The WEB history :

The past 5-10 years have witnessed not only an explosion of activity, but the creation of entirelynew sectors within the optical industry. As the concept of WDM began to emerge, many new companiesdeveloping WDM transport equipment came into existence. The newer enterprises pushed theolder established equipment vendors to more aggressive deployment schedules and a constant downwardtrend for the corresponding prices of WDM transport equipment followed. In what appeared tobe an almost insatiable demand for more bandwidth, a situation arose that allowed the creation of thenew companies and the accompanying innovation. Not only did new equipment vendors emerge, butalso new national scale carriers were created. This trend is continuing as the concept of optical layering/networking is gaining acceptance and new optical equipment companies are being formed on aregular basis. They deal not only with “traditional” WDM transport equipment, but also with terrestrialultra long haul systems, regional and metro optimized systems, and various incarnations of opticalcross connects.There were hundreds of developments and contributions enabling this burst of activity. Many ofthe technical innovations are described in this book and its predecessors. However, perhaps the greatestsingle factor that fueled this phenomena was the belief and perception that traffic and hence neededcapacity were growing at explosive rates. This is a remarkable fact, especially when one recalls thataround 1990, both the traditional carriers and most of their equipment vendors still expected the trafficdemands to not vary much from the voice demand growths (which historically was around 10% peryear). In fact both carriers and equipment vendors were arguing that WDM would not be needed andthat going to individual channel rates of at most 10 Gb/s would be adequate. Also, around 1995,the conventional wisdom was that 8 channel WDM systems would suffice well into the foreseeablefuture. Now it almost appears as if the pendulum has swung the other way. Is too much capacity beingdeployed and are many of the reported traffic growth rates correct, and if so will they continue?As we explained in the previous section, the early skepticism about the need for high capacityoptical transport was rooted in the reality of the telecommunications networks. Up until 1990, theywere dominated by voice, which was growing slowly. Then, by the mid-1990s, they became to bedominated (in terms of capacity) by private lines, which were growing three or four times as fast. Andthen, in the late 1990s, they came to be dominated by the Internet, which was growing faster still.Before we go through the analyses for the traffic growth on the Internet we must first at least definethe Internet and describe the history and structure of it. This is paramount in helping put much of later12described growth analyses into perspective.When one now speaks of the Internet, it is usually described as an evolution from ARPANET toNSFNET, and finally to the commercial Internet that now exists. Arguably, the phenomenal growthof the Internet started in 1986 (more than 17 years after its “birth”) with NSFNet. However, the pathwas very complicated and full of many twists and turns in its roughly 40 year history [Cerf, Hobbes,Leiner].From the very early research in packet switching, academia, industry, and the US government havebeen intertwined as partners. Ironically, the beginnings of the Internet can trace itself back to the ColdWar and specifically to the launch of Sputnik in 1957. The US government formed the Advanced ResearchProject Agency (ARPA - the name was later changed to DARPA, Defense Advanced ResearchProject Agency, and later back to ARPA) the year after the launch with the stated goal of establishing aUS lead in technology and science (with emphasis on applications for the military). As ARPA was establishingitself, there were several pivotal works [Klein1, Baran] in the early 1960s on packet switchingand computer communications. These works and the efforts they spawned laid many of the foundationsthat enabled the deployment of distributed packet networks. J.C.R. Licklider (of MIT) [LickC]wrote a series of papers in 1962 in which he “envisioned a globally interconnected array of computerswhich would enable ‘everything’ to easily access data and programs from any of the sites”. Genericallyspeaking, this idea is not much different from what today’s Internet has become. Of importance is thefact the Licklider was the first head of the computer research program at DARPA (beginning in 1962),and in this role he was instrumental in pushing his concept of networks. Kleinrock published both thefirst paper on packet switching and the first book on the subject. In addition, Kleinrock convinced severalkey players of the theoretical feasibility of using packets instead of circuits from communications.One such person was Larry Roberts, one of the initial architects for the ARPANET. In the 1965-66time frame ARPA sponsored studies on “cooperative network of [users] sharing computers”[Leiner],and the first ARPANET plans were begun, with the first design papers on ARPANET being publishedin 1967. Concurrently the National Physical Laboratory (NPL) in England deployed an experimentalnetwork called the NPL Network making use of packet switching. It utilized 768 kb/s lines.A year before the Moon landing, in 1968, the first ARPANET requests for proposals were sentout, and the first ARPANET contracts were awarded. Two of the earliest contracts went to UCLA todevelop the Network Measurement Center, and to Bolt, Beranek and Newman (BBN) for the PacketSwitch contract (to construct the Interface Message Processors or IMPs - effectively the routers).Kleinrock headed the Network Measurement Center at UCLA and it was selected as the first node13on the ARPANET. The first IMP was installed at UCLA and the first host computer was connectedin September of 1969. The second node was at Stanford Research Institution (SRI). Two other nodeswere added at UCSB and in Utah, so that by the second half of 1969, just months past the first moonlanding, the initial four node ARPANET became functional. This was truly the initial ARPANET, andthus a case can be made that this was when the Internet was born. The first message carried over thenetwork went from Kleinrock’s lab to SRI. Supposedly the first packet sent over ARPANET was sentby Charley Kline and as he was trying to log in the system crashed as the letter “G” of “LOGIN” wasentered.One of the next major innovations for the fledgling Internet (i.e., ARPANET) was the introductionof the first host-to-host protocol called Network Control Protocal or NCP, which was first used inARPANET in 1970. By 1972 all of the ARPANET sites had finished implementing NCP. Hence theusers of ARPANETcould finally begin to focus on the development of applications - another paramountdriver for the phenomenal growth and sustained growth of the internet. It was also in 1970 that the firstcross-country link was established for ARPANET by AT&T between UCLA and BBN (at the blindingrate of 56 kb/s). By 1971, the ARPANET had grown to 15 nodes and had 23 hosts. However, perhapsthe most influential work that year was the creation of an email program that could send messagesacross a distributed network. (Email was not among the original design criteria for the ARPANET, andits success caught the creators of this network by surprise.) Ray Tomlinson of BBN developed this,and his original program was based on 2 previous ones [Hobbes]. Tomlinson modified his program forARPANET in 1972, and at that point its popularity quickly soared. In fact it was at this time that thesymbol “@” was chosen. Arguably Internet email as we know it today can trace its origins directly tothis work. Internet email was clearly one of key drivers for the popularity (and hence the phenomenaltraffic growth demands) of the Internet and was the first “killer app” for the Net. It was every bit ascritical to the Internet’s “success” as the spreadsheet applications were to the popularization of the PC.Internet email provided a new model of how people could communication with each other and alter thevery nature of collaborations.Although there was already considerable work being done on packet networks outside the US, thefirst international connections to the ARPANET (to England via Norway) took place in 1973. To putthe time frame in perspective this was the same year that Robert Metcalfe did his PhD which describedhis idea for Ethernet. Also during this year the number of ARPANET “users” was estimated to be 2000and that 75% of all the ARPANET traffic (in terms of bytes) was email. One needs to note that in only1-2 years from its introduction onto the Internet email became the predominant type traffic. The same14behavior took place several years later for html (i.e., Web traffic), and to a somewhat lesser degree, thiswas seen for Napster-like traffic within many networks a few years later.Several other key developments began to take place in the mid 1970s. The initial design specificationfor TCP published by Vint Cerf and Bob Kahn in 1974 [CerfK]. The NCP protocol which was beingutilized at the time, tended to act like a device driver, whereas the future TCP (later TCP/IP) would bemuch more like a communications protocol. As is discussed later, the evolution from ARPANET’s NCPprotocol to TCP (which in 1978 was split into TCP and IP) was critical in allowing the future growthand scalability of today’s Internet. DARPA had three contracts to implement TCP/IP (at the time stillcalled TCP), at Stanford (led by Cerf), BBN (led by Ray Tomlinson) and UCLA (led by Kirsten). Stanfordproduced the detailed specification and within a year there were 3 independent implementationsof TCP that could interoperate.It is noted that the basic reasons that led to the separation of TCP (which guaranteed reliable delivery)from IP actually came out of work that was done trying to encode and transport voice througha packet switch. It was found that a tremendous amount of buffering was needed, in order to allow forthe appropriate reassembly after transmission was completed. This in turn led to trying to find a wayto deliver the packets without requiring a guaranteed level of reliability. In essence, the UDP (UserDatagram Protocol) was created to allow users to make use of IP. In addition, it was also in 1978 thatthe first commercial version of ARPANET came into existence as BBN opened Telenet.In 1981-82 the first plans were being made to “migrate” from NCP to TCP. It is claimed by somethat it was this event (TCP was established as THE protocol suite for ARPANET) was truly the birthof the Internet - defined as a connected set of networks, specifically those with TCP/IP. A few yearslater (in 1983) another major development occurred, which later enabled the Internet to scale with the“explosive” growth and popularity of the future Internet. This was the development of the name server(which evolved into the DNS) [Cerf, Leiner]. The name server was developed at the University ofWisconsin [Hobbes] This made it easy for people to use the network since hosts were assigned namesand it was not necessary to remember numeric addresses. Much of the credit for the invention of theDNS (domain name server) is credited to Paul Mockapetris of USC/ISI [Cerf].The year 1983 was also the date for two other key developments on ARPANET. The first one wasthe cutover from NCP to TCP on the ARPANET. Secondly, ARPANET was split into ARPANET andMILNET. Although the road was convoluted, this split was one of the key bifurcations points thatlater allowed NSFNET to come into existence. Soon thereafter (in 1984) the number of hosts on theARPANET had grown to 1000, and the next year in 1985 the first registered domain was assigned in15March.In 1985 NSFNET was created with a backbone speed of 56 kb/s. Initially there were 5 supercomputingcenters that were interconnected. One of the paramount benefits of this was that it allowed anexplosion of connections (most importantly from universities) to take place. Two years later in 1987,NSF agreed to work with MERIT Network to manage the NSFNet backbone. The next year (1988)the process of upgrading the NSFNet backbone to one based on T1 (i.e., 1.5 Mb/s links) was begun.In 1987 the number of hosts on the Internet broke the 10,000 number. Two year later in 1989 this hadgrown to around 100,000, and 3 years after that in 1992 it reached the 1,000,000 value. It is noted thatif you look at how the number of hosts had been growing from 1984 to 1992 that it was still prettymuch tracking a growth curve that was LESS than tripling each year (i.e., doubling every 9 months).In the 1985-86 time frame key decision was made that had very long term impact: that TCP/IP wouldbe mandatory for the NSFNet program.In the 1988-1990 time frame a conscious decision was made to connect the Internet to electronicmail carriers, and by 1992 most of the commercial email carriers in the US were “like the Internet”.This was still another development that cemented email as the single most important application to takeadvantage of the Internet.In 1990 the ARPANET ceased to exist, and arguably NSFNet was the essence of the Internet. Thefollowing year Commercial Internet Service Providers began to emerge (PSI, ANS, Sprint Link, toname a few) and the Commercial Internet Xchange (CIX) was organized in 1991 by commercial ISPsto provided transfer points for traffic. NSF’s lifting the restriction on the commercial use of the Net wasagain one of the pivotal decisions. This was again a key bifurcation point, in that this helped set thestage for the complete commercialization of the Net that would follow only a few years later. In 1991the upgrading of the NSFNet backbone continued as the work to upgrade to a T3 (i.e., 45 Mb/s links)began. It also interesting to note that it was the next year (1992) than the term “surfing the Internet”was first coined by Jean Armour Polly [Polly], only two years before the ARPAnet/Internet celebratedits 25th anniversary.It was in the 1993-1995 time period that several major events seemed to emerge which fueled analmost explosive growth in the popularity of the Internet. One of the key ones was the introduction of“browsers” most notable Mosaic. This led to the creation of Netscape that went public in 1995. Evenas early as 1994 WWW(i.e., predominantly html) traffic was increasing in volume on the Net. By thenit was the second most popular type of traffic, surpassed only by ftp traffic. However, in 1995 WWWtraffic surpassed ftp as the greatest amount of traffic. In addition the traditional online dial up systems16such as AOL, Prodigy and Compuserve began to provide Internet access.In 1996 the net truly became public with the NSFNet being phased out. Soon thereafter majorinfrastructure improvements were made within the transport part of the Internet. The Internet began toupgrade much of its backbone to OC3-OC12 (up to 622 Mb/s) links, and in 1999 upgrades began formuch of the Net to OC-48 (2.5 Gb/s) links.