NETWORK FLOW TERMINATION

• US 20120033553A1
• Filed: 03/25/2010
• Published: 02/09/2012
• Est. Priority Date: 03/31/2009
• Status: Active Grant

• Current Assignee: British Telecommunications PLC (BT Group PLC)
• Original Assignee: British Telecommunications PLC (BT Group PLC)
• Inventors: Strulo, Ben, Wennink, Marc
• Granted Patent: US 8,625,426 B2
• Time in Patent Office: Days
• US Class Current: 370/232
• CPC Class Codes: H04L 47/10 Flow control; Congestion control
• CPC Class Codes: H04L 47/2408 for supporting different services, e.g. a differentiated services [DiffServ] type of service
• CPC Class Codes: H04L 47/26 using explicit feedback to the source, e.g. choke packets
• CPC Class Codes: H04L 47/31 by tagging of packets, e.g. using discard eligibility [DE] bits
• CPC Class Codes: H04L 47/32 by discarding or delaying data units, e.g. packets or frames

Abstract

A network has a plurality of edge nodes (7a, 7b, 7c, 7d) and core nodes (3a, 3b, 3c) for carrying flows of data from an ingress gateway (7a, 7b, 7c, 7d) to an egress gateway (7a, 7b, 7c, 7d) along a path of core nodes. For any given path in the network work from an ingress node to an egress node, the number of flows is controlled. When network congestion occurs, the egress gateway (7a, 7b, 7c, 7d) provides an indication of the level of network congestion. Terminating flows is disruptive and therefore the ingress gateway (7a, 7b, 7c, 7d) can mark other flows with a congestion marker and send them to the egress gateway (7a, 7b, 7c, 7d). These marked flows are routed by the core and egress gateway (7a, 7b, 7c, 7d) but ignored for calculating network congestion. When the network congestion is alleviated, the marked flows can be un-marked and treated as normal flows, hence the network avoids unnecessary flow termination.

Claims

1. A method of flow control on a network formed of a plurality of edge nodes and a plurality of core nodes, in which a plurality of flows, each formed of data packets, are introduced into the network from one of the edge nodes functioning as an ingress gateway, and travel on a predefined path via at least one of the core nodes to a different edge node functioning as an egress gateway, the method comprising:

each core node;

comparing a first reception rate of data packets against a first threshold value; and

if the first reception rate is greater than the first threshold value, adding a first congestion marker (TT) to the header of received packets;

the egress gateway;

monitoring the number of received unmarked data packets, packets marked with the first congestion marker and packets marked with a second congestion marker (RT);

sending a network congestion message containing the monitored data as an indicator of network congestion, to the ingress gateway;

the ingress gateway;

receiving the network congestion message from the egress gateway;

terminating a first subset of the flows such that data packets belonging to those flows are not sent to the at least one core node;

selecting a second subset of flows for possible termination;

marking received data packets belonging to the second subset of flows with the second congestion marker (RT); and

sending data packets marked with the second congestion marker (RT) to the egress gateway via the network,

wherein the method further comprises;

the core nodes ignoring any data packets containing the first congestion marker (TT) or the second congestion marker (RT) in the comparison of the first reception rate; and

the ingress gateway unselecting flows from the second subset of flows, if the ingress gateway receives a network congestion message indicating that network congestion has decreased.

2. A method according to claim 1, further comprising:

the egress gateway;

comparing a second reception rate of packets containing either the first congestion marker (TT) or the second congestion marker (RT) against a second threshold value;

if the second reception rate is greater than the second threshold value, determining a third reception rate of unmarked data packets for sending to the ingress gateway;

the ingress gateway;

calculating a difference value between the received third reception rate and a sending rate of unmarked flows being introduced onto the network;

wherein;

the number of terminated flows in the first subset of flows is a first predetermined proportion of the calculated difference; and

the number of selected flows in the second subset of flows is a second predetermined proportion of the calculated difference, and

the sum of terminated flows and marked flows in accordance with the first and second proportion is less than the calculated difference.

3. A method according to claim 1, further comprising:

the egress gateway;

calculating a ratio of the number of received unmarked packets against the number of received data packets carrying the first or second congestion marker (TT, RT) for sending to the ingress gateway; and

;

the ingress gateway;

marking flows with the second congestion marker (RT) at a predetermined marking rate; and

terminating flows from said first subset of flows at a predetermined termination rate which is a proportion of the predetermined marking rate.

4. A method according to claim 1, wherein the ingress gateway selects the set of flows for possible termination by:

accessing a flow processing policy and choosing flows in order of flow priority.

5. A network edge node for performing flow control within a network of edge nodes and core nodes for transporting a plurality of flows, each formed of data packets, along a predefined path across the network, the network edge node comprising:

a network interface for communication with at least one core network node and at least one data external network entity;

a data packet receiver for receiving data packets from the at least one core node and from the at least one external network entity;

a data packet processor determining where to send received data packets for onward transmission; and

a data packet transmitter for sending data packets to the at least one core node and to the at least one external network entity,

wherein the edge node has a first mode of operation as a network ingress gateway and a second mode of operation as a network egress gateway;

wherein in the second mode as an egress gateway;

the data packet processor is operable to monitor the number of received data packets, and distinguish between an unmarked data packet, a data packet having a first congestion marker (TT) and a data packet having a second congestion marker (RT); and

generate a network congestion message containing the monitored data as an indicator of network congestion to another edge node in the network operating as an ingress gateway; and

wherein in the first mode as an ingress gateway;

the data packet processor is operable to;

determine whether a network congestion message has been received;

terminate a first subset of flows;

select a second subset of flows for possible termination; and

mark received data packets belonging to the second subset of flows with the second congestion marker (RT);

wherein in the first mode the ingress gateway is operable to unselect flows from the second subset of flows, if a network congestion message indicating that network congestion has decreased is received.

6. An edge node according to claim 5, wherein in the second mode as an egress node the data packet processor is further operable to:

compare a second reception rate of packets containing either the first congestion marker (TT) or the second congestion marker (RT) against a second threshold value; and

determine a third reception rate of unmarked data packets for sending to the ingress gateway if the second reception rate is greater than the second threshold value; and

wherein in the first mode as an ingress gateway the data packet processor is operable to;

calculate a difference value between the received third reception rate and a sending rate of unmarked flows being introduced onto the network; and

wherein;

the number of terminated flows in the first subset of flows is a first predetermined proportion of the calculated difference; and

the number of selected flows in the second subset of flows is a second predetermined proportion of the calculated difference, and

the sum of terminated flows and marked flows in accordance with the first and second proportion is less than the calculated difference.

7. A network edge node according to claim 5, wherein in the second mode as an egress gateway the data processor is operable to:

calculate a ratio of the number of received unmarked packets against the number of received data packets carrying the first or second congestion marker (TT, RT) for sending to the ingress gateway; and

;

in the first mode as an ingress gateway the data processor is operable to;

mark flows with the second congestion marker (RT) at a predetermined marking rate; and

terminate flows from said first subset of flows at a predetermined termination rate which is a proportion of the predetermined marking rate.

8. A network edge node according to claim 5, wherein in the first mode the data packet processor is operable to choose flows for termination and marking with the second congestion marker in order of flow priority.

9. A data network for transporting flows of data packets, comprising:

a plurality of edge network nodes configured as an ingress gateway as claimed in claim 5;

a plurality of edge network nodes configured as egress gateways; and

a plurality of core nodes for routing flows across the network from the ingress gateways to the egress gateways.