TopWare Tools

home *** CD-ROM | disk | FTP | other *** search

/ TopWare Tools / TOOLS.iso / tools / top1379 / history.txt < prev next >

Wrap

Text File | 1994-02-22 | 13.4 KB | 243 lines

1) The History of Ethernet Ethernet was born on May 22, 1973 in a memo at the Xerox Palo Alto Research Center which proposed a local network called the Ether Network which would connect personal computers together. The Ether referred to the "luminiferous ether" which was thought to be the passive medium that served in the propagation of electromagnetic waves through space. Through convenience the Ether Network name was soon shortened to Ethernet and was released to the public under this name in a paper written by its Xerox developers, Robert M. Metcalfe and David R. Boggs in 1976. Their paper, "Ethernet: Distributed Packet Switching for Local Computer Networks", described Ethernet at a branching broadcast communication system for carrying digital data packets among locally distributed computing stations. Their paper also covered a experimental Ethernet designed to handle 256 stations connected to low-loss coaxial cable with a data transfer speed of 3 megabits per second. In 1979 Digital Equipment Corporation saw the possibilities of using Ethernet which would allow its computers to be interconnected with Xerox computers. Intel Corporation also saw this need which led to a joint collaboration of the three companies on developing a standard for the Ethernet. What was produced in September 1980 was Version 1 of a 10 Mbps Ethernet specification that invented an open, non-proprietary, industry standard local network. In late 1981, formal liaison was established between the IEEE and the European Computer Manufacturers Association (ECMA), and the result was the stated objective that all standards should be in accordance with the open system interconnection (OSI) reference model of the International Standards Organization (ISO). The IEEE then began work on Local Area Network standard that would bring the Ethernet specification within the framework of the OSI reference model. In November 1982, Xerox, Digital Equipment Corp, and Intel Corp. released an updated version 2 of its Ethernet specification. The purpose of this update was to bring the Ethernet specification more in line with the specification the IEEE was working on. On June 4, 1983, the IEEE approved the 802.3 Carrier Sense Multiple Access with Collision Detection Access Method and Physical Layer Specifications. The American National Standards Institute approved the 802.3 standard the following year on December 21, 1984. Even though the name "Ethernet" is registered with Xerox Corporation it is still used in the networking community to refer to the IEEE standard as 802.3 Ethernet. Supplements to the 802.3 standard have been approved by IEEE covering implementation of the 803.2 standard on different physical media. 802.3a for type 10BASE2 "thin coax" and 802.3b type 10BROAD36 were approved by IEEE in 1985. Type 802.3e 1BASE5 "twisted pair" was approved by IEEE in 1987. 2) Operation of Ethernet All of the Ethernet specifications provide precise detailed definitions of the lowest two layers of an overall network architecture defined in the ISO reference model for Open Systems Interconnection (see Fig. A-1) The importance of conforming to the OSI model is to provide compatibility between Ethernet implementations that allow a stable, easy-to-use, economical, and long-term communication solution for its users. The Physical layer defines the all the essential physical characteristics of Ethernet, including bit encoding, timing, voltage levels, and transmission media type. The Data link layer defines the medium-independent operations of Ethernet which covers two main functions of 1)Link management and 2) Data encapsulation/decapsulation. OSI MODEL APPLICATION PRESENTATION IEEE LAN MODEL SESSION > TRANSPORT Logical Link Control (LLC) IEEE 802.2 NETWORK < Media Access Control (MAC) DATA LINK IEEE 802.3 < > PHYSICAL Physical IEEE 802.3 < > Figure A-1 Link management is concerned with the channel allocation and contention resolution of the shared physical medium managed by the distributed control policy of CSMA/CD. An Ethernet station wishing to transmit is contending for the use of the shared communications channel until it acquires the channel. After the channel is acquired, the station uses it to transmit a packet. Before acquiring the channel, the station checks whether the channel is being used by monitoring the carrier and deferring transmission of the packet until the channel is quiet (no carrier). A carrier on the channel represents another station using the channel to transmit. While a station is transmitting a packet it is also listening for collisions which occur when other stations are attempting to use the channel simultaneously. In a correctly functioning system, collisions occur only within a short time interval following the start of transmission, since after this interval all stations will detect carrier on the channel and defer transmission. This time interval is referred to as the "collision widow" and is a function of end-to-end propagation delay between the farthest two points on the physical medium. If no collisions occur during this time, a station has acquired the channel and continuous transmission of the packet. If a station detects a collision, the transmission of the rest of the packet is immediately aborted. All stations that detected a collision now send a jamming signal onto the channel is order to ensure that all stations who were parties to a collision have detected it. Each station involved in a collision reschedules its packet for transmission at some later random time period. The station can attempt to retransmit 15 more times before it gives up which results in a packet error. A random delay time period is done by each station in order to minimize possible repeated collisions. Truncated Binary exponential back-off algorithm is used to calculate the delay period (called retransmissions slot) which is the number of slot times to delay before the nth transmission attempt is chosen as a uniformly distributed random integer r in the range: 0 <= r <= 2 ^ k where k= min(n,10). Stations accept packets addressed to them and discard any that are found to be in error. The deference of packets reduces the probability of collision, and collision detection allows the timely retransmissions of a packet. Packet delivery cannot be fully guaranteed due to a station not being in receive mode will miss the error-free packet sent to it. It is then up to the higher level protocol operating outside of Ethernet to detect this error. Data encapsulation function of the Data Link layer comprises the construction and processing of frames which includes addressing and error detection. All Ethernet packets begin with a known pattern of bits called the preamble. This is used by the receiver to establish bit synchronization and then locate the first bit of the packet. The preamble is inserted by the sending station and stripped off by the receiving station. A 64-bit preamble is used to ensure synchronization of phase-lock loop receivers. The Ethernet packet includes fields for both source and destination addresses. These are the physical addresses of the stations attached to the channel which should be distinct from the physical addresses of all other stations on the channel. A packet sent from one station to another station is called single-station addressing. Multicast address is the name given to a mechanism by which packets may be targeted to more than one destination. Broadcast is a predefined multicast address which always denotes the set of all stations attached to the channel. The broadcast address should be used only when strictly necessary since excessive broadcasting of packets can quickly overload stations. The first bit of a data link address distinguishes physical from multicast addresses. A 0 in bit position one indicates a physical address while a 1 in bit position one indicates a multicast address. Ethernet Version 1,2 Packet Format PREAMBLE DESTINATION SOURCE TYPE DATA CRC 8 Bytes 6 6 2 46-1500 4 IEEE 802.3 Packet Format PREAMBLE SFD DESTINATION SOURCE LENGTH DATA PAD CRC 7 Bytes 1 6 6 2 46-1500 4 Each Ethernet packet is ended with a frame check sequence(FCS) which is used to detect all packet errors due to lost or added bits in the frame from time of transmission to reception. A 32-bit cycle redundancy check (CRC) is performed to calculate the FCS value. Upon detection of an incorrect FCS by the receiver has two options to act upon; either discard the packet or deliver the damaged packet with an appropriate status indicating a FCS error. The length of an Ethernet packet has a minimum size limit of 64 bytes and a maximum size limit of 1518 bytes. The minimum size of 64 bytes was chosen since the time taken to send 64 bytes is greater than the collision window time period. Several factors have limited the maximum packet size to 1518 bytes. The desire to limit buffer size on both receiving and sending stations and the need to avoid tying up the channel that increases delay for other stations trying to gain access to the channel. Frames must have a minimal spacing period between them of 9.6 microseconds on the channel. This is required to allow a station to be able to receive consecutive packets without missing one. A station has some small delay in restarting its receiver from the time it receives a packet. 3) Differences between Ethernet V1.0, V2.0, and 802.3 specifications The main differences among the three versions lie in the connection from the data-terminal equipment (DTE) controller to the media-access unit (MAU) transceiver interface. For instance, the line-idle state was changed from 0.7 volts idle (Version 1) to zero volts idle (Version 2). As a result, the first bit of the message preamble for Version 2 starts with a voltage half step instead of a voltage full step. Ethernet 2.0 and 802.3 include a "heartbeat" function which is a signal sent from the MAU to the station that confirms that MAU collision signalling is working and connected to the station. Without this signal, the station is unsure whether the frame was actually sent without a collision. Ethernet 1.0 does not have this function. In addition, both 802.3 and Ethernet 2.0 have a jabber function. This is a self-interrupt capability that allows a MAU to inhibit transmit data from reaching the medium if the transmission occurs for longer than the maximum frame size. Because of these differences, a controller designed according to one specification may not work with a transceiver designed according to another. The Ethernet (both versions) frame format differs significantly from the IEEE 803.3 frame format. Ethernet 1.0 and 2.0 do not include a length field, and therefore, expects some higher layer protocol to do the padding. The Ethernet 1.0 and 2.0 type field is used to determine which client protocol the frame is for. 4) Ethernet Performance The first step in analyzing an Ethernet Local Area Network is to measure the network's overall health. This is done by measuring and understanding the network's current utilization (both continuous and peaks), traffic variation over time, and data link errors. Utilization is the time the network is busy moving data versus the time the network could be moving data within the rules of the access method. This is often calculated as a percentage of utilization, frames per second, or bits per second. Although frames per second may be a good measure of arrival times for a node, it is a poor measure of network utilization because of the wide variation in packet sizes. Depending on needs, different measures are appropriate. For utilization of the network, percentage utilization or aggregate bit rate is most appropriate. When looking at a station's ability to process incoming data, the rate at which packets arrive may be more important. Please see the section 'Getting the most from LAn BOSS'.