Internet traffic

From Wikipedia, the free encyclopedia
Jump to: navigation, search

Internet traffic is the flow of data across the Internet.

Because of the distributed nature of the Internet, there is no single point of measurement for total Internet traffic. Internet traffic data from public peering points can give an indication of Internet volume and growth, but these figures exclude traffic that remains within a single service provider's network as well as traffic that crosses private peering points.

Traffic sources[edit]

File sharing constitutes a large fraction of Internet traffic.[1] The prevalent technology for file sharing is the BitTorrent protocol, which is a peer-to-peer (P2P) system mediated through indexing sites that provide resource directories. The traffic patterns of P2P systems are often described as problematic and causing congestion.[2] According to a Sandvine Research in 2013, Bit Torrent’s share of Internet traffic decreased by 20% to 7.4% overall, reduced from 31% in 2008.[3]

Streaming media provides users with video and audio resources, such as YouTube and Spotify.

Traffic management[edit]

Internet Connectivity Distribution & Core.svg

The Internet does not employ any formally centralized facilities for traffic management. Its progenitor networks, especially the ARPANET established early backbone infrastructure which carried traffic between major interchange centers for traffic, resulting in a tiered, hierarchical system of internet service providers (ISPs) within which the tier 1 networks provided traffic exchange through settlement-free peering and routing of traffic to lower-level tiers of ISPs. The dynamic growth of the world wide network resulted in ever-increasing interconnections at all peering levels of the Internet, so that a robust system developed that could mediate link failures, bottlenecks, and other congestion at many levels.[citation needed]

Tax on Internet use[edit]

A planned tax on Internet use in Hungary introduced a 150 forint (US$0.62, €0.47) tax per gigabyte of data traffic, in a move intended to reduce Internet traffic and also assist companies to offset corporate income tax against the new levy.[4] Hungary achieved 1.15 billion gigabytes last year and another 18 million gigabytes accumulated by mobile devices. This would have resulted in extra revenue of 175 billion forints under the new tax based on the consultancy firm eNet.[4]

According to Yahoo News, economy minister Mihály Varga defended the move saying "the tax was fair as it reflected a shift by consumers to the Internet away from phone lines" and that "150 forints on each transferred gigabyte of data – was needed to plug holes in the 2015 budget of one of the EU’s most indebted nations".[5]

Some people argue that the new plan on Internet tax would prove disadvantageous to the country’s economic development, limit access to information and hinder the freedom of expression.[6] Approximately 36,000 people have signed up to take part in an event on Facebook to be held outside the Economy Ministry to protest against the possible tax.[5]

Traffic classification[edit]

Traffic classification describes the methods of classifying traffic by observing features passively in the traffic, and in line to particular classification goals. There might be some that only have a vulgar classification goal. For example, whether it is bulk transfer, peer to peer file sharing or transaction-orientated. Some others will set a finer-gained classification goal, for instance the exact number of application represented by the traffic. Traffic features included port number, application payload, temporal, packet size and the characteristic of the traffic. There are a vast range of methods to allocate Internet traffic including exact traffic, for instance port (computer networking) number, payload, heuristic or statistical machine learning.[7] are the main methods.

Accurate network traffic classification is elementary to quite a few Internet activities, from security monitoring to accounting and from quality of service to providing operators with useful forecasts for long-term provisioning. Yet, classification schemes are extremely complex to operate accurately due to the shortage of available knowledge to the network. For example, the packet header related information is always insufficient to allow for an precise methodology. Consequently, the accuracy of any traditional method are between 50%-70%.

Bayesian analysis techniques[edit]

Work[8] involving supervised machine learning to classify network traffic. Data are hand-classified (based upon flow content) to one of a number of categories. A combination of data set (hand-assigned) category and descripttions of the classified flows (such as flow length, port numbers, time between consecutive flows) are used to train the classifier. To give a better insight of the technique itself, initial assumptions are made as well as applying two other techniques in reality. One is to improve the quality and separation of the input of information leading to an increase in accuracy of the Naive Bayes classifier technique.

The basis of categorizing work is to classify the type of Internet traffic; this is done by putting common groups of applications into different categories, e.g., "normal" versus "malicious", or more complex definitions, e.g., the identification of specific applications or specific Transmission Control Protocol (TCP) implementations.[9] Adapted from Logg et al.[10]

Survey[edit]

Traffic classification is a major component of automated intrusion detection systems.[11][12][13] They are used to identify patterns as well as indication of network resources for priority customers, or identify customer use of network resourrs that in some way contravenes the operator’s terms of service. Generally deployed Internet Protocol (IP) traffic classification techniques are based approximately on direct inspection of each packet’s contents at some point on the network. Source address: port and destination address are included in successive IP packet's with similar if not the same 5-tuple of protocol type. ort are considered to belong to a flow whose controlling application we wish to determine. Simple classficaiton infers the controlling application’s identity by assuming that most applications consistently use’ well known’ TCP or UDP port numbers. Even though, many candidates are increasingly using unpredictable port numbers. As a result, more sophisticated classification techniques infer application type by looking for application-specific data within the TCP or User Datagram Protocol (UDP) payloads.[14]

Global Internet traffic[edit]

Aggregating from multiple sources and applying usage and bitrate assumptions, Cisco Systems, a major network systems company, has published the following historical Internet Protocol (IP) and Internet traffic figures:[15]

Global Internet traffic by year
 
Year
IP Traffic
(PB/month)
Fixed Internet traffic
(PB/month)
Mobile Internet traffic
(PB/month)
1990 0.001 0.001 n/a
1991 0.002 0.002 n/a
1992 0.005 0.004 n/a
1993 0.01   0.01   n/a
1994 0.02   0.02   n/a
1995 0.18   0.17   n/a
1996 1.9     1.8     n/a
1997 5.4     5.0     n/a
1998 12       11       n/a
1999 28       26       n/a
2000 84       75       n/a
2001 197       175       n/a
2002 405       356       n/a
2003 784       681       n/a
2004 1,477       1,267       n/a
2005 2,426       2,055       0.9   
2006 3,992       3,339       4      
2007 6,430       5,219       15      
2008 9,927       7,639       38      
2009 14,414       10,676       92      
2010 20,197       14,929       256      
2011 27,483       20,634       597      
2012 - 31,338       885      

"Fixed Internet traffic" refers perhaps to traffic from residential and commercial subscribers to ISPs, cable companies, and other service providers.
"Mobile Internet traffic" refers perhaps to backhaul traffic from cellphone towers and providers.
The overall "Internet traffic" figures, which can be 30% higher than the sum of the other two, perhaps factors in traffic in the core of the national backbone, whereas the other figures seem to be derived principally from the network periphery.

Internet backbone traffic in the United States[edit]

The following data for the Internet backbone in the US comes from the Minnesota Internet Traffic Studies (MINTS):[16]

US Internet backbone traffic by year
Year Data (TB/month)
1990 1
1991 2
1992 4
1993 8
1994 16
1995 n/a
1996 1,500
1997 2,500–4,000
1998 5,000–8,000
1999 10,000–16,000
2000 20,000–35,000
2001 40,000–70,000
2002 80,000–140,000
2003 n/a
2004 n/a
2005 n/a
2006 450,000–800,000
2007 750,000–1,250,000
2008 1,200,000–1,800,000
2009 1,900,000–2,400,000
2010 2,600,000–3,100,000
2011 3,400,000–4,100,000

The Cisco data can be seven times higher than the Minnesota Internet Traffic Studies (MINTS) data not only because the Cisco figures are estimates for the global—not just the domestic US—Internet, but also because Cisco counts "general IP traffic (thus including closed networks that are not truly part of the Internet, but use IP, the Internet Protocol, such as the IPTV services of various telecom firms)".[17] The MINTS estimate of US national backbone traffic for 2004, which may be interpolated as 200 petabytes/month, is a plausible tree-fold multiple of the traffic of the US's largest backbone carrier, Level(3) Inc., which claims an average traffic level of 60 petabytes/month.[18]

See also[edit]

References[edit]

  1. ^ "Data volume of global file sharing traffic from 2013 until 2018". Statista. 2014. Retrieved 18 October 2014. 
  2. ^ Milton Kazmeyer. "What are the causes of Internet traffic?". Demand Media. Retrieved 18 October 2014. 
  3. ^ Paul Resenikoff (12 November 2013). "File-Sharing Now Accounts for Less Than 10% of US Internet Traffic…". Retrieved 18 October 2014. 
  4. ^ a b Marton Dunai (2014). "http://uk.reuters.com/article/2014/10/22/uk-hungary-internet-tax-idUKKCN0IB0RI20141022". 
  5. ^ a b "Anger mounts in Hungary over internet tax". Yahoo News. 25 October 2014. Retrieved 18 October 2014. 
  6. ^ Margit Feher (2014). "Public outrage mounts against hunger's plan to tax internet use". Retrieved 18 October 2014. 
  7. ^ "Internet traffic classification". National Science Foundation. 2013. Retrieved 18 October 2014. 
  8. ^ Denis Zuev (2013). "Internet traffic classification using bayesian analysis technique" (PDF). Retrieved 18 October 2014. 
  9. ^ J.Padhye; S.Floyd (June 2001). "Identifying the TCP Behavior of Web Servers". In Proceedings of SIGCOMM 2011, San Diego, CA. 
  10. ^ C.Logg; L.Cottrell (2003). http://www.slac.stanford.edu/comp/net/slac-netflow/html/SLAC-netflow.html. Retrieved 21 October 2014.  Missing or empty |title= (help)
  11. ^ of August 14, 2007 (accessed 21 October)
  12. ^ Bro intrusion detection system – Bro overview, http://bro-ids.org, as of August 14, 2007.
  13. ^ V. Paxson, ‘Bro: A system for detecting network intruders in real-time,’ Computer Networks, no.31 (23-24), pp. 2435-2463, 1999
  14. ^ S. Sen., O. Spats check, and D. Wang, ‘Accurate, scalable in network identification of P2P traffic using application signatures,’ in WWW2004, New York, NY, USA, May 2004.
  15. ^ "Visual Networking Index", Cisco Systems
  16. ^ Minnesota Internet Traffic Studies (MINTS), University of Minnesota
  17. ^ http://www.dtc.umn.edu/mints/
  18. ^ 2004 Annual Report, Level(3), April 2005, p.1

Further reading[edit]

  • Williamson, Carey (2001). "Internet Traffic Measurement". IEEE Internet Computing 5 (6): 70–74. doi:10.1109/4236.968834. 

External links[edit]