Multiplex communications – Data flow congestion prevention or control – Control of data admission to the network
Reexamination Certificate
1998-12-01
2002-12-10
Kizou, Hassan (Department: 2662)
Multiplex communications
Data flow congestion prevention or control
Control of data admission to the network
C370S232000, C370S233000, C370S477000
Reexamination Certificate
active
06493316
ABSTRACT:
FIELD OF INVENTION
The invention resides generally in the field of digital data transmission through a network. In particular, it relates to an apparatus for and method of transmitting digital data in streams of packets while observing guaranteed minimum and/or maximum bandwidth allocations.
BACKGROUND OF INVENTION
In contrast to circuit based transport systems, a packet based transport system allows the access bandwidth to be dynamically allocated. Remote nodes can be represented as logical ports but there is no commitment of bandwidth when this is not needed. The physical access link is fully available for traffic to any destination. In a packet transport system, virtual pipes are provided between any two transport access points. These pipes may be guaranteed some minimum rate of transmission but certainly it is required that an access point can make opportunistic use of spare bandwidth up to some maximum amount. Frame relay as a packet based transport allows more efficient use of bandwidth by permitting statistical multiplexing of data streams, thus allowing it to exploit unused bandwidth. However, there is no mechanism (protocol) to ensure reliable delivery of the frames and under congestion conditions frames are discarded and the higher layer protocols must compensate for the loss. The discard of frames is not sensitive to the impact on the higher layer protocols and the frame flows do not directly adapt to the network conditions. ATM with an effective flow control can provide a lossless but dynamic transport. However, it relies on some reasonable level of complexity at the transport switching points to achieve the flow control, the effectiveness of which has not yet been proven in the field. ATM without flow control requires that cells be discarded under congestion conditions and this discard should be aware of packet boundaries and the impact on the higher layer protocol such as TCP protocol. These issues are just beginning to be understood.
It is commonly understood in the field of the present invention that a layer under the networking layer is called “transport” layer and provides pipes between networking layer nodes. This is in contrast to the layered model of the OSI (open systems interconnect) in which the transport layer resides upon the network layer, which in turn sits on top of the data link layer. The data link layer provides similar functionalities to those of the transport layer of the present description. Throughout the specification, the former designation is used.
Therefore in the TCP/IP model, IP layer resides under TCP layer. The IP layer is the network layer in which IP (Internet protocol) runs. An internetwork differs from a single network, because different parts may have wildly different topologies, bandwidth, delays, packet sizes, and other parameters. The TCP layer is the transfer layer in which the TCP (transmission control protocol) runs. The TCP has been used for ensuring reliable transfer of byte stream data between two end points over an internetwork, which may be less reliable. TCP allows a shared and adaptive use of available bandwidth of a transmission link between two end points. It does this by having the sender gradually increase the rate of sending until a packet is lost whereupon it reduces its rate significantly and repeats the gradual increase. Thus, TCP tends to give connections their proportional share of the available bandwidth on a link although different connection characteristics can cause large variations in the sharing.
In Internet terminology, aggregating traffic streams by encapsulating them into a single IP stream is often called tunneling. Applicant's copending patent application Ser. No. 09/066,888 filed on Apr. 28, 1998 describes an invention, which re-uses TCP in a packet based transport to provide TCP tunneling which can conveniently be called “TCP trunking”. The use of TCP provides for reliable delivery of data between two transport access points while permitting that transport to offer elasticity and bandwidth sharing. Aggregating traffic streams into TCP tunnels reduces the size of buffers and tables in the transport switches. TCP is well suited to the use of first-in-first-out queues and allows simple implementations at the switching nodes. TCP is also inherently provides for re-sequencing of out-of-order packets which can occur when switching nodes spread load over multiple links.
It is expected that in future networks, particularly those using TCP trunking between aggregation points, it will be required to assure a minimum sending rate for the connection while still allowing the connection to probe for more throughput if it is available. This opportunistic acquisition of bandwidth might be limited to some maximum. TCP currently has no capability to support a guaranteed minimum bandwidth or a maximum permitted bandwidth but this will be essential as IP networks introduce virtual private networks with performance guarantees. The present invention offers a good solution to such a problem.
It should be noted that in a carrier owned transport network the TCP functionality would be in transport access points rather than host computers, the variability of paths will be low and parameters such as connection round trip time will be very stable. These features make it much easier to envisage modifications to the TCP protocol for these networks. However, the use of this invention can be more generally applied to other TCP hosts and other sliding window protocols.
Many flows on today's networks carry only a small amount of information, say ten or twenty packets, and the rate adaptation feature of TCP is not as relevant as the reliability feature. Providing reliability without incurring TCP time-outs on the loss of a single packet would greatly improves user throughput and network efficiency.
OBJECTS OF INVENTION
It is therefore an object of the invention to provide sliding window based flow control for point-to-point connections in a packet network, which constrain the minimum and/or maximum bandwidths available to that connection.
It is another object of the invention to provide modifications to the TCP protocol to provide for constraining the minimum and/or maximum bandwidths achieved by the connection.
It is yet an object of the invention to ensure the absence of TCP retransmission timeouts for flow with small windows by maintaining a minimum rate of sending.
SUMMARY OF INVENTION
Briefly stated, in accordance with one aspect, the invention is directed to a method of sending data in packets via a connection by way of sliding window algorithm in which a flow of data into the connection is controlled in response to acknowledged packets and the connection observing either or both of a guaranteed minimum bandwidth and a maximum permitted bandwidth. The method comprises steps of (1) calculating a congestion window hereinafter called C-WND of the connection, (2) setting either or both of a guaranteed minimum bandwidth and a maximum permitted bandwidth and then respectively calculating either or both of a guaranteed minimum bandwidth window hereinafter called MIN-WND and a maximum permitted bandwidth window hereinafter called MAX-WND. The method further includes steps of (3) determining if the MIN-WND or MAX-WND is invoked on the connection, based on their relationship with C-WND and (4) allowing the transmission of one or more packets of data into the connection if either MIN-WND or MAX-WND permits said transmission.
According to another aspect, the invention is directed to an apparatus for sending data in packets via a connection by way of sliding window algorithm in which a flow of data into the connection is controlled in response to acknowledged packets and the connection observing either or both of a guaranteed minimum bandwidth and a maximum permitted bandwidth. The apparatus comprises a flow control module for controlling a flow of packets into the connection in response to acknowledged packets, a congestion window arithmetic module for calculating a congestion window hereinafter called C-WND of the connecti
Chapman Alan Stanley John
Kung Hsiang-Tsung
Haszko Dennis R.
Kizou Hassan
Nortel Networks Limited
Tsegaye Saba
LandOfFree
Apparatus for and method of managing bandwidth for a packet... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Apparatus for and method of managing bandwidth for a packet..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Apparatus for and method of managing bandwidth for a packet... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2925877