Ether ring architecture for local area networks

United States Patent

[19]

Miller

US006028837A [ii] Patent Number: 6,028,837 [45] Date of Patent: *Feb. 22,2000

[54] ETHER RING ARCHITECTURE FOR LOCAL AREA NETWORKS

[75] Inventor: David J. Miller, Lawrenceville, Ga.

[73] Assignee: National Semiconductor Corporation,

Santa Clara, Calil.

[ * ] Notice: This patent issued on a continued prosecution application filed under 37 CFR 1.53(d), and is subject to the twenty year patent term provisions of 35 U.S.C. 154(a)(2).

[21] Appl. No.: 08/703,454 [22] Filed: Aug. 27, 1996

[51] Int. CI.7 H04L 12/433

[52] U.S. CI 370/222; 370/452; 370/909;

375/212; 340/825.06

[58] Field of Search 370/293, 257,

370/258, 221, 222, 246, 241-243, 248, 462, 463, 501, 502, 507, 508, 516-519, 522, 528, 509, 908-910, 492; 340/825.05, 825.06; 375/211, 212, 213, 214

[56] References Cited

U.S. PATENT DOCUMENTS

4,514,841 4/1985 Sem-Sandberg 370/463

4,701,910 10/1987 Ulug 370/228

4,759,009 7/1988 Casady et al 370/221

4,760,571 7/1988 Schwarz 370/445

4,845,709 7/1989 Matsumoto et al 370/452

5,155,726 10/1992 Spinney et al 370/296

5,168,272 12/1992 Akashi et al 340/825.05

5,181,202 1/1993 Walser et al 370/452

5,243,335 9/1993 Kato et al 340/825.05

5,247,518 9/1993 Takiyasu et al 370/466

5,257,259 10/1993 Tsurami 370/354

5,274,637 12/1993 Sakamura et al 370/455

5,305,306 4/1994 Spinney et al 370/296

5,317,198 5/1994 Husbands 370/222

5,377,228 12/1994 Ohara et al 375/211

5,530,694 6/1996 Guezou et al 370/223

5,555,540 9/1996 Radke 370/462

5,652,758 7/1997 Donley 370/503

A ring topology based architecture for a local area network. The transmission medium is unshielded twisted-pair wire. A set of two such twisted-pairs are used to connect the nodes of the network together, with the transmit twisted-pair wires from one node being connected to the receive twisted-pair wires of the next node. When a node of the network is powered up, a relay is opened which disconnects the incoming receive lines from the outgoing transmit lines. This brings the node on line. The relay is controlled via software command and a timer which is used to take a preceding node off-line for a predetermined period in the event jabber from that node is detected. A 16 byte FIFO bypass built into the medium access control hardware of the system is used to reduce the ring latency which normally results from the need to store entire data packets at each node, prior to the packet being forwarded. The FIFO bypass is used to detect the source and destination addresses for the data packets, and to determine if a packet should be forwarded along the ring. If a node is transmitting a packet when it receives a new packet, the incoming data is buffered until the current transmission is completed. This avoids the collision problem common in bus topologies.

9 Claims, 2 Drawing Sheets

1

ETHER RING ARCHITECTURE FOR LOCAL
AREA NETWORKS

TECHNICAL FIELD

The present invention relates to architectures for local area networks, and more specifically, to a ring topology networking architecture using unshielded, twisted-pair fullduplex technology which provides improved performance and lower cost compared to existing ring and bus-based topology networks.

BACKGROUND OF THE INVENTION

Local area networks (LANS) are data transmission systems consisting of a set of nodes which are interconnected 15 by links. The nodes may be terminals, computers, printers, storage devices, etc. The links may be coaxial cable, twistedpair wires, or fiber-optic cable. LANs are used for such purposes as transmission of messages among the nodes, resource sharing, and transfer of data or files from one 20 storage medium to another.

Three characteristics that are used to differentiate LAN architectures are network topology, the transmission medium, and the method of controlling access to the medium. Each combination of these characteristics has 25 inherent advantages and disadvantages which determine its suitability for a particular application.

The network topology refers to the manner in which the endpoints or nodes of the network are interconnected. The specific topology used in a LAN is important as it deter- 30 mines the data paths that may be used between any pair of nodes. The four primary network topologies used in LAN systems are the bus, tree, ring, and star topologies.

In a ring topology, the network consists of a set of 3J repeaters joined by point-to-point links in a closed loop. Each repeater receives data from one link and transmits the data, bit by bit, on the second link to which it is attached. The data is transmitted as fast as it is received with no buffering at the repeater. The links are unidirectional so that data is 4Q only transmitted in one direction. Each node or station of the network attaches to the LAN at a repeater. Data is transmitted in packets, with each packet containing control information used by each repeater in the ring to determine

whether to pass the data to the station attached to that „

45

repeater (the destination station for the data) or to the next repeater in the ring.

In a bus topology based LAN, the network is the transmission medium. All of the nodes are directly attached by means of the appropriate hardware to a linear transmission 50 medium or bus. Data transmitted from any one node propagates the length of the medium and can be received by any other node connected to the medium.

The transmission medium is the path between the nodes (or repeaters) of a LAN. Typical mediums used in LANs are 55 twisted-pair, coaxial cable, and optical fibers. Twisted-pair is the medium used to connect the telephones in a building together, or to connect the phones in a local geographic area to a central exchange. It consists of two insulated wires arranged in a spiral pattern. A wire pair serves as a com- go munication link between nodes or stations.

One well-known LAN architecture is that named "Ethernet", which is based on a bus topology. In a twistedpair medium implementation of Ethernet, the nodes or stations are attached to a central hub. The hub functions both 65 to connect all of the nodes to a common bus and also acts as a repeater for the data packets. In the Ethernet architecture,

2

the nodes are connected to the central hub by means of two twisted-pair transmission mediums, one for transmission to the hub, the other for receiving data from the hub. When a node transmits a data packet, it is received by the hub and then re-transmitted to each of the nodes connected to the hub. An Ethernet LAN typically operates at a data rate of 10 mBit/sec (10 million bits/sec).

Although the Ethernet architecture is suitable for use in many environments, it does have some disadvantages. Ethernet systems which utilize twisted-pair wires as the transmission medium require a central hub or repeater, which in the case of a 100 mBit/sec (100 million bits/sec) data transfer rate system can be quite expensive. This cost limits the use of such high data rate systems for home or small business use.

A second disadvantage is that because all nodes in an Ethernet system monitor the same bus, only one node can transmit a data packet at a time. If multiple nodes attempt to transmit at approximately the same time, collisions between data packets may occur. These collisions are detected by the nodes, resulting in the production of a jam signal which causes the nodes to cease transmitting. The nodes then reschedule their respective transmissions based on a probabilistic analysis which includes consideration of how many times each nodes' packet has collided. This method of controlling access to the medium is termed carrier sense, multiple access with collision detection, or CSMA/CD. The possibility of collisions, and the CSMA/CD control protocol used in response complicate the system requirements by requiring an additional layer of overhead. This degrades the system's performance by reducing system throughput.

Yet another disadvantage of Ethernet-type bus-based LAN architectures is that the maximum latency of transmissions between nodes is not predictable. This is because a data packet transmitted by one node or station is received by each of the other stations. As a result, it is not possible to predict the maximum time delays between transmissions from one station to another. This complicates attempts to use such an architecture for real-time interactive communications such as audio and video conferencing.

Although ring-based networks do not suffer from the collision problem noted with regards to bus networks, they also have some inherent disadvantages. In a ring-based LAN architecture, as data packets circulate around the ring, the receiver in each station recovers the binary data from the received signal. To perform this function accurately, the receiver needs to know the starting and ending times of each data bit. This allows a proper sampling of the received signal. However, accurate knowledge of this timing information requires synchronization of the stations on the ring. This is typically accomplished by encoding the clock data used to synchronize the stations into the data streams. However, in a ring-based LAN, the recovered clock signal may deviate from its intended value due to several reasons: (1) noise during transmission of the data; (2) receiver circuitry induced deviations; and (3) propagation delay distortion of the signal. This clock deviation is termed "timing jitter". In a ring architecture, this jitter accumulates as a data packet circulates from station to station, causing the data to become less reliable.

Another disadvantage of ring topology networks is the use of "tokens" for controlling access to the transmission medium and for determining packet transferral around the ring. A token is a control frame that circulates around each node of the ring when all of the stations are idle. A station having a data packet to transmit must wait until a token is