Previous | Table of Contents | Next |
When users access a Web page, they transmit a universal resource locator (URL) address that represents the document they wish to view. This address is transported using the HTTP within a packet. The HTTP consists of an address that totals fewer than 100characters, which are used to frame the address the message is being transported to, as well as the address of the originator of the request. The destination Web server uses the document address to locate the requested page on the server, retrieves it from disk, and forms a packet using the source address from the incoming packet as the destination address for the outgoing packets. If the requested document contains a full screen of text, the packet contain close to 2,000 characters, because a full screen of text consists of 80 columns by 24 rows of data (i.e., 1,920 characters). However, because a typical Web page contains one or more graphics, the total amount of data transmitted from the server to the user will be in actuality substantially more than 2,000 characters. For example, it is assumed that the Web page in question includes a 3 ¥ 3-inch photograph, drawing, or schematic diagram that has been scanned using a resolution of 300 dots per inch. Regardless of the color of the image, each square inch of the image requires 11,250 bytes of storage. If the image was scanned using a 256-color resolution, each pixel requires a byte to represent its color, resulting in 90,000 bytes of storage per square inch. Thus, a 3 ¥ 3-inch color image requires 270,000 bytes of storage.
Because HTTP breaks large files into small packets for transmission, the image might be carried by a sequence of approximately 100 packets, each roughly 2,700 bytes in length, to include packet overhead. Thus, the short, 100-character transmission from a user can result in a response of 280,000 bytes. Because a user connected to the Web typically clicks on hotlinks that represent document addresses to view other documents, most Web operations represent asymmetrical transmission, that is, more transmissions return to the user than the user actually originates. Thus, a high speed cable channel with a low speed reverse path occurring over the switch telephone network may actually be sufficient for most data transmission applications.
The previously described asymmetrical transmission operation of users was also recognized by Intel Corporation, which took it into consideration when designing its CablePort cable modem system. That cable modem is designed to provide an average downstream data rate of 27M bps and a 96K bps upstream rate. One interesting difference between Zenith and Intel concerning their cable modem systems is in the type of adapter card required to be used in the PC. Because Intel provides a higher downstream operating rate than what is usable by a 10BASE-T adapter card, the user must install a Fast Ethernet (100M bps) adapter card in the PC to be able to use the Intel cable modem. Although no commercial costs were provided by Zenith or Intel for field trial operations, it is worth noting that a Fast Ethernet adapter has a retail cost of approximately $250, whereas a 10BASE-T adapter can be obtained for less than $50.
A second difference between the Zenith and Intel modems concerns their upstream capability. Although Zeniths new architecture permits support of the switched telephone network for locations where CATV operators cannot provide reverse direction transmission, the Intel system did not offer this capability when this chapter was researched.
Although the technology of cable modems is in its infancy, the data center manager can still plan for their use. Whereas the type of cable modem offered will depend upon the CATV operators cable infrastructure (i.e., either unidirectional or bidirectional), as well as the cable modem vendor the data center manager selects, each cable modem requires the use of RG-11 coaxial cable. Thus, if the manager has previously installed such cabling as part of a legacy terminal to mainframe or legacy LAN wiring system and are considering its removal, he or she may wish to leave the cabling in place. If RG-11 cabling has not been installed, the data center manager may wish to consider contacting the local CATV operator to determine when cable modems will be supported and the type the operator intends to use. If it intends to use bidirectional transmission via RF modulation, the data center manager can develop a wiring plan that requires only the use of RG-11 cable. If the CATV operator says it intends to provide reverse transmission via the public switched telephone network, the wiring plan must be modified to ensure that each cable modem user will have an available telephone jack. By understanding how cable modems operate and planning the organizations wiring infrastructure to use this evolving technology, the data center manager will be prepared for its future use.
Previous | Table of Contents | Next |