Internet est-il performant ? Les Cottrell – SLAC École SIG et nouvelles Technologies, République Démocratique du Congo, 12-17 Septembre, organisée par l’Université de Kinshasa Translated by Guillaume Cesieux, SLAC www.slac.stanford.edu/grp/scs/net/talk11/perform.ppt

Plan Caractéristiques d’Internet Comment est-il utilisé Utilisateurs, capacités, satellites, taille de paquets, protocoles, routage, flux Comment est-il utilisé Applications etc. Quelle est la performance mondiale d’Internet Vue par de nombreuses mesures et métriques Les besoins des applications Introduction to how the Internet is working today Later (in 3rd lecture) will get into details of how protocols work, i.e. formats, addressing, acknowledgements, timeouts. This lecture is more on top level observations on what is passing through the Internet, e.g. packet sizes, protocol usage, web page sizes, hops, ISP/AS, routes applications and flows performance dependencies for various applications, what are important metrics how well is the Internet performing for these metrics by regions etc. Will introduce terms and explain but only at an overview level.

Utilisateurs

Internet Usage growth ‘95-’10 1500 1000 500 Asie Europe Amer N. Amer L. Africa M. Orient AustrAsia 2,09 milliards mars 2011 Millions d’utilisateurs 0 200 400 600 Millions d’utilisateurs Amer. N. Australie Europe Amer. L M. Orient Asie Afrique Monde See http://www.webnewswire.com/node/464312 increasing number of Internet users every year indicates growing awareness among the people across the world. According to Internet World Stats June-09 statistics, World population reached 6.76 billions, increasing 1.4% YOY. At the same time, worldwide Internet usage penetration reached 24.7%, which means almost a quarter of the world population is accessing Internet. As the western world is saturating, the emerging world is fast catching up with the usage numbers. One Out of Four Persons Accessing Internet According to the Internet World Stats, worldwide Internet users reached 1.66 billions penetrating 24.7% with a growth rate of 2.8%. The number of online users at midyear 2008 was 1,463,632,361 with penetration rate of 21.9%. Now almost one out of every four persons is accessing the Internet. Earlier it was one out of five. North America Dominates the Penetration Rate North America is dominating with higher penetration rate of Internet users with 215 million Internet users with 73.9% penetration rate, United States and Canada being the major contributors. When it comes to total users, Asian region has the highest number of Internet users. More than 42% of world’s Internet users are coming from Asian countries with 18.5% penetration rate. On the other hand, Australia and Europe are holding second and third positions After North America with 60.1% and 50.1% respectively. Top 20 Countries Hold More Than 75% Top 20 countries with highest number of users are holding 76.8% of total usage. China is in top position with 298 million users and United States holding second position with 227 million users. However, when we compare penetration rates of these two countries, United States is way ahead of China. United States has more than 74%, where as China has only 22% penetration rate. Japan, India, Brazil, Germany, UK, France, Russia and South Korea are holding next positions respectively in total usage. Overall, there is a significant improvement in worldwide Internet users compared to last year. Penetration rates improved rapidly in the emerging countries, which propelled the major growth. Even though one out of four persons is accessing Internet, still there is huge scope for improvement across the globe. Pénétration % Année 0 20% 50% 80% 95 00 05 09

Exemple: La Chine La Chine n’était pas connectée à Internet jusqu’en mai 1994 La première connexion permanente était à IHEP/Pékin en utilisant une liaison satellite via SLAC www.computerworld.com.au/article/128099/china_celebrates_10_years_being_connected_internet

Utilisateurs d’internet en 2002 Où sont-ils ? Villes connectées à Internet Utilisateurs d’internet en 2002 2,8% de croissance/an ~¼ pop mondiale l’utilise Pays développés le sature Les pays en dev. rattrapent 73% pénétration au USA 43% utilisateurs en Asie

Bande passante

Qu’est ce qui est disponible? Capacité Source: Telegeography Capacity

Qui utilise encore le satellite? GEOS GEOS (Satellite à orbite géostationnaire) Bonne couverture mais cher en $/Mbps 50 fois plus cher qu’aux USA > 800% du salaire moyen mensuel c.f. 20% aux USA ET beaucoup de latence min RTT > 450ms, souvent bien plus à cause de la congestion Min RTT (ms) Terrestrial RTT = round trip time = Temps d’aller retour Facile à détecter –signature claire

Types de paquets et tailles

Taille de paquet Principalement 3 tailles: POURQUOI? Proche du minimum=telnet et ACKs, 1500 (max Ethernet payload, e.g. FTP, HTTP); ~ 560 octets pour les implémentations de TCP n’utilisant pas la MTU discovery Moy ~ 420 octets, médian ~ 80 octets Mesuré en Février 2000 au Ames Internet eXchange Paquets ~ 84 million de paquets < 0,05% fragmentés Probabilité cumulée % octets Taille de paquet (octets)

Usage des protocoles d’Internet Il y a 3 principaux protocoles utilisés sur Internet: UDP (sans connexion, datagrammes, best effort ) TCP (orienté connexion, délivrance ordonnée assurée) ICMP (Internet Control Message Protocol) Aujourd’hui TCP domine Flux/protocols à SLAC ICMP In TCP Flux/10min Out UDP Période Fev-Mai 2001

Routage

Sauts (Hops) Nombre de sauts entre 4 sites (Japon, S. Cal, N. Cal, E. Canada - 10-15 sauts en moyenne) Faible dépendance du RTT sur le nombre de sauts RTT 95% 50% 5% Sauts Nombre de sauts

Force de la connectivité Angle = longitude du siège de l’AS dans les entrées WHOIS Radius=1-log(outdegree(AS)+1)/(maxoutdegree + 1) Outdegree = number of next Hops As’ accepting traffic Bleu foncé & rouge = plus de connexions Tous les AS dans le top 15 sont aux USA – Excepté un au Canada Peu de liens entre ISPs en Europe et Asie ISP = internet service provider = fournisseur d’accès à internet

Une position moins centrale pour les USA www.nytimes.com/2008/08/30/business/30pipes.html Inventé aux USA Les 30 première années la majorité du trafic passait par les USA 70%=>20% en 10 ans Pas de contrôle central Patriot act => stockage des informations en dehors des USA Investissement fort de la Chine, l’Inde et le Japon Situation plus équitable Plus dur pour la CIA ! SAN FRANCISCO — The era of the American Internet is ending. Skip to next paragraph Enlarge This Image iStock Multimedia Graphic Rerouting the Web Tengku Bahar/Agence France-Presse — Getty Images “Suppose the Internet was entirely confined to the U.S., which it once was? That wasn’t helpful,” said Vint Cerf of Google. Invented by American computer scientists during the 1970s, the Internet has been embraced around the globe. During the network’s first three decades, most Internet traffic flowed through the United States. In many cases, data sent between two locations within a given country also passed through the United States. Engineers who help run the Internet said that it would have been impossible for the United States to maintain its hegemony over the long run because of the very nature of the Internet; it has no central point of control. And now, the balance of power is shifting. Data is increasingly flowing around the United States, which may have intelligence — and conceivably military — consequences. American intelligence officials have warned about this shift. “Because of the nature of global telecommunications, we are playing with a tremendous home-field advantage, and we need to exploit that edge,” Michael V. Hayden, the director of the Central Intelligence Agency, testified before the Senate Judiciary Committee in 2006. “We also need to protect that edge, and we need to protect those who provide it to us.” Indeed, Internet industry executives and government officials have acknowledged that Internet traffic passing through the switching equipment of companies based in the United States has proved a distinct advantage for American intelligence agencies. In December 2005, The New York Times reported that the National Security Agency had established a program with the cooperation of American telecommunications firms that included the interception of foreign Internet communications. Some Internet technologists and privacy advocates say those actions and other government policies may be hastening the shift in Canadian and European traffic away from the United States. “Since passage of the Patriot Act, many companies based outside of the United States have been reluctant to store client information in the U.S.,” said Marc Rotenberg, executive director of the Electronic Privacy Information Center in Washington. “There is an ongoing concern that U.S. intelligence agencies will gather this information without legal process. There is particular sensitivity about access to financial information as well as communications and Internet traffic that goes through U.S. switches.” But economics also plays a role. Almost all nations see data networks as essential to economic development. “It’s no different than any other infrastructure that a country needs,” said K C Claffy, a research scientist at the Cooperative Association for Internet Data Analysis in San Diego. “You wouldn’t want someone owning your roads either.” Indeed, more countries are becoming aware of how their dependence on other countries for their Internet traffic makes them vulnerable. Because of tariffs, pricing anomalies and even corporate cultures, Internet providers will often not exchange data with their local competitors. They prefer instead to send and receive traffic with larger international Internet service providers. This leads to odd routing arrangements, referred to as tromboning, in which traffic between two cites in one country will flow through other nations. In January, when a cable was cut in the Mediterranean, Egyptian Internet traffic was nearly paralyzed because it was not being shared by local I.S.P.’s but instead was routed through European operators. The issue was driven home this month when hackers attacked and immobilized several Georgian government Web sites during the country’s fighting with Russia. Most of Georgia’s access to the global network flowed through Russia and Turkey. A third route through an undersea cable linking Georgia to Bulgaria is scheduled for completion in September. Ms. Claffy said that the shift away from the United States was not limited to developing countries. The Japanese “are on a rampage to build out across India and China so they have alternative routes and so they don’t have to route through the U.S.” Andrew M. Odlyzko, a professor at the University of Minnesota who tracks the growth of the global Internet, added, “We discovered the Internet, but we couldn’t keep it a secret.” While the United States carried 70 percent of the world’s Internet traffic a decade ago, he estimates that portion has fallen to about 25 percent. Internet technologists say that the global data network that was once a competitive advantage for the United States is now increasingly outside the control of American companies. They decided not to invest in lower-cost optical fiber lines, which have rapidly become a commodity business. That lack of investment mirrors a pattern that has taken place elsewhere in the high-technology industry, from semiconductors to personal computers. The risk, Internet technologists say, is that upstarts like China and India are making larger investments in next-generation Internet technology that is likely to be crucial in determining the future of the network, with investment, innovation and profits going first to overseas companies. Skip to next paragraph Multimedia “Whether it’s a good or a bad thing depends on where you stand,” said Vint Cerf, a computer scientist who is Google’s Internet evangelist and who, with Robert Kahn, devised the original Internet routing protocols in the early 1970s. “Suppose the Internet was entirely confined to the U.S., which it once was? That wasn’t helpful.” International networks that carry data into and out of the United States are still being expanded at a sharp rate, but the Internet infrastructure in many other regions of the world is growing even more quickly. While there has been some concern over a looming Internet traffic jam because of the rise in Internet use worldwide, the congestion is generally not on the Internet’s main trunk lines, but on neighborhood switches, routers and the wires into a house. As Internet traffic moves offshore, it may complicate the task of American intelligence gathering agencies, but would not make Internet surveillance impossible. “We’re probably in one of those situations where things get a little bit harder,” said John Arquilla, a professor at the Naval Postgraduate School in Monterey, Calif., who said the United States had invested far too little in collecting intelligence via the Internet. “We’ve given terrorists a free ride in cyberspace,” he said. Others say the eclipse of the United States as the central point in cyberspace is one of many indicators that the world is becoming a more level playing field both economically and politically. “This is one of many dimensions on which we’ll have to adjust to a reduction in American ability to dictate terms of core interests of ours,” said Yochai Benkler, co-director of the Berkman Center for Internet and Society at Harvard. “We are, by comparison, militarily weaker, economically poorer and technologically less unique than we were then. We are still a very big player, but not in control.” China, for instance, surpassed the United States in the number of Internet users in June. Over all, Asia now has 578.5 million, or 39.5 percent, of the world’s Internet users, although only 15.3 percent of the Asian population is connected to the Internet, according to Internet World Stats, a market research organization. By contrast, there were about 237 million Internet users in North America and the growth has nearly peaked; penetration of the Internet in the region has reached about 71 percent. The increasing role of new competitors has shown up in data collected annually by Renesys, a firm in Manchester, N.H., that monitors the connections between Internet providers. The Renesys rankings of Internet connections, an indirect measure of growth, show that the big winners in the last three years have been the Italian Internet provider Tiscali, China Telecom and the Japanese telecommunications operator KDDI. Firms that have slipped in the rankings have all been American: Verizon, Savvis, AT&T, Qwest, Cogent and AboveNet. “The U.S. telecommunications firms haven’t invested,” said Earl Zmijewski, vice president and general manager for Internet data services at Renesys. “The rest of the world has caught up. I don’t see the AT&T’s and Sprints making the investments because they see Internet service as a commodity.”

Les routes sont asymétriques Advanced vers U. Chicago Min, 50% & 90% RTTs mesurés par une étude A révélé de grandes disparités entre RTTs Peut être dû à des chemins différents dans les 2 sens, ou à différentes charges sur les liens RTT ms U. Chicago vers Advanced RTT ms

Flux

Taille des flux SNMP Real A/V AFS Serveur de fichier Distribution lourde, in ~ out, flux UDP plus court que les TCP, paquets ~ octets 75% TCP-in < 5ko / 75% TCP-out < 1,5ko (<10 paquets) UDP 80% < 600 octets (75% < 3 paquets), ~10 * plus de TCP que d’UDP Top UDP = AFS (>55%), Real (~25%), SNMP(~1.4%) On peut grossièrement caractériser cela en une loi de puissance avec pente & interception

Durée des flux 60% des flux TCP durent moins d’une seconde Mesuré avec netflow Flux restreints à 30 minutes TCP outbound flows Temps d’activité en secondes 60% des flux TCP durent moins d’une seconde On s’attendait à ce qu’ils durent plus longtemps Mais 60% des flux UDP durent plus de 10 secondes – Peut être dû à l’usage intensif d’AFS au SLAC Une autre étude (CAIDA) reporte que les flux UDP durent moins longtemps que les flux TCP

Applications

Web Mi 2012 IE < 50% - Chrome dépassera Firefox

Usages Le P2P fait face à des poursuites légales (RIAA) Tendance à la vidéo et aux réseaux sociaux La vidéo à la demande double tout les 2 ans (2008-13) iPhones (désormais accessoirement un téléphone) Le trafic des mobiles double chaque année Yahoo See http://findarticles.com/p/articles/mi_m0DXK/is_18_21/ai_n6358702/ A decade after the World Wide Web opened up convenient Internet use to millions of Americans, 10 major trends have emerged to reveal how the Internet most affects its users, according to a comprehensive study of online technology impact by the University of Southern California's Annenberg School's Center for the Digital Future. Released as "The Digital Future Project, Year Four: Surveying the Digital Future" (formerly the UCLA Internet Report), the report provides a year-to-year exploration of the Internet's influence on Americans. The project has explored the behavior and views of a national sample of 2,000 Internet users and nonusers. "Of these findings, 10 broad trends have emerged that have particular relevance as we reach the 10-year anniversary of the opening of the Internet to the public: 10 years, 10 trends," says Dr. Jeffrey I. Cole, the director of the Center for the Digital Future. Related Results TRENDS IN SEARCHING FOR COMMERCE RELATED INFORMATION ON WEB SEARCH ENGINES Evolutions of All-in-One Devices and Internet-TV Functionality Will Fuel the... Nearly 1 Million Users Confirm Web is Becoming De Facto Business Desktop -... "2007 Digital Camera End-User Survey Analysis: United States" by... Cisco Visual Networking Index Projects Global IP Traffic to Reach Over Half a... The trends are as follows: 1. In America, the digital divide is closing, but is not yet closed as new divides emerge: The Digital Future Project found that about 75 percent of Americans can access the Internet from some location--home, work, school and libraries. The fastest-growing Internet user populations are groups traditionally considered the digital divide's forgotten people: Latinos, African Americans and older Americans. "Some lingering digital divide issues remain: There is still a divide based on Internet access at home. And, a new divide is coming that will bring with it a new set of ramifications: the divide between those who have broadband and those who use traditional telephone modern access," Cole says. 2. The media habits of the nation have changed, and continue to change: For the past 50 years, Americans' time at home has been dominated by television. Increasingly over the last 10 years, Internet users have "bought" their time to go online from the time they previously spent watching television. And, the more experience users have with the Internet, the less television they watch. 3. The credibility of the Internet is dropping: The credibility of information on the Internet rated high among users through the first three years of the Digital Future Project, and that credibility rates fairly high in year four. Nevertheless, the project found that the high credibility level for online information began to fall in the third year of the study, and fell even further in year four. 4. We have just begin to see the changes to come in buying online: Several developments have affected online purchasing in the past four years. Though concerns about credit card security while buying online remain extremely high, the intensity of that concern is declining. Three years ago, two-thirds of Internet users were very concerned or extremely concerned about privacy of information while buying online. In the current report, that number has declined to less than half (46 percent). And, while concerns remain high, Internet users are buying more often. 5. The "Geek-Nerd" perception of the Internet is dead: Since the beginning of the Digital Future Project, its studies found that going online has not lessened the social lives of users. The Internet has little or no impact on time spent with family or friends, or on sleeping, exercising or most other personal activities (other than watching television). 6. Privacy and security: concerns remain, but the high levels are changing: In each of the four studies by the Digital Future Project, Internet users and non-users alike have expressed very high levels of concern about privacy and security. They fear not only for their personal security, but are also concerned about companies or individuals tracking what they do online. 7. The Internet has become the No. 1 source for information for Internet users: The Internet has become the most important source of current information for users--the primary place they go for research, general information, hobbies, entertainment listings, travel, health and investments. 8. The benefits--and drawbacks--of the Internet for children are still coming into focus: "Perhaps the greatest conflicts about the Internet emerge in our exploration of how adults perceive the role of the Internet in their children's lives," Cole says. 9. E-mail: "E-Nuff" already? E-mail is still the single most important reason people go online. E-mail is a tremendous convenience, and for most users, it is a free service with enormous benefits. E-mail opens opportunities to communicate more often and with a much broader circle of people than we ever reach by telephone or by mail. 10. Broadband will change everything--again: Just as the arrival of the Internet created a flood of social change, the spread of broadband technology as a means of access to the Internet is causing a revolution of its own. Broadband is changing entirely its users' conduct with the Internet at home--how often Americans go online, how long they stay online, and what they do online. To download the full text of the Digital Future Project, visit <www.digitalcenter.org>. Also see: http://www.techcrunch.com/2008/04/28/morgan-stanleys-march-internet-trends-report-social/ And http://www.cisco.com/en/US/solutions/collateral/ns341/ns525/ns537/ns705/ns827/white_paper_c11-481360_ns827_Networking_Solutions_White_Paper.html YouTube Google Facebook

Croissance de la vidéo Le trafic P2P, encore majoritaire aujourd’hui, diminuera en tant que pourcentage global du trafic La vidéo sur internet, en direct et en téléchargement, occupe une part croissante de la bande passante et atteindra presque 60% du trafic internet en 2014 Definitions Web, email, and data: Includes web, email, instant messaging, and other data traffic (excluding file sharing) File sharing: Includes peer-to-peer traffic from all recognized P2P systems such as BitTorrent, eDonkey, etc., as well as traffic from web-based file sharing systems Gaming: Includes casual online gaming, networked console gaming, and multiplayer virtual world gaming Video communications: Includes PC-based video calling, webcam viewing, and web-based video monitoring VoIP: Includes traffic from retail VoIP services and PC-based VoIP, but excludes wholesale VoIP transport Internet video to PC: Free or pay TV or VoD viewed on a PC, excludes P2P video file downloads Internet video to TV: Free or pay TV or VoD delivered via Internet but viewed on a TV screen using an STB or media gateway Web, Email, and Data This is a general category that encompasses web browsing, email, instant messaging, data (which includes file transfer using HTTP, FTP, etc.) and other Internet applications. Note that "data" may include the download of video files that are not captured by the "Internet video to PC" forecast. It includes traffic generated by all individual Internet users. An Internet user is here defined as someone who accesses the Internet through a desktop or laptop at home, school, Internet café, or other location outside the context of a business.

Utilisation d’internet et causes asert.arbornetworks.com/2009/08/the-internet-after-dark Enterprise & tier 1 The Internet After Dark (Part 1) by Craig Labovitz After dark when the dinner dishes are put away and the kids are safely tucked into bed, the Internet subtly changes. Starting in the twilight of early evening, business traffic slows to a crawl, previously dormant applications flicker on home computer screens, and like clockwork, Internet activity begins its nightly climb towards a regular after hours bandwidth peak. But before we get too carried away with metaphor and innuendo, some background. In our last post blog post, we found (somewhat unexpectedly) that the pattern of North American daily Internet traffic differs from Europe and Asia. Unlike European Internet traffic which peaks around 7pm GMT and then quickly drops off until morning business hours, US Internet traffic reaches its peak at 11pm EDT and then stays relatively high until 3am in the morning (i.e. stays above 60% of peak or more). This uniquely American traffic pattern holds true across dozens of individual ISPs, tens of millions of subscribers, and petabytes of daily Internet traffic. The question is what are Americans doing at night? To begin answering this question, we first recap Internet Observatory data from our earlier post. The below graph shows the daily average traffic fluctuations of 40 North American consumer / regional providers (taking the average of 10 weekdays in July). To make the graph more readable, we show traffic as a percentage of peak traffic levels. All times are EDT. The way to interpret the graph above is that at 6am EDT North American traffic volumes are at 50% of their daily peaks. Traffic then climbs to a local maxima at 4pm and then a daily peak around 11pm EDT before again dropping during the early morning hours. To understand the two North American traffic peaks at 4pm and 11pm graphed above, it helps to look only at consumer Internet traffic (i.e. as opposed to enterprise and tier1 transit). Below we overlay in yellow the average daily traffic from only US and Canadian consumer providers (i.e. only showing cable / DSL and excluding tier2, research, content, tier1, etc.). From the graph its pretty clear consumer traffic plays a large role in the midnight North American traffic peak. We also see consumer traffic tends to climb later in the day (i.e. consumer traffic crosses 50% threshold after 9am as opposed to the broader Internet average of 6am) and consumer traffic trends towards filling the bulk of after-hours traffic. Perhaps most telling is the change in slope of graphed average consumer traffic around 6pm and then again around 8-9pm — all likely related to Americans turning to the Internet after dinner and during evening leisure hours. But though we now know it is consumers driving the late night Internet traffic peak, we still have not answered what they are doing. In response to our last post on the somewhat mysterious differences between American and European traffic patterns, readers offered a range of theories including: More so than Europe, American traffic grows with web surfing and at night Larges surges of P2P would explain North American traffic spikes Americans watch more video and related adult entertainment late at night In general, Europeans use the Internet less at night, have better social interactions, eat better food and generally live better lives We finish this blog post by exploring which of the above theories do not account for the large midnight spike in North American traffic. I have no way to evaluate the last bullet point around higher European quality of life (although I’m pretty sure my high school French teacher is still insisting this is true). Web During both the day and night the single largest Internet application is the web (52% of all Internet traffic on average). But while web surfing plays a large role in North American traffic trends, the graph below shows web does not provide the complete explanation behind the American bandwidth peaks. The “Daily Web Traffic” graph shows web as an average daily percentage of all North American Internet traffic. For purposes of this blog post, we define “web” as traffic on port 80, 8080 and 443. We note that web traffic includes both html page downloads as well as video and other applications running over HTTP. From a daily low of 42%, web traffic grows by 10% at night to account for 52% of all Internet traffic. So web accounts for slightly over half of the late night traffic, but what is consuming the other half of American traffic? P2P Given all the press and provider angst over P2P traffic, many commenters suggested (incorrectly) that P2P is the source of the post midnight bulge in American Internet traffic. As a category, P2P is the second largest source of American Internet traffic coming in at roughly 15-20% of all North American traffic. [Note: We'll devote an article on the evolution of P2P traffic in an upcoming post. And, of course, the upcoming "2009 Internet Observatory" report goes into far more detail on the statistics and methodology than our more casual blog postings.] Since most P2P does not use standard ports and/or includes encryption, we extrapolate the data below using a combination of Observatory port data with statistics from application payload characterization across several large US and Canadian cable operators. We again graph P2P as an average daily percentage of all North American Internet traffic. Unlike the web and almost all other applications, the daily average P2P cycle does not coincide with broader traffic trends. In fact, the P2P daily trend is pretty much completely inverted from daily traffic. In other words, P2P reaches it low at 4pm when web and overall Internet traffic approaches its peak. P2P traffic only bursts from a low of 8% to a high of 17% of Internet traffic after midnight and then drops off at 6am. As a side note, the cyclical inverted traffic pattern of P2P is interesting in its own accord. The inversion is highly suggestive of either persistent congestion or, more likely, evidence of widespread provider manipulation of P2P traffic rates. So P2P also does not explain the midnight spike of American Internet traffic. What does? Next week we’ll complete this blog post in Part II and explore the applications behind North American traffic after dark.   September 2, 2009 Update: Given some of the comments / questions around P2P as a percentage of Internet traffic versus a percentage of peak, a new graph below: Looking at P2P as a percentage of P2P peak traffic shows even more clearly the inverted pattern (i.e. since Internet traffic begins to climb at 6am, the earlier graph previously obscured the even more pronounced 6am peak in relative P2P traffic levels).

Caractéristiques des usages du Web La taille des objets varie de site à site, de serveur à serveur et en fonction de l’heure En 2000 des valeurs moyennes typiques sont entre 1 500 et 4 000 octets par objet Cela dépend aussi du type d’objet, des ordres de grandeurs sont ~100 ko a qq. Mo pour une petite vidéo, ~100 ko pour du postscript & audio, html applets & image ~ 1 000 Kb La taille moyenne des pages web a triplé en 5 ans (2003-2008) www.websiteoptimization.com/speed/tweak/average-web-page/ Bytes

Pourquoi une telle augmentation? Nouveau utilisateurs (accès facilité, plus de diffusion) Nouvelles applications: You-Tube, recherche climatique… Nouveaux outils de développement: Manuel  Génération automatique Web 2: Ajax, Javascript, CSS Haut débit  Contenus plus élaborés/attractifs possibles Application sur le web e.g. email, calendrier, albums photos, jeux...

Impact sur les réseaux - Historique du trafic sur ESnet Le trafic sur ESnet est en moyenne multiplié par 10 tous les 47 mois Apr 2006 1 PBy/mo. Nov 2001 100 TBy/mo. July 2010 10 PBy/mo. Jul 1998 10 TBy/mo. 53 months Oct 1993 1 TBy/mo. Téraoctets / mois Aug 1990 100 MBy/mo. 40 months 57 months 38 months Tracé du trafic mensuel d’ESnet, Janvier 1990 – Décembre 2008

Performance par les métriques

De quoi depend la performance? La performance d’internet de bout en bout vue par les applications dépend de: RTT (Round Trip Time - Temps d’aller retour) Perte de paquets (losses) Gigue (Jitter) L’accessibilité Les goulot d’étranglement Implémentation/configurations Du besoin des applications Des données présentes dans les paquets

RTT de SLAC vers le monde RTT ~ distance/(0.6*c) + hops * router delay Router delay = queuing + clocking in & out + processing msec. From the 300 msec have to be subtracted the processing tmes etc. for dejitter, endcoding, compression etc.The problem with > 300msec is “talker overlap”. "When I work with game companies, I like to design for a 200 to 250 msec round-trip latency and tolerate gracefully the rare instances when latencies might hit 1 second." http://www.gamasutra.com/features/19970905/ng_01.htm ITU G.114 300 ms RTT seuil pour la voix 2/3 des pays sont Ok pour la voix, le reste est principalement en Afrique Quel est le probleme avec > 300 ms?

RTT de Californie vers le monde Europe Cote E. Bresil Cote O. US Cote E. US 300ms RTT (ms) Europe & Amérique du Sud 0.3*0.6c Longitude (degrés) Skitter host at F name server in Palo Alto California to ~36,000 customers Note the diagonal minimum envelope, the outliers at 80W are due to incorrect coordinates in database Observe that the minima indicate that can get to most of the globe and back within the ITU RTT constraint. 300 ms Fréquence Source = Palo Alto CA, Cote O. Pourquoi ces distributions? RTT (ms.) Données fournies par le CAIDA Skitter project

RTT du Japon vers le monde RTT (ms) Longitude Mesuré depuis le Japon

Gigue (Jitter) “Jitter” = IQR(ipdv); ipdv(i) =RTT(i) – RTT(i-1) Variation du RTT, différentes manières de le mesurer “Jitter” = IQR(ipdv); ipdv(i) =RTT(i) – RTT(i-1) Principalement aux extrémités, donc souvent indépendant des distances Fort impact les petits flux (VoIP, vidéo, temps réel) Haptics (Chirurgie à distance) < 1 ms; H.323 < 40 ms avec buffer On peut améliorer la voix avec des buffers anti-gigue, e.g 70 ms de cache pour faciliter le transfert Mais…. Gigue internet depuis SLAC vers le monde Sept.08

Pertes de paquets Indépendant de la distance Fort impact Sur les bons liens c’est souvent dû à la congestion Atténuation du signal dB, équipements sans fil Souvent au « last mile » Indépendant de la distance Fort impact Temps réel, jeux, voix, frappe dupliquée 1% de perte suffisent à perturber la VoIP

Débit dérivé (Derived Throughput) Derived throughput ~ 8 * 1460 /(RTT * sqrt(loss)) Mathis et. al Retard par rapport a l'Europe: 5 ans: Russie, Amérique Latine, Moyen Orient 6 ans: Asie du SE 9 ans: Asie du Sud 12 ans: Asie centrale 16 ans: Afrique 1993 Central Asia, and Africa are in Danger of Falling Even Farther behind In 10 years at the current rate Africa will be 1000 times worse than Europe 36

Où est le meilleur débit?

Voix sur IP Affectée par: Pertes, RTT, gigue Qualité mesurée par le score MOS (Mean Opinion Score) On peut calculer le MOS à partir du RTT, des pertes et de la gigue Score MOS: 1=Mauvais; 2=Pauvre; 3=Acceptable; 4=Bon; 5=excellent La VoIP nécessite raisonnablement un MOS entre 3,5 et 4,2 La Russie et l’Amérique latine ont fait de gros progrès en 2002 en passant du satellite au terrestre. USA, Europe, Asie de l’E, Russie et Moyen Orient ont tous un MOS > 3.5 Bien: Asie du S.E, Moyen: Asie du Sud - l’Asie Centrale progresse L’Afrique commence à émerger

Besoins des applications Basé sur ITU Y1541 & Stanford (Haptics) Le seuil de perte de 0,001 pour la VoIP était auparavant 0,25 mais c’était en assumant des pertes aléatoires En pratiques les pertes sont souvent regroupées Perte de paquet dans les mémoires des routeurs Perte de synchronisation dans les circuits, reconvergence du spanning tree, changement de routes Application Temps réel VoIP WAN connectivity Web free services Flux video Haptics (Chirurgie a distance) 1 way delay 150ms 1000ms undefined 400ms 160ms ‘jitter” 50ms 17ms 1ms Loss 10-3 10-5 0.1

Et ensuite? Terminaux mobiles 40G (transatlantique, US), cœur de réseaux à 100Gb Bande passante dédiée à la demande (couches 1 & 2) Réservation d’un chemin à une certaine bande passante pour un certain temps Utilisation de la QoS HEP, radio astronomie, recherche climatique IPv6 Cypress, CA, October 9, 2009 –  National LambdaRail (NLR), the coast-to-coast, high-performance network owned by the U.S. research and education community, announces its updated Costs for Services, effective immediately. For the first time, NLR is enabling individual institutions and research project groups to join NLR directly, with new membership categories and a corresponding, affordable cost structure of only $70,000 and $25,000 per year respectively.  This includes unlimited, shared use of non-dedicated circuits on NLR FrameNet (Layer 2).  As a result, educators and researchers now have more choices and more affordable options for taking advantage of NLR’s cutting-edge production and experimental networks.  If they do not already belong to a regional optical network member of NLR, they may now work directly with NLR, also leveraging NLR’s individualized research project support services. NLR is also announcing today the availability of native 40-Gigabit Ethernet (GE) circuits across NLR WaveNet, NLR’s national, Layer 1 optical infrastructure.  NLR is also announcing that the NLR backbone can support 100-GE alien waves today for users providing the equipment for the wave.  Cost for 100-GE support on the NLR backbone is now $18,002 per route segment for Class A Sustaining Members, Individual Institution Participants or Individual Research Project Groups.  For Class B or Class C Members the cost is $24,315 per route segment.  100-GE circuits will be fully available on the NLR platform by January 2011. In addition, NLR is making highest-resolution video and audio technologies affordable to all categories of NLR members in order to further enable quality collaboration between researchers around the country and across the world.  Class A Sustaining and Individual Institution Participant Members will enjoy as part of their membership access to NLR’s TelePresence Exchange services based on Cisco TelePresence at no additional cost.  For Class B and C Members the cost is a nominal $25,000 per year.  NLR’s TelePresence Exchange, the only telepresence exchange of any research and education network, can connect up to 12 different physical locations in a single session.  For complete, updated  NLR Membership Benefits and Cost Guidelines, please see: http://www.nlr.net/docs/NLRMembership_Costs_for_Services.pdf # # # About National LambdaRail (NLR) Owned by the U.S. research and education community and dedicated to serving the needs of researchers and research groups, NLR is the innovation platform for a wide range of academic disciplines and public-private partnerships.  NLR’s coast-to-coast, high-performance network infrastructure offers unrestricted usage and bandwidth, a choice of cutting-edge network services and applications, and customized service for individual researchers and projects.  For more information, please visit www.nlr.net. Media Contact Kristina Scott, NLR, 650.678.9034, kscott@nlr.net Internet2 Launches New ION ServiceNew Production-Class Dynamic Circuit Network Service To Support Leading Edge Research by the Internet2 Community ANN ARBOR, Mich - October 9, 2009 – Internet2 this week announced a new production network service that allows researchers the ability to easily create dedicated, customized circuits on demand. Internet2 ION allows users to provision bandwidth when they need it to meet their applications’ specific time and capacity requirements. Using a simple web interface, dedicated circuits can be created instantly or reserved in advance and can be provisioned on the Internet2 Network across partner networks like the Department of Energy’s ESnet to provide guaranteed network resources for the most demanding of scientific experiments. “Researchers in a wide range of disciplines including high energy physics, radio astronomy and climate research are eager to have predictable capabilities and capacities for the intense requirements of their science. Internet2 ION provides researchers that require reliable dedicated network resources to support their work and to ensure new scientific discovery is only limited by imagination and not by network resources,” said Rob Vietzke, Internet2 executive director of network services. “We look forward to working with our regional network partners to roll out this service and to make it useful to our community across the country.” Internet2 ION is built on technology developed in collaboration with key partners including GÉANT and ESnet. The service leverages OSCARS technology, developed in collaboration with ESnet in part through funding by the Department of Energy’s Office of Science. The service also leverages DRAGON technology, developed in part through National Science Foundation funding by MAX, USC ISI East, and George Mason University. By leveraging these technologies, which are designed using widely deployed protocols, Internet2 ION is able to dynamically create circuits across the Internet2 Network and into partner regional, national and international networks, connecting researchers with their colleagues worldwide. For many distributed users and applications of today, this is an absolute requirement. “The Internet2 ION service opens up a whole new range of options for High-Energy Physicists using the network. Using ION initiated connections will allow us to optimize our transport of high-priority data without having to worry about potentially disruptive protocols interfering with other users,” said Shawn McKee, research scientist at the University of Michigan. “In addition, having the ability to create circuits via ION will provide us with the possibility of prioritizing and protecting the most critical data flows within our globally distributed infrastructure. As such, ION should help us expedite the anticipated scientific discoveries as part of the Large Hadron Collider research.” To create a circuit, the user signs into a simple, secure, web interface and follows a wizard-like, six-step process, setting start and end points, start and end times, bandwidth (available in 50Mbps increments up to 10Gbps) and other parameters. Reasonable default settings and helpful widgets speed the process, and context-sensitive help appears on screen. Users can also view sortable lists of current and past circuit reservations, create favorites, cancel reservations or modify start and end times, even jump-start a new circuit reservation by cloning an existing one. Monitoring screens and maps are also provided, allowing users to review their circuit status and usage statistics—the up/down state of optical and Ethernet links, utilization, dropped packets, discarded packets and errored packets for ingress and egress points. The research prototype version of the service was first known as DCN (Dynamic Circuit Network), but that term now refers to the underlying technology that enables services like Internet2 ION. The service's release signals the launch of a leading-edge production service, which represents the Internet2 community’s gateway to realizing the vision of dynamic circuit networking. “ION brings to fruition the vision of dynamic circuit networking that has been developed through a community-based network research initiative over the past five years,” said George Loftus, president of OSHEAN and chair of the Internet2 Architecture and Operations Advisory Council. “We are confident that Internet2 members who need to deliver the most cutting edge services to researchers will find the ION service invaluable.” Both the Internet2 AOAC, which advises on operations and architecture of the Internet2 Network, and the Research Advisory Council (RAC), which advises on network-focused research, have provided guidance on the development of the service and have endorsed its capabilities for the Internet2 community.

Questions & liens www.internetworld.stats.com www-iepm.slac.stanford.edu/pinger www.slac.stanford.edu/comp/net/wan-mon/tutorial.html www.slac.stanford.edu/xorg/icfa/icfa-net-paper-jan09/report-jan09.doc http://www.cablemap.info/

Comparaison avec les indices de développement Abv. Name Organization Countries Date of Data GDP Gross Domestic Product per capita CIA  229 2001-2006 HDI Human Development Index UNDP 175 2004 DAI Digital Access Index ITU 180 1995-2003 NRI Network Readiness Index World Economic Forum 120 2007 TAI Technology Achievement Index 72 1995-2000 DOI Digital Index 2004-2005 OI Index 139 1996-2003 CPI Corruption Perception Index Transparency Organization Choose most: up-to-date, countries, important factors HDI & DOI