Electrical computers and digital data processing systems: input/ – Input/output data processing – Flow controlling
Utility Patent
1998-04-03
2001-01-02
Lee, Thomas C. (Department: 2782)
Electrical computers and digital data processing systems: input/
Input/output data processing
Flow controlling
C710S036000, C710S107000, C709S226000, C709S235000, C370S902000, C370S231000, C370S232000, C370S235000, C370S236000, C370S227000
Utility Patent
active
06170022
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to data communications networks and more particularly to a method and system for providing congestion control in such a network.
BACKGROUND OF THE INVENTION
When data processing systems first were used commercially on a widespread basis, the standard system configuration was an autonomous mainframe or host computer which could be accessed only through locally-attached terminals. Few people, at that time, perceived any significant benefit from interconnecting host computers.
Over time, it came to be understood that significant commercial advantages could be gained by interconnecting or networking host computers. Data originating with users at one host computer could readily and rapidly be shared with users located at other host computers, even if those other host computers were many miles away. Also, the functional capabilities of a given host computer could be treated as a resource that could be shared not only among locally-attached users but also among remote, network-attached users. Mainframe networks of this type came to be generically referred to as Wide Area Networks, commonly abbreviated to WANs.
Certain parallels exist between the development of mainframe computer technology and the later development of personal computer technology. Early personal computers were relatively unsophisticated devices intended for use by a single user in a standalone configuration. Eventually, the same kinds of needs (data sharing and resource sharing) that drove the development of mainframe networks began to drive the development of networks of personal computers and auxiliary devices, such as printers and data storage devices.
While mainframe networks developed primarily using point-to-point connections among widely-separated mainframes, personal computer networks developed using shared or common transmission media to interconnect personal computers and auxiliary devices within a geographically-limited area, such as a building or even an area within a building. Networks of this type came to be generically referred to as Local Area Networks or LANs.
Different LAN technologies exist. Currently, the most popular LAN technology is Ethernet technology. In an Ethernet LAN, personal computers and auxiliary devices share a common bi-directional data bus. In the following description, LAN-attached devices are generically referred to as stations or LAN stations. Any transmission-capable Ethernet LAN station may initiate transmission on the bus. Every transmission propagates in both directions and is received by every Ethernet LAN station attached to the same bus, including the transmitting station.
Because several Ethernet LAN stations can attempt to claim the bus at the same time, a Collision Sense Multiple Access/Carrier Detect (CSMA/CD) protocol exists to resolve control data flow within an Ethernet LAN. The protocol is relatively simple. When a station has data to transmit, it “listens” to the bus to see if the bus is “busy”; that is, already carrying data from another station. If the bus is quiet, the listening station begins its own transmission immediately. If the bus is busy, the station with data to send waits for a predetermined interval before restarting the bus acquisition process.
Since electrical signals require time to propagate down any conductor, two or more Ethernet stations can listen, find the bus quiet at the same time, and begin transmitting simultaneously. If that happens, data from the transmitting stations collides and becomes corrupted. If a transmitting station doesn't detect the same data it transmitted, that station sends a short jamming signal and stops transmitting. The jamming signal increases the chances that all other Ethernet transmitting stations will detect the collision and stop transmitting themselves. Following a random delay, each transmitting station restarts the bus acquisition process.
Another well known LAN technology is token ring technology. In a token ring LAN, individual LAN stations are connected in a unidirectional ring. In a basic token ring, flow control is achieved by requiring that a station having data to send first acquire a token from the ring, a token being a special purpose frame representing a permission to send. Once a station acquires a token, it can insert its data onto the ring. As the inserted data circulates around the ring in a single direction, each station to which the data is addressed copies the data into station memory but does not remove the data from the ring. The data continues to circulate through the ring from one station to the next. When the data finally returns to the originating station, it is stripped from the ring.
The same user needs (data sharing and resource sharing) which drove the development of local area networks have driven the creation of LAN networks consisting of multiple LANs (Ethernet or token ring) interconnected through boundary devices known as LAN bridges or switches. Point-to-point connections or links between LAN switches permit traffic originating in any given LAN to be transported to a LAN station connected to any other LAN in the same switched network. A given switch-to-switch link typically carries traffic from multiple sources concurrently.
For a single Ethernet LAN, the CSMA/CD protocol provides a fairly effective flow control mechanism. Similarly, for a single token ring, the token protocol provides an effective flow control mechanism. However, where LANs are interconnected through switched networks, CSMA/CD and token protocols are unable to provide effective flow control on the interconnecting links.
To provide flow control on switch-to-switch links carrying Ethernet traffic, at least one standards group, the IEEE 802.3 working group, has developed a flow control standard (IEEE 802.3x) intended for that purpose. Under the standard (as currently developed), a downstream switch monitors traffic flow through the switch to detect any congested conditions. Conventionally, a switch is considered to be congested when the amount of traffic stored in a switch buffer (input and/or output) exceeds a predetermined threshold; for example, 80% of the buffer capacity. When congestion is detected, the switch can act to inhibit or pause transmission of data from one or more upstream switches. The congestion-detecting switch generates and transmits one or more pause frames, each of which contains, among other things, a pause time and a reserved multicast address. Pause times are expressed as a number of time slots, with a time slot being the time required to transmit a sixty-four byte packet on the link.
An upstream switch identified by the reserved multicast address responds to such a pause frame by temporarily suspending (pausing) traffic addressed to the downstream switch. The upstream switch resumes sending traffic when the pause time specified in the pause frame has elapsed. Currently, flow control is defined only on a link level rather than on a per-station level.
More specifically, if an upstream Ethernet switch receives a pause command originating in a downstream Ethernet switch, the upstream switch responds by suspending all traffic on its link to the downstream switch. The upstream switch ordinarily receives its traffic from multiple upstream sources, only some of which may actually be producing congestion-inducing traffic. However, where flow control is performed only on a link level, all sources to the upstream switch will be impacted by a pause frame, whether those sources are contributing to a congested condition or not.
To avoid penalizing upstream traffic sources which are not actually contributors to a downstream congested condition, it has been proposed that flow control be extended from a link level to the level of individual stations or traffic sources. Only stations or sources producing congestion-inducing traffic would be required to pause their traffic. Other stations would remain unaffected by pause commands.
The discussion above has been limited to flow control proposals for switched Ethernet network environments. Similar pr
Linville John Walter
Makrucki Brad Alan
Suffern Edward Stanley
Warren Jeffrey Robert
International Business Machines - Corporation
Lee Thomas C.
Schuster Katherina
Woods Gerald R.
LandOfFree
Method and system for monitoring and controlling data flow... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and system for monitoring and controlling data flow..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and system for monitoring and controlling data flow... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2457463