NETWORK

Technical Field

The present invention relates to the field of communication networks, in particular to control of traffic flows with different CoS (Class of Services) in a communication network.

Background Art

Nowadays, Ethernet is widely deployed in LANs (Local Area Networks) or MANs (Metropolitan Area Networks). Besides, due to its inherent easiness of implementation:, topological flexibility and cost-effectiveness, Ethernet is currently being adopted also in carrier networks for delivering different types of services, such as best effort, voice, video, CES (Circuit Emulation Services), etc.

In particular, Circuit Emulation Services (briefly, CES) provide for delivery of TDM traffic coming from a TDM network (e.g. T1/E1 traffic coming from an SDH/SONET network) through an Ethernet network. Implementing a CES service requires connecting the TDM network to the Ethernet network by means of inter-working functions that convert the TDM traffic to Ethernet packets and vice versa. The performance requirements of CES services are generally very stringent, in particular as far as packet delay, packet delay variation (jitter) and packet loss are concerned.

It is known in the Ethernet to implement Quality of Service (briefly, QoS) mechanisms at the MAC layer of the OSI model, or at a higher layer (e.g, the network layer). The QoS mechanisms typically provide for classifying the various Ethernet packets in different Classes ot Service (briefly, CoS) according to the traffic they carry, and to provide at each Ethernet node a separate queue for each CoS. The CoS of each Ethernet packet is defined by means of a 3-bits user priority field in the VLAN header of the Ethernet frame. Typically, best effort traffic is inserted in Ethernet packets with lower priority. Besides, voice, video and CES traffic is inserted in Ethernet packets having the highest priority, due to the stringent loss, delay and jitter requirements of the services they transport. QoS mechanisms provide for dropping of packets from the queues of the Ethernet nodes in case of traffic congestion and queue overflow.

Ethernet further provides for a flow control mechanism. Generally speaking, a flow control mechanism enables a downstream node to communicate an existing congestion of one or more of its input queues to an upstream node. In the Ethernet, the flow control mechanism is typically implemented by using a special type of Ethernet MAC control frame, lhal is termed "pause frame". In case of congestion of one of its input queues (i.e. the filling level of the queue exceeds a predefined threshold), the downstream node sends a pause frame to the upstream node, which in turn temporarily stops the transmission of ali Ethernet packets directed to the downstream node. The pause frame can carry a parameter field which specifies the duration of the stop. In that case, the upstream node stops transmission for the duration specified in the pause frame. Alternatively, the downstream node can issue another pause frame when the filling level of the congested input queue decreases again underneath the predefined threshold. In that case, the upstream node resumes transmission upon reception of the other pause frame. The above described flow control mechanism does not distinguish between different CoS, and accordingly is not implemented in Ethernet carrier networks, that typically implement QoS mechanisms .

Therefore, when different CoS are present, no flow control is implemented in the Ethernet network, and the risk of congestions is minimized by over- provisioning the bandwidth of Ethernet links and Ethernet nodes.

WO 2009/046869 A1 discloses a technique for transmitting data in an Ethernet network including nodes that support differentiated queues for different Classes of Service and queue congestion handling, which involves the steps of recognizing at one node in the network that a given queue for a given Class of Service is congested, and causing the node to send to another node in the network pause frames that explicitly identify that queue for a given Class of Service as congested. The node receiving the pause frames may then control the transmission rates for different outbound queues towards the first node according to the congestion information carried by the pause frames.

Summary of the invention

The inventors have noticed that over- provisioning of bandwidth in Ethernet links and Ethernet nodes for minimizing Hie risk of congestions when different CoS are present is disadvantageous^' expensive and implies an unacceptable waste of resources.

On the other hand, dropping of packets for managing queue congestion based on CoS disadvantageous^ requires substantial buffering capabilities at each Ethernet node.

Besides, as mentioned above, the above described Ethernet flow control mechanism is not CoS aware. This means that, in case of congestion, the flow control mechanism stops all packets, regardless of the CoS they belong to. Therefore, disadvantageous, both low priority and high priority traffic experience delay and jitter. Typically, congestion of input queues is mainly due to bursts of low priority traffic, such as best effort, which generally has a not constant bit-rate. On the contrary, typically, higher priority traffic has a known and constant bit-rate but, due to congestions, may experience a non controllable latency, resulting in uncontrolled and excess delay and jitter.

Accordingly, the inventors have addressed the problem of providing a network node for a communication network (in particular, but not exclusively, an Ethernet communication network) capable of controlling traffic flows with different CoS, which overcomes the aforesaid drawbacks .

In particular, the inventors have addressed the problem of providing a network node for a communication network (in particular, but not exclusively, an Ethernet communication network) capable of controlling traffic flows with different CoS, which exploits the above described Ethernet flow control mechanism and, at the same time, properly handles the different QoS, allowing controlling excess deiay and jitter on higher priority traffic.

According to a first aspect, the present invention provides a network node for a packet-switched communication network comprising: - a first queue and a second queue;

- an input module configured to receive from a further network node first packets associated to a first CoS and second packets associated to a second CoS, the first CoS having QoS requirements more stringent than the second CoS. and to forward the first packets to the first queue and the second packets to the second queue;

- a queue manager configured to detect a congestion of the second queue; and

- a transmitting module configured to transmit to the further network node a pause frame specifying a stop duration for which the further network node should stop transmitting the first packets and the second packets, wherein the stop duration is lower than or equal to a maximum allowable delay variation on the first packets.

Preferably, the queue manager is configured to detect the congestion of the second queue by comparing a filling level of the second queue with a filling threshold.

Preferably, the transmitting module is further configured to periodically retransmit the pause frame with a re-transmission period.

Preferably, the re-transmission period is such that an effective transmission bit-rate at which the first packets and second packets are received at the network node is lower than or equal to a bit-rate available downstream of the network node.

Preferably, the re-transmission period is calculated as follows:

wherein SD is the stop duration, CR is a maximum transmission rate at which the first packets and second packets are received at the network node and RR is the bit-rate available downstream of the network node.

Preferably, the re-transmission period is such that the effective transmission bit-rate is higher than or equal to a constant bit-rate of the first packets. According to a second aspect, the present invention provides a packet- switched communication network comprising a network node as set forth above.

According to a third aspect, the present invention provides a method for controlling traffic flows with different CoS in a communication network, the traffic flows comprising first packets associated to a first CoS and second packets associated to a second CoS, the first CoS having QoS requirements more stringent than the second CoS, Ihe method comprising, at a network node:

a) providing a first queue and a second queue;

b) receiving the first packets and the second packets from a further network node, storing the first packets in the first queue and storing the second packets in the second queue;

c) detecting a congestion of the second queue;

d) transmitting to the further network node a pause frame specifying a stop duration for which the further network node should stop transmitting the first packets and the second packets,

wherein step d) comprises setting the stop duration to a value lower than or equal to a maximum allowable delay variation on the first packets.

Preferably, step c) comprises detecting the congestion of the second queue by comparing a filling level of the second queue with a filling threshold.

Preferably, step d) is periodically repeated with a re-transmission period. Preferably, step d) comprises setting the re-transmission period so that an effective transmission bit-rate at which the first packets and second packets are received at the network node is lower than or equal to a bit-rate available downstream of the network node.

Preferably, step d) comprises calculating the re-transmission period as follows: wherein SD is the stop duration. CR is a maximum transmission rate at which the first packets and second packets are received at the network node and RR is the bit-rate available downstream of the network node.

Preferably, step d) comprises setting the re-transmission period so that the effective transmission hit-rate is higher than or equal to a constant bit-rate of the first packets.

According to a fourth aspect, the present invention provides a computer program product comprising computer-executable instructions for performing, when the program is run on a computer of a network node for a communication network, a method for controlling traffic flows with different CoS in a communication network, the traffic flows comprising first packets associated to a first CoS and second packets associated to a second CoS, the first CoS having QoS requirements more stringent than the second CoS, the method comprising:

- when the first packets and the second packets are received at the network node from a further network node, storing the first packets in a first queue and storing the second packets in a second queue;

- detecting a congestion of the second queue;

- generating a pause frame specifying a stop duration for which the further network node should stop transmitting the first packets and the second packets,

wherein the step of generating comprises setting the stop duration to a value lower than or equal to a maximum allowable delay variation on the first packets.

According to a fifth aspect the present invention provides a digital data storage medium storing a computer program product as set forth above.

Brief description of the drawings

Embodiments of the invention will be better understood by reading the following detailed description, given by way of example and not of limitation, to be read with reference to the accompanying drawings, wherein: - Figures 1a and 1b schematically show a portion of a communication network according to an embodiment of the present invention;

- Figure 2 shows an exemplary flow chart illustrating the steps of the method as a function of time according to an embodiment of the present invention; and

- Figure 3 shows an exemplary time diagram of the number of transmitted bits in the communication network of Figures 1 a and 1b when the method of Figure 2 is implemented.

Detailed description of preferred embodiments of the invention

Figures 1a and 1b schematically show a portion of a communication network CN suitable for implementing a method for controlling traffic flows according to an embodiment of the present invention. Preferably, the communication network CN is a packet-switched communication network, more preferably an Ethernet communication network.

The portion of communication network CN shown in Figure 1a comprises a transmitting node TN, a receiving node RN and a link EL connecting the transmitting node TN and the receiving node RN, The communication network CN may comprise other nodes and links that are not shown in Figures 1a and 1b, since they are not relevant to the present description.

Preferably, the link EL is a packet-switched link, more preferably an Ethernet link. Although in the following description the nodes TN and RN arc termed "transmitting" and "receiving" respectively, the link EL is preferably a bidirectional link, i.e. both the transmitting node TN and the receiving node RN have a transmitting side and a receiving side. However, for sake of clarity of the following description, only the transmitting side of TN and the receiving side of RN are considered.

The link EL is suitable for carrying packets EP distributed in a number of traffic flows transmitted from the transmitting node TN to the receiving node RN. The packets EP are preferably Ethernet packets. Each traffic flow (and therefore each packet EP belonging thereto) is preferably associated to a Class of Service (CoS). Preferably, a number N of CoS is defined. For instance, N may be equal to three, the three CoS being Best Effort, Guaranteed Bandwidth and Fixed Bandwidth.

Packets EP are preferably marked with a priority corresponding to the associated CoS. For instance, packets associated to the Best Effort CoS are marked with a low priority, packets associated to the Guaranteed Bandwidth CoS are marked with a medium priority and packets associated to the Fixed Bandwidth CoS are marked with a high priority.

According to a preferred embodiment of the invention, the transmitting node

TN comprises a number N of output queues Q° _n (n=1 , 2 N) and an output module O .

The output queues Q° are suitable for buffering the packets EP that have to be transmitted from the transmitting node TN to the receiving node RN. Preferably, each output queue Q° _n is configured to buffer packets EP associated to a respective CoS. Since, as mentioned above, it has been assumed that N is equal to three, the transmitting node TN comprises three output queues Q Q°2 and Q 3. each corresponding to a respective CoS. For instance, the output queue Q°i may be associated to the Fixed Bandwidth CoS, the output queue Q°2 may be associated to the Guaranteed Bandwidth CoS and the output queue Q° ₃ may be associated to the Best Effort CoS. This is merely exemplary, since the transmitting node TN can comprise any number of output queues, according to the number N of defined CoS.

Preferably, the output module OM is suitable for scheduling the transmission of the packets EP over the link EL.

According to a preferred embodiment of the present invention, the receiving node RN comprises an input module I , a number of input queues Q' _n (n=1 ,2, .... N) and a queue manager Q .

Preferably, the input module IM is configured to receive packets EP from the transmitting node TN and, for each packet, to identify the associated CoS. On the basis of the identified CoS. the input module IM is configured for forwarding each packet to a respective input queue Q' _n .

The input queues Q' _n are preferably suitable for buffering the packets EP received at the receiving node RN from the transmitting node TN. Preferably. each input queue Q' _n is configured to buffer packets associated to a respective CoS. Since, as mentioned above, it has been assumed that N is equal to three, the receiving node RN comprises three input queues Q\, Q' ₂ and Q' ₃ , each corresponding to a respective CoS. For instance, the input queue θΊ may be associated to the Fixed 8andwidth CoS, the input queue Q'a may be associated to the Guaranteed Bandwidth CoS and the input queue Q' ₃ may be associated to the Best Effort CoS. Obviously, this is merely exemplary, since the receiving node RN can comprise any number of input queues, according to the number !M of defined CoS.

Preferably, a filling threshold T _n (n=1 ,2, .... N) is associated to each input queue Q'n. Each filling threshold T _n identifies the input queue filling level above which the input queue Q' _n is considered to be "congested". Preferably, different filling thresholds T _n are associated to different input queues Q'n. Preferably, the filling thresholds T _n of the input queues Q' _n are chosen on the basis of the requirements (i.e. delay and jitter) of the associated CoS. More preferably, the filling threshold T„ increases as the CoS associated to the input queue Q'n decreases (i.e. its requirements become less stringent).

The queue manager QM is preferably configured for monitoring the filling level of each input queue Q' _n , for comparing it with the respective filling threshold T ₀ , and for detecting possible congestions of the input queue Q' _n when the filling level exceeds the filling threshold T _n .

In Figure 1a, where the receiving node RN comprises three input queues Q'L Q' ₂ and Q' ₃ , the respective filling thresholds Ti, T ₂ and T ₃ are schematically indicated. The input queue Q i, being associated to the Fixed Bandwidth CoS, has the lowest filling threshold T-,, whereas the input queue Q 3, being associated to the Best Effort CoS, has the highest filling threshold T ₃ . Indeed, the traffic flows associated lo the Best Effort CoS do not have a constant bandwidth, and accordingly the packets EP included in such traffic flows may be grouped in "packet bursts" that may induce fast and non- predictable filling of the input queue Q' _:J .

The method according to a preferred embodiment of the present invention will be now described in detail, referring also to the flow chart of Figure 2. In the operative situation shown in Figure 1a, the transmitting node TN is transmitting packets EP to the receiving node RN through the link EL (step 20).

it is assumed that, at time to. the queue manager Q detects that one of the input queues Q' _n becomes congested (step 21 ). In particular, it is assumed that the filling level of input queue Q' ₃ , corresponding to the Best Effort CoS, exceeds the filling threshold T ₃ and thus the input queue Q'3 is congested. This is merely exemplary, since congestion can occur to each of the input queues Q'i . Q'2 and O 3.

According to a preferred embodiment of the invention, at time to (i.e. upon detection by the queue manager QM that the input queue Cr ₃ is congested), the receiving node RN issues a pause frame PF and sends it (by means of a transmitting module not shown in Figures 1a and 1b) to the transmitting node TN. Preferably, the pause frame PF is an Ethernet pause frame containing a parameter field that specifies a stop duration SD.

The transmitting node TN preferably receives the pause frame PF by means of a receiving module (not shown in Figures 1a and 1 b). After a delay, due to the propagation of the pause frame PF along the link EL and the processing of the pause frame PF at the transmitting node TN, the transmuting node TN preferably stops the transmission of ail packets EP directed to the receiving node RN (step 22). Preferably, the transmission is stopped for a time interval having a duration equal to the stop duration SD specified in the received pause frame PF. After the stop duration SD is elapsed, the transmitting node TN preferably restarts the transmission of all packets EP directed to the receiving node RN.

Preferably, the stop duration SD is set lower than or equal to a maximum allowable delay variation on the packets EP associated to the CoS with more stringent QoS requirements (i.e. Fixed Bandwidth, in the above example). Preferably, the maximum allowable delay variation on the packets EP belonging to the traffic flow associated to the CoS with more stringent QoS requirements is related to the QoS requirements (i.e. delay and jitter) of the traffic flow. In particular, for CES traffic flows, the maximum allowable delay variation is chosen so thai it can be absorbed by the buffering capability of the inter-working functions interposed between the TDM network from which the TDM traffic to be transported according to the CES service comes and the communication network CN {in particular, the transmitting node TN) of Figures 1a and 1b. For example, a typical value of the maximum allowable delay variation for CES traffic flows is equal to about 1 ms.

Advantageously, selecting a stop duration SD lower than or equal to a maximum allowable delay variation on the packets EP associated to the CoS with more stringent OoS requirements allows controlling the delay variation induced by the pause frame mechanism on the traffic flows with CoS with more stringent QoS requirements.

According to a preferred embodiment of the present invention, the pause frame PF specifying the stop duration SD is sent at time to (as described above) and successively, it is periodically re-transmitted by the receiving node RN to the transmitting node TN. In Figure 2, the re-transmission period between two consecutive transmissions of the pause frame PF by the receiving node RN is referred to as PFP. The re-transmission period PFP is preferably longer than the slop duration SD.

Therefore, in each re-transmission period PFP the transmitting node TN stops transmission of all packets EP for a stop duration SD. For simplicity, a single re-transmission is shown in Figure 2 and is indicated as step 24. Similarly, the step of stopping transmission at reception of a pause frame PF following the first re-transmission step 24 is indicated in Figure 2 as step 25, while the step of restarting transmission after a stop duration SD is elapsed is indicated in Figure 2 as step 26. Steps 24, 25 and 26 are periodically repeated in each re-transmission period PFP. For simplicity, such repetitions are not shown in Figure 2.

Preferably, the re-transmission period PFP depends on the stop duration SD and on the difference between the transmission bit-rate CR {expressed e.g. in Mb/s) of the link EL and the bit-rate RR {expressed e.g. in Mb/s) available downstream of the receiving node RN. In particular, the re-transmission period PFP is computed so as to provide, on the link EL, an effective transmission bit-rate lower than or equal to the bit- rate RR available downstream of the receiving node RN. Besides, according to a preferred embodiment of the present invention, the re-transmission period PFP is computed so as to provide, on the link EL, an effective transmission bit-rate higher than or equal to the constant bit-rate associated to the CoS with more stringent QoS requirements, i.e. the Fixed Bandwidth CoS. This advantageously allows avoiding congestion of the input queues at the receiving node RN. while at the same time it guarantees that the traffic flows associated to the Fixed Bandwidth CoS are properly delivered.

With reference to Figure 3, the computation of the re-transmission period PFP Is described in further detail, by considering an exemplary case.

In particular, it is assumed that a CES traffic flow is transmitted on the link EL from the transmitting node TN to the receiving node RN. The CES traffic flow is associated to the Fixed Bandwidth CoS. The maximum allowable deiay variation on the packets EP belonging to the CES traffic flow is assumed to be equal to 1 ms. Besides, it is assumed that the transmission bit-rate CR of the link EL is equal to 100 Mb/s, and the bit-rate RR available downstream of the receiving node RN is equal to 20 Mb/s.

In Figure 3, the operation of the transmitting node TN for a fixed time window TW has been taken into consideration. The time window TW is equal to 10 ms.

As shown in Figure 3, the maximum number of bits that, in a time window TW of 10 ms, can be processed by the receiving node RN is equal to RR*TW=20 Mb/s*10 ms=200 Kb. On the other hand, since the transmission bit-rate CR of the link EL is equal to 100 Mb/s, in a time window TW of 10 ms the link EL can carry CR*TW=100 Mb/s*10 ms=1 Mb of traffic.

As mentioned above, for avoiding congestions of the input queues at the receiving node RN and for guaranteeing proper delivery of the traffic flows associated to the CoS with more stringent QoS requirements, the effective transmission bit-rate of the link EL should be set (by suitably tailoring the retransmission period PFP) to a value lower than or equal to the bit-rate RR, and higher than or equal to the constant bit-rate of the Fixed Bandwidth traffic flow. In this example, it is assumed that the re-transmission period PFP is set in order to provide an effective transmission bit-rate equal to the bit-rate RR available downstream of the receiving node RN. To this purpose, the total number of bits to be stopped in the time window TW of 10 ms is (CR- RR)*TW=(100 b/s-20 Mb/s)/10 ms=800 Kb.

Since, as described above, each stop period preferably has a stop duration SD equal to or lower than the maximum allowable delay variation on the packets EP associated to the CoS with more stringent QoS requirements (i.e. 1 ms for packets of the CES traffic flows, in this example), the 800 Kb should be stopped using stop periods with a stop duration SD lower than or equal to 1 ms each.

According to the example considered in Figure 3, It is assumed that the stop duration SD is equal to 1 ms. Therefore, in each stop period, CR*SD=100 Mbit/s*1 ms=100 Kb are stopped. In order to stop 800 Kb in a time window TW of 10 ms, the transmitting node TN should therefore receive eight pause frames PF inside the time window TW of 10 ms.

Since, as mentioned above, the pause frames PF are periodically transmitted with a re-transmission period PFP, according to the example considered in Figure 3, the re-transmission period PFP is thus equal to:

PFP=TW/8= 10 ms/8=1 ,25 ms.

Therefore, in each time window TW of 10 ms, transmission of the packets EP is periodically stopped and restarted, with a period of 1 ,25 ms. Every 1 ,25 ms, the transmitting node TN transmits packets EP at a maximum rate of CR=100 bit/s for 0,25 ms, while it stops transmission for 1 ms, as indicated by the curve B of Figure 3.

More generally, therefore, if (CR-RR)*TW is the number of bits that should be stopped in a time window TW and CR*SD is the number of bits stopped during a stop period having stop duration SD, the number of stop periods that should be provided in a time window TW is: The re-transmission period PFP is therefore equal to:

In this way, advantageously, the effective transmission bit-rate on the link EL is lowered from the transmission rate CR to the bit-rate RR available downstream of the receiving node RN (thereby preventing congestion of the input queues at the receiving node RN) while, at the same time, the maximum delay variation experienced by all the packets EP (and, therefore, also by the packets of the CES traffic flow) is equal to the maximum delay variation of 1 ms.

It can be noticed that stopping 800 Kb in each time window TW of 10 ms may be achieved for instance by operating the transmitting node TN so that, in each time window TW of 10 ms, it transmits packets EP for 2 ms and then it stops transmission for 8 ms, as indicated by the curve A of Figure 3. This means that packets buffered in the output queues Q\, Q'z and Q' ₃ at the transmitting node TN may undergo a latency equal to 8 ms. This value is however much higher than 1 ms, i.e. the maximum allowable delay variation on the packets EP belonging to the CES traffic flow.

Therefore, periodically sending pause frames PF to the transmitting node TN in order to periodically stop transmission as disclosed above advantageously prevents congestion of the input queues Q' _n at the receiving node RN, or at least of the input queue associated to the CoS with more stringent QoS requirements. This is achieved by suitably setting the stop duration SO and the re-transmission period PFP so that the transmission bit- rate CR of the link EL matches the bit-rate RR available downstream of the receiving node RN, or at least the bit-rate associated to the services with highest CoS. In this way, advantageously, the transmission of the services with highest CoS (e.g. Fixed Bandwidth) is guaranteed without loss of packets.

On the other hand, the QoS requirements of the related services (e.g. the CFS services) in terms of delay and delay variation are fuifilied, by setting the stop duration SD to a value lower than or equal to the maximum delay variation that packets transporting CES services may undergo. This advantageously allows to control the delay and jitter of the delivered services and thus, in other words, to properly handle the QoS of the services themselves.

Moreover, advantageously, the output queues Q° _n of the transmitting node TN are periodically emptied and thus, differently from the congestion management mechanisms, no substantial buffering capabilities are required at the transmitting node TN.