The Internet is a network of networks that interconnects computers around theworld, supporting both business and residential users. In 1994, a multimediaInternet application known as the World Wide Web became popular. The higherbandwidth needs of this application have highlighted the limited Internet accessspeeds available to residential users. Even at 28.
8 Kilobits per second(Kbps)the fastest residential access commonly available at the time of thiswritingthe transfer of graphical images can be frustratingly slow. Thisreport examines two enhancements to existing residential communicationsinfrastructure: Integrated Services Digital Network (ISDN), and cable televisionnetworks upgraded to pass bi-directional digital traffic (Cable Modems). Itanalyzes the potential of each enhancement to deliver Internet access toresidential users.
It validates the hypothesis that upgraded cable networks candeliver residential Internet access more cost-effectively, while offering abroader range of services. The research for this report consisted of casestudies of two commercial deployments of residential Internet access, eachintroduced in the spring of 1994: Continental Cablevision and PerformanceSystems International (PSI) jointly developed PSICable, an Internet accessservice deployed over upgraded cable plant in Cambridge, Massachusetts; Internex, Inc. began selling Internet access over ISDN telephone circuitsavailable from Pacific Bell. Internex’s customers are residences and smallbusinesses in the “Silicon Valley” area south of San Francisco,California. 2.
0 The Internet When a home is connected to the Internet,residential communications infrastructure serves as the “last mile” ofthe connection between the home computer and the rest of the computers on theInternet. This section describes the Internet technology involved in thatconnection. This section does not discuss other aspects of Internet technologyin detail; that is well done elsewhere.
Rather, it focuses on the services thatneed to be provided for home computer users to connect to the Internet. 2.1 ISDNand upgraded cable networks will each provide different functionality (e.g. typeand speed of access) and cost profiles for Internet connections. It might seemsimple enough to figure out which option can provide the needed level of servicefor the least cost, and declare that option “better.” A key problemwith this approach is that it is difficult to define exactly the needed level ofservice for an Internet connection. The requirements depend on the applicationsbeing run over the connection, but these applications are constantly changing.
As a result, so are the costs of meeting the applications’ requirements. Untilabout twenty years ago, human conversation was by far the dominant applicationrunning on the telephone network. The network was consequently optimized toprovide the type and quality of service needed for conversation.
Telephonetraffic engineers measured aggregate statistical conversational patterns andsized telephone networks accordingly. Telephony’s well-defined and stableservice requirements are reflected in the “3-3-3” rule of thumb reliedon by traffic engineers: the average voice call lasts three minutes, the usermakes an average of three call attempts during the peak busy hour, and the calltravels over a bidirectional 3 KHz channel. In contrast, data communications arefar more difficult to characterize. Data transmissions are generated by computerapplications. Not only do existing applications change frequently (e.
g. becauseof software upgrades), but entirely new categoriessuch as Web browserscomeinto being quickly, adding different levels and patterns of load to existingnetworks. Researchers can barely measure these patterns as quickly as they aregenerated, let alone plan future network capacity based on them.
The onegeneralization that does emerge from studies of both local and wide- area datatraffic over the years is that computer traffic is bursty. It does not flow inconstant streams; rather, “the level of traffic varies widely over almostany measurement time scale” (Fowler and Leland, 1991). Dynamic bandwidthallocations are therefore preferred for data traffic, since static allocationswaste unused resources and limit the flexibility to absorb bursts of traffic.This requirement addresses traffic patterns, but it says nothing about theabsolute level of load. How can we evaluate a system when we never know how muchcapacity is enough? In the personal computing industry, this problem is solvedby defining “enough” to be “however much I can affordtoday,” and relying on continuous price-performance improvements in digitaltechnology to increase that level in the near future.
Since both of theinfrastructure upgrade options rely heavily on digital technology, anothercriteria for evaluation is the extent to which rapidly advancing technology canbe immediately reflected in improved service offerings. Cable networks satisfythese evaluation criteria more effectively than telephone networks because: Coaxial cable is a higher quality transmission medium than twisted copper wirepairs