Electrical computers and digital data processing systems: input/ – Input/output data processing – Flow controlling
Reexamination Certificate
1998-03-06
2001-02-20
Lee, Thomas C. (Department: 2782)
Electrical computers and digital data processing systems: input/
Input/output data processing
Flow controlling
C710S036000, C710S052000, C710S053000, C710S240000, C709S232000, C709S235000
Reexamination Certificate
active
06192422
ABSTRACT:
FIELD OF THE INVENTION
This invention relates generally to computer networking devices. More particularly, this invention relates to method and apparatus for increasing the data throughput of local area computer networks (LANs), particularly networks defined by IEEE standard 802.3 and commonly known as Ethernet networks.
BACKGROUND OF THE INVENTION
A local area network (LAN) is a system for directly connecting multiple computers so that they can directly exchange information with each other. LANs are considered local because they are designed to connect computers over a small area, such as an office, a building, or a small campus. LANs are considered systems because they are made up of several components, such as cable, repeaters, network interfaces, nodes (computers), and communication protocols.
Every LAN type has a set of rules, called topology rules, that dictate how the components of the network are physically connected together. Ethernet is one such set of topology rules. Background information on the Ethernet specifications and computer networks can be found in a number of references such as the IEEE 802.3 standards, Fast Ethernet (1997) by L. Quinn et al., and
Computer Networks
(3rd Ed. 1996) by A. Tannenbaum, which are incorporated herein by reference. Ethernet operates as a bussed network in which each of the nodes connects to a common bus. On early Ethernet networks, all the nodes were literally attached to a single segment of cable (the bus) with T connectors. The network could be extended by connecting pieces of cable together with two-port repeaters. These repeaters “repeat” signals transmitted through the cable by restoring the signal's shape and strength to its original characteristics.
Newer Ethernet technologies improved on these early repeaters by introducing the concept of a repeater hub (often called just a hub or a repeater). A repeater hub, shown in a block diagram in
FIG. 1
, is a device that each node on the network plugs into instead of having a T connection to a common cable. A repeater hub replaces the cable and T connection of the bussed network but behaves just like the shared cable. Each node, which may be a personal computer, server, printer, etc., connects to a port of the central hub via a cable, with only one node per cable. This arrangement creates a “Hub and Spoke” or “Star” topology as shown in
FIG. 1
that operates as a bussed network. Inside the hub is a digital bus that connects to multiple ports. The ports of a typical repeater hub operate exactly as the ports of early repeaters, except that a hub has many more ports than the two found in the early repeaters.
Conventional repeater hubs, however, do not address the problem that the maximum size of a LAN shrinks as the data rate of the LAN increases. This increase has occurred as Ethernet, which operates at 10 megabits per second (Mbps) has been extended to Fast Ethernet (100 Mbps) and presently to Gigabit Ethernet (1000 Mbps). With conventional network components including repeater hubs, the increased rate requires a reduction in the diameter of the local area network (the maximum cable distance between two nodes). In the move from standard Ethernet at a network speed of 10 Mbps to Fast Ethernet at a network speed of 100 Mbps, the allowable network diameter shrank from 2.5 kilometers to 250 meters, a factor of 10. The same effect will occur when network speed is increased from 100 Mbps to 1000 Mbps. The theoretical allowable maximum diameter will be reduced to 25 meters.
The reason for this reduction in network diameter relates to the media access rules adopted by the IEEE 802.3 committee for Ethernet networks (known as the CSMA/CD rules for Carrier Sense Multiple Access with Collision Detection) and to the physical nature of the components making up a network. Briefly, under the CSMA/CD rules, each of the multiple nodes on a network (forming a “collision domain”) first listens for a carrier on the shared network media (e.g., cable) before transmitting a data packet to other nodes (the carrier sensing). Once the network media is free, nodes with a pending packet may transmit the packet. If two or more nodes simultaneously transmit packets on the network media, however, the packets collide. Ideally the sending nodes detect the collision and resend the corrupted packets after random delay times. These access rules are implemented in each node by a media access controller (MAC) device.
Collisions occur because signal propagation delay within the network components prevents a second node from immediately sensing when a first node has begun a transmission. For example, assume that the network media is clear and that the first and second nodes have packets to transmit. The first node then begins its transmission. The second node will not be aware of the first node's transmission until it actually reaches the second node because of the propagation delay. During that delay the second node may begin its own transmission, and the two packets will collide. This situation is called contention, as the nodes contend for control of the media rather than defer to one another.
The time difference, in terms of propagation delay, between two particular nodes on an Ethernet network is called the Path Delay Value (PDV). The PDV for two particular nodes is calculated by adding up the individual propagation delays of each component between the MACs at each node and multiplying the total by two. This is the time it takes for a bit to travel round trip from one node to another and then back. The maximum PDV for a network is called the “collision window” and is directly dependent on the network diameter. The larger the network diameter is, the larger the network's collision window.
The Ethernet specification defines the maximum allowable collision window to be
512
bit times This value is called the “slot time.” Two values are derived from the slot time: the minimum frame size of 512 bits (64 bytes) and the maximum allowable network diameter. The network diameter must be small enough that a signal can start from a MAC on one node and travel to a MAC on any other node and back inside the slot time.
Stated another way, the network's collision window must be less than or equal to the slot time.
A maximum collision window is specified because, under the CMSA/CD rules, a node only detect collisions while it is sending a frame.
Once the node completely sends the frame, it assumes the transmission was successful. If the network's collision window exceeds the slot time, a node can completely send a frame before the node detects that the frame has collided with other data on the media. At that point, however, it is too late. The node will not automatically retransmit the frame. The network protocol must then recover the lost frame in a lengthy process that temporarily but significantly degrades network performance.
With this as background, it now can be understood why, with conventional network components, faster network speeds require a reduction in network diameter. As described above, a node is assured of detecting a collision only during the time it is transmitting a minimum-sized frame. This is the slot time, which is specified to be 512 bit times. The maximum allowable collision window is thus also 512 bit times. As the network speed increases from 10 to 100 to 1000 Mbps, the slot time required to transmit
512
bits decreases by a factor of 100 to 512 nanoseconds. Because of the signal propagation delay, the maximum allowable network diameter must be reduced accordingly, or the nature of the network makeup itself must be changed, to ensure that collisions are detected within the reduced slot time.
A common solution to this problem of network size reduction is to break up a network consisting of a single collision domain into multiple smaller collision domains and connect the multiple domains together with frame switches. See, for example, chapter 12 of
Fast Ethernet
(1997) by L. Quinn et al. Each of the smaller collision domains has the maximum allowable network diameter. The entir
Daines Bernard Nelson
Lawrence Frank S.
Alcatel Internetworking, Inc.
Christie Parker & Hale LLP
Lee Thomas C.
Nguyen Tanh
LandOfFree
Repeater with flow control device transmitting congestion... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Repeater with flow control device transmitting congestion..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Repeater with flow control device transmitting congestion... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2563847