed to enable non-real-time download of video on demand (VOD). Personal Video Recorders (PVRs) in subscribers’ set-top boxes that are deployed by network or service providers could host P2P video in order to offload traffic from network video servers. A number of control, oversight, and copy protection issues are being addressed to make this practical [28]. Other work has enabled fault tolerant and cost-effective P2P Video on Demand (VOD) [29, 30]. Studies have explored the bandwidth and performance of time-shift TV (TSTV), which allows users to pause, fast forward, or rewind a title that was viewed elsewhere [31].

Streaming video with P2P poses unique delay and bandwidth challenges. New work is enabling P2P to stream linear broadcast TV by using application-layer multicast [32]. Router-level multicast infrastructure is used by network providers to limit core bandwidth, and multicast can be combined with P2P in interesting ways, for example allowing “VCR” controls for P2P linear broadcast service [33, 34]. Most successful P2P streaming application-layer multicast systems use fairly simple tree structures, but more recent work is looking at using multiple trees, meshes [35, 36], data chunks, and swarms. Recent measurements of a large-scale live streaming video P2P multicast reported that most traffic could be offloaded by P2P, with about 15% of traffic still streamed from servers [37].

Popular P2P protocols, as exemplified by BitTorrent, download many different chunks of a file from many different peers, lowering the impact of individual upstream bottlenecks. Users form a “swarm” to share content, and use a “torrent file” that identifies pieces of data to retrieve in the swarm. BitTorrent's non-contiguous download methods do not appear to be suited to real-time streaming, however, recent work has modified BitTorrent to “LiveSwarms” [38] which enables P2P streaming. The “Swarm Player” is being developed as an open source project for P2P streaming.

There are now a number of companies and platforms that provide P2P streaming including Joost, Veoh, Nextshare, Vudu, PiCast, Vatata, End System Multicast (ESM), Gridmedia, PPStream, PPLive, Zattoo, Octoshape, Sopcast, Tvkoo, Roxbeam, Tribler, Ustream, RedSwoosh, Mediamelon, RawFlow, Selfcast.com, Velocix, NeoKast, BitTorrent, uTorrent, Brightcove, and so forth. [2]. The Vudu Internet set-top box is said to predownload the first few minutes of the most popular movies to give the viewer a “head start” using a sort of proprietary P2P technology to retrieve content from other Vudu boxes.

The European P2P-Next consortium has created a set-top box (STB) prototype that allows P2P streaming, which is called “NextshareTV” and was developed at the Pioneer Digital Design Center in London. NextshareTV is being developed for PCs as well as STBs, and it makes use of some of the open-source P2P streaming technology of the Swarm Player. Octoshape has P2P video streaming technology called the “grid delivery system” that is said to enable HDTV (at about 2.5 Mbps).

There is clearly much current interest and many new developments in localized P2P, network-controlled P2P, P2P video, and P2P streaming.

3. Bandwidth Evaluation Methodology

The remainder of this report shows calculations of video bandwidth usage and the impact of localizing P2P video. Models are created representing typical usage.

3.1. Video Demand Model

The most popular video titles are viewed far more than the least popular. This is particularly true for TV shows, first-run movies, and recently released DVD rentals. TV, movie, and video rental rating statistics [3941] were examined, and an overall model for the demand of video titles was created by matching these statistics. The model first rank orders all video titles, from most popular to least popular. The title number increases with decreasing popularity. A probability density model is assigned to the title numbers. A series of curve fits were made to several datasets of number of viewers of different types of video content, with the video model here constructed as a composite of these curve fits to represent all overall video usage.

Long-tailed distributions often follow a “Zipf-law” probability distribution, also known as a power-law or hyperbolic probability. The video demand model here uses this power-law to fit the long tail, formally called the hyperbolic probability density. A recent paper [42] also showed that the power law is a good fit to Netflix and YouTube video demand, which are both very long tailed. The YouTube data matched a power-law with correlation 0.8. Statistics for higher title numbers, out in the very long tail, are represented by Netflix [42] and video rental data here. The long tail is somewhat supply-driven: as more titles are offered, someone will eventually view them.

On the other hand, some video content is not long tailed and is instead highly weighted toward the most popular titles. It was found that the popularity of linear broadcast TV channels and new movie releases are modeled closely by an exponential probability density function, with a few titles very popular and the less popular content rapidly dropping off with increasing title number. These include weekend movie gross.

A combined video demand model is used here; this is a mixture of exponential and hyperbolic probability densities. This model is a slightly modified version of a previous model [1]; there is more weight on the tail in the new model to include the Netflix data. The statistics and model are shown in Figure 3 for popular titles with low title numbers.

562832.fig.003
Figure 3: Video demand statistics and model. Statistics and probabilities are normalized so that the probability density of title 1 = 1.

The model density is a combination of hyperbolic (Zipf or power-law) and exponential densities. The hyperbolic component models the long-tail Netflix data, while the exponential component models content such as new movie releases. The density is truncated to limit to a finite number of available titles, title such that The mathematical definition of the model is with where is an indicator for a Bernoulli random variable; Pr Pr and satisfying

Further, the truncated hyperbolic density is where A is a constant, and The mean of the truncated hyperbolic probability density is The truncated exponential density is where is a constant, and The mean of the truncated exponential density is The overall mean is The TV title demand model parameters used in numerical evaluations here are For all results in this report, the top 20 most popular titles are broadcast/multicast, the rest (64%) is unicast VOD or P2P. The broadcast/multicast titles are not considered here other than that they reduce the overall demand for unicast video. The total available number of different video titles is 10 000 here.

3.2. Number of Titles Demanded by Each Subscriber

Another aspect is the number of video streams that are viewed by each subscriber. Statistics [40] show that the maximum busy hour (prime time peak) has about 66% of all households watching TV. Also, each home averages a little less than two simultaneous TV viewings per home in the busy hour.

According to FCC statistics in 2005, 90% of US homes had TVs, and 67.5% subscribed to cable TV. The average number of TV sets per household was 2.62. The average TV viewing hours per day are listed in Table 1.

tab1
Table 1: TV viewing hours per day in 2003, FCC data.

IP video subscribers may watch a little more video than the average person. A simple discrete and independent model for the number of video viewings per home suffices here, an example of which is in Table 2. The average total number of streams to each subscriber is 1.8.

tab2
Table 2: Model used in analyses here of probability of number of video streams per IP video subscriber in the busy hour (prime time). Average = 1.8 streams per sub.
3.3. Serving Area Sizes

Previous work [41] showed that a Gamma probability can closely model statistics of telecom serving area sizes. The video demand model presented here uses a Gamma model to determine the number of subscribers per Central Office (CO).

The gamma probability density function (pdf), defined as Define the mean of the Gamma to be and the standard deviation to be Then Here the gamma model determines serving area radii. Given the radius, of the serving area, a uniform subscriber density is assumed and the number of subscribers in the serving area is simply the average subscriber density multiplied by the serving area size,

The gamma model parameters used for modeling CO-serving area size here are as in Table 3 with: resulting in the probability shown in Figure 4, which is truncated to a maximum value due to computer memory limitations in the evaluations here.

tab3
Table 3: Simulated serving area sizes for P2P results.
562832.fig.004
Figure 4: Model of CO-serving area sizes.
3.4. NG-PON Limits

P2P traffic could exhaust streaming and upstream capacity, and so realistic capacity limitations on P2P traffic are imposed in this study. Downstream capacity is not considered here because downstream bandwidth would be used anyway if the video was network-provided and not P2P. Each peer is assumed to be capable of simultaneously sourcing at most two P2P video streams. This work is forward-looking, so all video is assumed to be HDTV at 19.3 Megabits per second (Mbps) per stream.

P2P should not be allowed to exhaust all resources, and so up to at most 1/3 of any resource is allowed to be used by P2P.

Most results are for a next-generation passive optical network (NG-PON) access network, which is assumed to have the following capacity limitations:

(i)Each Optical Line Terminal (OLT) port can support 32 subscribers.(ii)Each OLT port has upstream bandwidth of 2.488 Gbps, which can support 42 P2P video streams.(iii)The OLT switching capacity is 400.0 Gigabits per second (Gbps), which can support 6908 P2P video streams.(iv)The OLT uplink bandwidth = 40.0 Gbps, which can support 690 P2P video streams(v)The CO-router throughput is one Terabyte.

In any cases where these limits would be exceeded, the demanded video streams are not served by P2P but instead by network servers. Much of the localized P2P traffic is served by, and goes to, subscribers on the same OLT, which would use no OLT uplink capacity, and so the OLT uplink capacity usually does not limit P2P here.

4. Bandwidth Impacts

Monte-Carlo simulations here repeatedly randomly generate CO-serving areas, video demand, P2P supply, and so forth, and collect statistics on P2P bandwidth usage. The stochastic models described in the previous section are used for serving area sizes, numbers of P2P titles, title selection, and storage. PON, OLT, and upstream streaming sizing are held constant to represent one fiber connection per home, with one fiber service provider.

A number of subscribers per CO is randomly generated by the model of Section 3.3, then these subscribers are assigned to OLT ports, and OLTs. An OLT supports a randomly-generated number of subscribers that is uniformly distributed between 1024 and 2048. OLT ports are filled with 32 subscribers. Each subscriber is randomly assigned some number of simultaneously demanded video streams according to the model of Section 3.2. A title is assigned to each demanded video stream using the model of Section 3.1.

A percentage of all subscribers are each assumed to store some number of P2P titles. The identity of these stored titles is determined by the same model as used for video demand in Section 3.1. Each of these stored titles can be delivered if demanded by other subscribers, from the closest source that stores the demanded title and has capacity for another stream. The simulation first searches to see if the demanded title can be served from a subscriber on the same OLT port (the PON); if there are none then all subscribers on the OLT are searched, if there are none then the all subscribers in the CO are searched (the CO-router area), and then if there are none the title is delivered through the core and aggregation network from the headend (not localized).

Statistics count up the number of video streams in each segment of the network. Simulations regenerate all serving area sizes, demanded video titles, stored P2P titles, and so forth, 250 times.

The most popular titles are highly likely to be demanded, and they are highly likely to be stored for P2P availability. This results in the localized P2P system here very often delivering content from nearby sources. If only very long-tailed contents were available for P2P, then less traffic would be localized.

4.1. Detailed Results

A P2P stream is switched or routed from upstream back to downstream at the “hairpin” location. Figure 5 shows results in terms of number of unicast video stream per CO in the busy hour (recall that the top 20 titles are broadcast/multicast and not included here). These figures vary the number of different video titles that are stored at each peer and are available for streaming to another peer along the -axis. Recall that each peer can only transmit two P2P streams simultaneously. As more titles are stored, it is more likely that any demanded title can be transmitted from a local peer. As the number of stored titles available from each subscriber increases, the most local traffic at the same OLT port increases, and the most distant traffic from non-P2P unicast servers decreases. Traffic hairpin routed at the intermediate local points, the OLT chassis and the CO router, rises and falls.

562832.fig.005
Figure 5: Detailed results: average numbers of streams. NG-PON.

The most telling curve in Figure 5 is the number of non-P2P unicast streams beyond the CO (from headend servers). With NG-PON and localized P2P, these aggregation and core network streams can drop to nearly zero, less than 0.1% of the total unicast video traffic. This is because the most popular video titles are also the most commonly stored, and demand can be met via P2P. Background traffic has little effect on these results since the GPON resources are sufficient.

562832.fig.006
Figure 6: Number of streams with GPON limits on upstream P2P Capacity.
4.2. GPON Compared to NG-PON

Simulations were run identical to previous Sections 3 and 4.1, except with access network bandwidth limitations representative of a Gigabit-Passive Optical Networks (GPON):

(i)1.244 Gbps per-PON upstream instead of NG-PON 2.488 Gbps,(ii)10 Gbps OLT uplink instead of NG-PON 40 Gbps,(iii)200 Gbps OLT backplane instead of NG-PON 400 Gbps.

Results are shown in Figure 6. Unlike NG-PON, the number of non-P2P unicast streams beyond the CO (from headend servers) does not approach zero. Instead, it reaches a floor of about 7500 streams per CO or about 32% of the total number of unicast video streams.

4.3. Video Core Bandwidth Reduction

This section distills previous results further to determine overall impact of localized P2P in alleviating core and aggregation network bandwidth. Here, the number of video streams is converted into bandwidth by assuming HDTV at 19.3 Mbps per stream.

The figures here clearly show the impact of localized P2P on core and access bandwidth. Figures 7 and 8 show that nearly all unicast video (99.9%) can be delivered locally with P2P. Figure 9 shows that NG-PON can fully enable this localization, nearly as well as a theoretical system with infinite capacity (“Unlimited”). However, GPON has insufficient bandwidth to fully enable localized P2P video.

562832.fig.007
Figure 7: Localized versus non-localized unicast video bandwidth. P2P hairpins at the OLT chassis or OLT port. Customers on the same OLT can share local P2P video. NG-PON.
562832.fig.008
Figure 8: Localized versus non-localized unicast video bandwidth. P2P hairpins at the OLT chassis, or OLT port, or CO-router. Customers at the same CO can share local P2P video. NG-PON.
562832.fig.009
Figure 9: Comparing unicast video streaming core and aggregation bandwidth with GPON, NG-PON, and unlimited upstream capacity for P2P.

IP Video is either unicast (one-to-one) or multicast (one to-many). Multicast only needs a fixed amount of core and aggregation bandwidth, less than about 5 Gbps. The unicast video bandwidths here are on the order of hundreds of Gbps, and so unicast bandwidth can be considered to be about equal to the entire video bandwidth.

4.4. Overall Reduction in Internet Core Bandwidth

Section 2.1 briefly presented the Cisco study projecting consumer Internet growth [13], and showed that these projections are reasonably similar to trends other reported elsewhere [12, 14, 15]. Numbers from the Cisco study are used in this section, primarily because Cisco projected the growth of the individual components of Internet traffic: IP video, P2P, IPTV, and so forth.

This section assumes that a carrier aggressively pursues network-controlled, localized P2P. Cisco estimated a Compound Annual Growth Rate (CAGR) of over 100% for Internet Video to TV, and for IPTV. A 100% CAGR of localized P2P video is assumed here from 2009 to 2012, resulting in the following percentages of traffic that becomes localized P2P in 2012: Internet video 50%, P2P video 60%, and IPTV (VOD only) 75%. There is currently (in 2008) almost no P2P video traffic that is localized to within a single CO, and so the starting point is zero in 2008. Also, 65% of P2P traffic are assumed to be video here, which is mid-way between the current values of 60% to 70% reported by Cisco. These assumptions and the data in Figure 10 generate the projections of localized P2P video impact on total consumer Internet bandwidth shown in Figure 11.

562832.fig.0010
Figure 10: Aggregated Cisco projections for global consumer Internet growth: Video is projected to grow to be almost 75% of total consumer broadband traffic.
562832.fig.0011
Figure 11: Combined overall projected growth: localized P2P growth and Cisco forecast.

The data in Figure 11 shows that in 2012, localized P2P is projected to grow to 58.2% of IP video traffic, and 43.5% of all consumer Internet traffic.

5. Summary

This report showed that localized P2P video distribution can greatly decrease core and aggregation network bandwidth growth.

Cisco estimates that video will grow to nearly 75% of consumer Internet traffic in 2012. Over 99% of Internet video plus IPTV bandwidth in the core network are expected to be unicast, since multicast is very efficient in the core.

Assuming ubiquitous use of technologies for P2P localization, calculations here showed that as much as 99.9% of all unicast video traffic can be localized to a single OLT serving area, and even more can be localized to a single CO serving area. Changing content appetites and practical considerations could lower the share of localized P2P somewhat to perhaps 90%, 80%, of IP video, but this would still be a highly substantial portion.

Assuming aggressive but reasonable penetration rates, in 2012 localized P2P is projected to be able to grow to 58.2% of IP video traffic, and 43.5% of all consumer Internet traffic. This localized P2P video traffic could exceed 10 ExaBytes per month—twice the entire traffic volume of the current consumer Internet!

There is a clear need for sufficient information flows, rights management, and incentives between subscribers and network service providers to enable localized P2P. Further, this report showed that P2P video localization down to an individual access multiplexer such as an OLT, or down to individual COs, can be highly beneficial and should be supported. Mechanisms are necessary to assure video flows with acceptable

References

  1. K. Kerpez, “Bandwidth impacts of localizing peer-to-peer IP video traffic in access and aggregation networks,” EURASIP Journal on Wireless Communications and Networking, vol. 2008, Article ID 491860, 7 pages, 2008. View at Publisher · View at Google Scholar
  2. Distributed Computing Industry Association (DCIA), http://www.dcia.info/.
  3. Minnesota Internet Traffic Studies (MINTS) data, http://www.dtc.umn.edu/mints/home.php.
  4. K. Cho, K. Fukuda, H. Esaki, and A. Kato, “The impact and implications of the growth in residential user-to-user traffic,” in Proceedings of the ACM Conference on Applications, Technologies, Architectures, and Protocols for Computer Communications, pp. 207–218, Pisa, Italy, September, 2006.
  5. “Draft Report of the Study Group on a Framework for Competition Rules to Address the Transition to IP-Based Networks,” “New competition promotion program 2010”, July 2006.
  6. K. Cho, K. Fukuda, H. Esaki, and A. Kato, “Observing slow crustal movement in residential user traffic,” in Proceedings of the 4th International Conference on Emerging Networking EXperiments and Technologies (CoNEXT '08), Madrid, Spain, December 2008. View at Publisher · View at Google Scholar
  7. M. Burke, “Ellacoya networks releases data on broadband subscriber and application usage in Europe,” September, 2005, http://www.ellacoya.com/news/pdf/2005/EllacoyaEuropeData.pdf.
  8. A. Parker, “P2P in 2005,” http://www.cachelogic.com/home/pages/studies/2005_01.php.
  9. H. Schulze and K. Mochalski, “Internet Study 2007,” http://www.ipoque.com/news_&_events/internet_studies/internet_study_2007.
  10. F. Dickson, “P2P: content's “Bad Boy”; tomorrow's distribution channel,” http://multimediaintelligence.com/index.php?page=shop.product_details&flypage=flypage.tpl&product_id=21&option=com_virtuemart&Itemid=80.
  11. D. Burstein, “p2p down to 20–25% of traffic,” DSL Prime, May, 2008.
  12. Sandvine, “2008 analysis of traffic demographics in North-American broadband networks”.
  13. Cisco, “Cisco visual networking index—forecast and methodology. 2007–2012,” http://www.cisco.com/en/US/solutions/collateral/ns341/ns525/ns537/ns705/ns827/white_paper_c11-481360_ns827_Networking_Solutions_White_Paper.html.
  14. In-Stat, “User-generated video, a global stage for you,” 2008.
  15. The Bridge, “Telco video status report—November, 2008,” http://www.mediabiz.com/thebridge/index.cfm.
  16. L. Plissonneau, J.-L. Costeux, and P. Brown, “Detailed analysis of eDonkey transfers on ADSL,” 0-7803-9455-0/06, IEEE, 2006.
  17. I. Stoica, R. Morris, D. Karger, F. Kaashoek, and H. Balakrishnan, “Chord: a scalable peer-to-peer lookup service for internet applications,” in Proceedings of the ACM Conference on Applications, Technologies, Architectures, and Protocols for Computer Communication, pp. 149–160, San Diego, Calif, USA, August 2001.
  18. R. A. Ferreira, A. Grama, and S. Jagannathan, “An IP address based caching scheme for peer-to-peer networks,” in Proceedings of IEEE Global Telecommunications Conference (GLOBECOM '03), vol. 7, pp. 3845–3850, San Francisco, Calif, USA, December 2003.
  19. G. Pfeifer, C. Fetzer, and T. Hohnstein, “Exploiting host name locality for reduced stretch P2P routing,” in Proceedings of the 6th IEEE International Symposium on Network Computing and Applications (NCA '07), pp. 134–141, Cambridge, Mass, USA, July 2007. View at Publisher · View at Google Scholar
  20. M. Cha, P. Rodriguez, S. Moon, and J. Crowcroft, “On next generation telco-managed P2P TV architectures,” in Proceedings of the 7th International Workshop on Peer-to-Peer Systems (IPTPS '08), Tampa Bay, Fla, USA, February 2008.
  21. H. Xie, A. Krishnamurthy, A. Silberschatz, and Y. R. Yang, “P4P: explicit communications for cooperative control between P2P and network providers,” http://www.dcia.info/documents/P4P_Overview.pdf.
  22. http://www.openp4p.net/front/p4pwg.
  23. http://digital.venturebeat.com/2008/05/30/bittorrents-plan-ride-the-video-wave-to-the-bank/.
  24. http://www.rightsflow.com/.
  25. http://www.velocix.com/.
  26. B. Falchuk, D. Gorton, and D. Marples, “Enabling revenue-generating digital content distribution for telecom carriers,” in Proceedings of the 3rd IEEE Consumer Communications and Networking Conference (CCNC '06), vol. 2, pp. 1144–1148, Las Vegas, Nev, USA, January 2006. View at Publisher · View at Google Scholar
  27. A. Sentinelli, G. Marfia, M. Gerla, L. Kleinrock, and S. Tewari, “Will IPTV ride the peer-to-peer stream?” IEEE Communications Magazine, vol. 45, no. 6, pp. 86–92, 2007. View at Publisher · View at Google Scholar
  28. T. T. Do, K. A. Hua, and M. A. Tantaoui, “P2VoD: providing fault tolerant video-on-demand streaming in peer-to-peer environment,” in Proceedings of IEEE International Conference on Communications (ICC '04), vol. 3, pp. 1467–1472, Paris, France, June 2004.
  29. Y. Guo, K. Suh, J. Kurose, and D. Towsley, “P2Cast: peer-to-peer patching scheme for VoD service,” in Proceedings of the 12th International World Wide Web Conference (WWW '03), Budapest, Hungary, May 2003.
  30. C. Huang, J. Li, and K. Ross, “Peer-assisted VoD: making internet video distribution cheap,” in Proceedings of the 6th International Workshop on Peer-to-Peer Systems (IPTPS '07), Bellevue, Wash, USA, February 2007.
  31. B. Cheng, X. Liu, Z. Zhang, and H. Jin, “A measurement study of a peer-to-peer video-on-demand system,” in Proceedings of the 6th International Workshop on Peer-to-Peer Systems (IPTPS '07), Bellevue, Wash, USA, February 2007.
  32. J. Liu, S. G. Rao, B. Li, and H. Zhang, “Opportunities and challenges of peer-to-peer internet video broadcast,” Proceedings of the IEEE, vol. 96, no. 1, pp. 11–24, 2008. View at Publisher · View at Google Scholar
  33. V. Janardhan and H. Schulzrinne, “Peer assisted VoD for set-top box based IP network,” in Proceedings of the ACM Conference on Applications, Technologies, Architectures, and Protocols for Computer Communications, Kyoto, Japan, August 2007.
  34. X. Y. Yang, P. Hernandez, F. Cores, et al., “DVoDP/sup 2/P: distributed P2P assisted multicast VoD architecture,” in Proceedings of the 20th International Parallel and Distributed Processing Symposium (IPDPS '06), Rhodes Island, Greece, April 2006.
  35. F. Picconi and L. Massoulie, “Is there a future for mesh-based live video streaming?” in Proceedings of the 8th International Conference on Peer-to-Peer Computing (P2P '08), pp. 289–298, Aachen, Germany, September 2008. View at Publisher · View at Google Scholar
  36. N. Magharei, R. Rejaie, and Y. Guo, “Mesh or multiple-tree: a comparative study of live P2P streaming approaches,” in Proceedings of the 26th IEEE International Conference on Computer Communications (INFOCOM '07), pp. 1424–1432, Anchorage, Alaska, USA, May 2007. View at Publisher · View at Google Scholar
  37. S. Agarwal, J. P. Singh, A. Mavlankar, P. Baccichet, and B. Girod, “Performance and quality-of-service analysis of a live P2P video multicast session on the internet,” in Proceedings of the 16th International Workshop on Quality of Service (IWQoS '08), pp. 11–19, Enschede, The Netherlands, June 2008. View at Publisher · View at Google Scholar
  38. M. Piatek, C. Dixon, A. Krishnamurthy, and T. Anderson, “LiveSwarms: adapting BitTorrent for end host multicast,” Tech. Rep. UW-CSE- 06-11-01, University of Washington, Seattle, Wash, USA, 2006.
  39. http://www.allyourtv.com/, http://www.hollywoodreporter.com/.
  40. http://en-us.nielsen.com/.
  41. K. J. Kerpez, “Statistical variables for evaluating compatibility of remote deployments,” ATIS standards contribution T1E1.4/2001-132, May, 2001.
  42. M. Cha, H. Kwak, P. Rodriguez, Y.-Y. Ahn, and S. Moon, “I tube, you tube, everybody tubes: analyzing the world's largest user generated content video system,” in Proceedings of the 7th ACM SIGCOMM Internet Measurement Conference (IMC '07), pp. 1–14, San Diego, Calif, USA, October 2007. View at Publisher · View at Google Scholar