Packet loss measurement in packet-switched communication networks

Abstract

A method for exchanging packets between two nodes of a packet switching communication network is disclosed. Each packet includes a tag field and a reflected tag field. Each node sets the value of the tag field in its outgoing packet for transmission to another node. This value alternates between two alternating tag values ​​every N outgoing packets. While receiving an incoming packet from another node, each node also sets the value of the reflected tag field of its outgoing packet according to the value of the tag field of the incoming packet. An observer placed between the two nodes can count packets with a tag field equal to any tag value and / or packets with a reflected tag field equal to any tag value, and use N and such counts to provide a packet loss measurement.

Description

Technical Field

[0001] The present invention relates to the field of communication networks. In particular, the present invention relates to a method for switching packets in a packet-switched communication network, a method for performing packet loss measurements and a packet-switched network configured to implement such a method. Background Art

[0002] In a packet-switched communication network, a packet stream is transmitted from a source node to a destination node through possible intermediate nodes. Examples of packet-switched networks are IP (Internet Protocol) networks, Ethernet and MPLS (Multi-Protocol Label Switching) networks.

[0003] Packets do not always reach their destination node, i.e., they may be lost during transmission through the network. Packets are lost for different reasons. For example, a node or a link may fail, or packets may be dropped by a node due to congestion at its port. Furthermore, packets may be dropped by a node because they contain bit errors.

[0004] When communication services (particularly real-time voice or data services such as calls, conference calls, video conferences, etc.) are provided by means of a packet-switched network, performance measurements on packet loss of packet flows carrying the service provide an indication of the quality of service (QoS) perceived by end users of the service. Furthermore, packet loss may require retransmissions, thus reducing the efficiency of the communication network. Measuring packet loss of packet flows in a communication network is therefore of particular interest to network operators.

[0005] As is known, QUIC (Quick UDP Internet Connections) is a transport layer (layer 4) network protocol designed to support multiplexed connections between two endpoints (client and server) over the User Datagram Protocol (UDP). In order to enable endpoints to detect possible packet losses, the QUIC protocol provides an acknowledgment for a transmission each time an endpoint receives a packet from the other endpoint. When a packet is transmitted and no acknowledgment is received within a predefined time threshold, the transmitted packet is declared lost.

[0006] Techniques are also proposed that enable packet loss measurements to be made by an intermediate observer located between the two endpoints.

[0007] M. Cociglio et al.: Internet Draft "New Spin bit enabled measurements with one or two more bits draft-cfb-ippm-spinbit-new-measurements-01", July 1, 2019, describes a round-trip packet loss measurement whereby a client generates a series of marked packets having a dedicated header bit (also called a loss bit) set to 1. The server reflects the series of marked packets, which are then reflected again by the client and then by the server. Thus, the series of marked packets travel back and forth twice between the client and the server. An observer located between the client and the server detects the series of marked packets transmitted in the same direction twice and calculates the round-trip packet loss based on the number of marked packets that the series of marked packets passes each time.

[0008] A. Ferrieux et al.: Internet Draft "Packet Loss Signaling for EncryptedProtocols draft-ferrieuxhamchaoui-quic-lossbits-03", January 16, 2020, describes an extension to the QUIC protocol that introduces two bits in the packet header, the Q bit (square signaling bit) and the L bit (loss event bit). Each endpoint sets the value of the Q bit in each outgoing packet. The value of the Q bit is toggled every N outgoing packets (N is typically 64). Each endpoint also implements an unreported loss counter that is incremented by 1 each time a packet loss is declared using QUIC's existing loss detection mechanism. If the counter is positive, then the L bit of the outgoing packet is set to 1, otherwise it is set to 0. The counter is decremented each time a packet with the L bit equal to 1 is transmitted. An observer located between the two endpoints counts packets with the Q bit equal to 1 or 0 and / or the L bit equal to 1 or 0 and provides different types of one-way packet loss measurements based on these numbers, including end-to-end measurements (i.e., packet loss from the transmitting endpoint to the receiving endpoint), upstream measurements (i.e., packet loss from the transmitting endpoint to the observer), and downstream measurements (i.e., packet loss from the observer to the receiving endpoint). Summary of the invention

[0009] Applicants have recognized a need to provide an improved technique to enable an intermediate observer to provide packet loss measurements, including unidirectional measurements, such as end-to-end measurements, upstream measurements, and downstream measurements.

[0010] According to the technique described by A. Ferrieux et al., indeed, the value of the L bit that each endpoint sets in an outgoing packet depends on the QUIC mechanism whereby the endpoint itself declares that a transmitted packet is lost, which in turn depends on whether the endpoint receives an acknowledgment from the other endpoint within a predefined time threshold.

[0011] However, the failure to receive an acknowledgment does not necessarily mean that the transmitted packet was actually lost. For example, the acknowledgment itself may be lost, or, due to congestion, it may be received at the endpoint after a time threshold has expired. In all of these cases, the endpoint that transmitted the packet declares the packet lost, increments the unreported loss counter, and then sets the L bit of the outgoing packet to 1. The number of outgoing packets with the L bit equal to 1 may not indicate the number of transmitted packets that were actually lost. Therefore, packet loss measurements based on counts of packets with the L bit equal to 1 or 0 detected by intermediate observers may not be accurate.

[0012] Furthermore, if the endpoint and the observer are managed by different entities, then the entity managing the observer has essentially no control over the mechanism by which the endpoint declares packet loss. If this mechanism is removed or modified by the entity managing the endpoint, then the mechanism for setting the L bit may no longer work, and all types of packet loss measurements based on the L bit will become infeasible.

[0013] In view of the above, the Applicant has solved the problem of providing a method for exchanging packets and a method for performing packet loss measurements on packets exchanged between two nodes of a packet switched communication network, thereby overcoming the above mentioned drawbacks.

[0014] In particular, the Applicant has solved the problem of providing a method for exchanging packets and performing packet loss measurements on packets exchanged between two nodes of a packet-switched communication network (for example - but not exclusively - two endpoints of a QUIC-based connection), which allows an observer placed between the two nodes to provide accurate packet loss measurements, including unidirectional measurements, such as end-to-end measurements, upstream measurements and downstream measurements.

[0015] According to an embodiment of the present invention, this problem is solved by a method in which a packet exchanged by two nodes includes a first field (also referred to herein as a "tag field") and a second field (also referred to herein as a "reflected tag field"). Each of the two nodes sets a value of the tag field of its outgoing packet, which alternates between a first and a second value (e.g., 1 and 0) every N outgoing packets, where N is a predefined integer. In addition, at least one of the nodes sets a value of the reflected tag field of its outgoing packet based on the value of the tag field of the incoming packet received from the other node.

[0016] This behavior of the two nodes results in the following exchange of packets.

[0017] The first node (e.g., a client of a QUIC-based connection) switches the value of the marker field in the first node's outgoing packets to be transmitted to the second node (e.g., a server of a QUIC-based connection) every, for example, N=64 packets. Thus, the stream of outgoing packets transmitted by the first node is in the form of a sequence of consecutive blocks of N=64 packets, the sequence comprising blocks of packets with the marker field equal to 1, which blocks of packets are temporally alternating with blocks of packets with the marker field equal to 0.

[0018] Since the second node behaves in the same way, the first node also receives from the second node a sequence of consecutive blocks of incoming packets, the sequence comprising blocks of packets with the marker field equal to 1, alternating in time with blocks of packets with the marker field equal to 0. The number of incoming packets in that block received by the first node is equal to N if there are no packet losses for a block, and is less than N otherwise.

[0019] When a first node receives such a flow of incoming packets, it sets the value of the reflected Tag field of its outgoing packets according to the value of the Tag field of the incoming packets received from the other node. For example, if the packet rates in both directions are substantially the same, then the first node may copy the value of the Tag field of each incoming packet into the reflected Tag field of the corresponding outgoing packet.

[0020] An observer placed between the two nodes can then detect and count packets transmitted from the first node to the second node whose tag fields and reflected tag fields are set as described above to 0 or 1. Knowing the value of N a priori, the observer can use such packet counts to provide different types of packet loss measurements, including unidirectional measurements, such as end-to-end measurements, upstream measurements, or downstream measurements.

[0021] For example, by counting the number of packets with a flag field equal to 1 or 0 (eg, by means of two separate counters), the observer may provide a measure of packet loss from the first node to the observer for each chunk.

[0022] In addition, the observer may also count the number of packets for which the reflected tag field is equal to 1 or 0 (e.g., by means of two separate counters). Since the value of the reflected tag field in these packets outgoing from the first node is related to the value of the tag field in the packets incoming at the first node from the second node, this number takes into account possible packet losses that occurred during transmission from the second node to the first node and possible packet losses that occurred during transmission from the first node to the observer. The observer may then provide a measurement of packet losses from the second node to the first node and further to the observer for each block.

[0023] The above measurements can be combined (possibly with similar measurements provided in the opposite direction if the observer is bidirectional) to provide further types of measurements, such as end-to-end measurements and / or downstream measurements. Other types of measurements can be provided if multiple observers can be implemented between two nodes.

[0024] As can be appreciated, these packet loss measurements - particularly those based on or involving reflected marker fields - are advantageously very accurate.

[0025] Since the first node sets the value of the reflected tag field in its outgoing packets according to the value of the tag field in the incoming packets from the second node as discussed above, the number of packets with the reflected tag field equal to 1 or 0 as detected by the observer indicates the number of packets transmitted by the second node that were actually lost, meaning they were not received by the first node. Other events that cause a node to declare a packet lost (such as the loss of an acknowledgment from another node or the receipt of an acknowledgment after a time threshold has passed) have no effect on the value of the reflected tag field bits in the outgoing packets.

[0026] Furthermore, the value of the reflected tag field is independent of the mechanism by which a node declares packet loss, or any other internal mechanism of the node provided for purposes other than enabling an observer to perform measurements according to the present invention. Thus, advantageously, the entity performing packet loss measurements has greater control over the packet loss measurement implementation, since it relies only on dedicated mechanisms implemented at the node (i.e., alternating tagging and reflection) for the specific purpose of supporting such measurements.

[0027] According to a first aspect, the invention provides a method for exchanging packets between two nodes of a packet-switched communications network, each packet comprising a tag field and a reflected tag field, the method comprising:

[0028] a) each of the two nodes sets a value of the tag field in an outgoing packet to be transmitted to the other node, the value switching alternately between a first tag value and a second tag value every N outgoing packets, where N is a predefined integer;

[0029] as well as

[0030] b) at least one of the two nodes, upon receiving an incoming packet from the other node, setting the value of the reflected tag field of the outgoing packet according to the value of the tag field of the incoming packet.

[0031] The incoming packet carries a marker signal consisting of a sequence of values ​​of the marker field in the incoming packet, and the outgoing packet carries a reflected marker signal consisting of a sequence of values ​​of the reflected marker field in the outgoing packet, and step b) preferably includes transmitting at least one cycle of the reflected marker signal in response to reception of at least one cycle of the marker signal.

[0032] Preferably, step b) comprises setting the width of the period of the reflected marker signal according to the width of the period of the marker signal, the width being expressed as the number of packets per period.

[0033] According to an embodiment, at step b), the period of transmitting the reflected marking signal comprises:

[0034] b1) when initiating transmission of the period of the reflected marker signal, setting the width of the period of the reflected marker signal to be equal to the width of the last period of the marker signal whose reception has been completed; and

[0035] b2) if reception of at least one additional cycle of the reflected marker signal is completed before transmission of the cycle of the reflected marker signal is terminated, then updating the width of the cycle of the reflected marker signal based on the width of the at least one additional cycle of the marker signal.

[0036] Preferably, updating the width of the period of the marker signal reflected by the litigation lake at step b2) comprises setting the width of the period of the reflected marker signal equal to the average width of the at least one further period of the marker signal.

[0037] Preferably, step b) comprises transmitting R periods of said reflected marker signal in response to reception of one period of said marker signal, R being an integer approximating the ratio between the transmission rate of said outgoing packets and the reception rate of said incoming packets.

[0038] In this case, the width of each of the R periods of the reflected marker signal is equal to the width of the one period of the marker signal, the width being expressed as the number of packets per period.

[0039] Alternatively, step b) comprises transmitting one period of said reflected marker signal in response to receipt of R periods of said marker signal, R being an integer approximating the ratio between the reception rate of said incoming packets and the transmission rate of said outgoing packets.

[0040] In this case, the width of the one period of the reflected mark signal is equal to the average value of the widths of the R periods of the mark signal, the width being expressed as the number of groups per period.

[0041] According to a second aspect, the invention provides a method for performing packet loss measurements on packets exchanged between two nodes of a packet switched communications network, each packet comprising a marker field and a reflected marker field, the method comprising the steps of the method as set out above and:

[0042] c) by an observer device, detecting the outgoing packet, providing a first parameter based on a value of the tag field in the detected outgoing packet, providing a second parameter based on a value of the reflected tag field in the detected outgoing packet, and performing the packet loss measurement based on the number N in combination with the first parameter and / or the second parameter.

[0043] According to a third aspect, the invention provides a packet switched communications network comprising two nodes for switching packets, each packet comprising a tag field and a reflected tag field, wherein:

[0044] - each of the two nodes is configured to set a value of the tag field in an outgoing packet to be transmitted to the other node, the value alternating between a first tag value and a second tag value every N outgoing packets, N being a predefined integer; and

[0045] - At least one of the two nodes is further configured to, upon receiving an incoming packet from the other node, set the value of the reflected tag field of the outgoing packet according to the value of the tag field of the incoming packet.

[0046] Preferably, the packet-switched communication network further comprises an observer device, which is configured to detect the outgoing packet, provide a first parameter based on the value of the tag field in the detected outgoing packet, provide a second parameter based on the value of the reflected tag field in the detected outgoing packet, and perform the packet loss measurement based on the number N in combination with the first parameter and / or the second parameter. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The present invention will become more apparent from the following detailed description, given by way of example and not limitation, which is to be read with reference to the accompanying drawings, in which:

[0048] - Figure 1 schematically illustrates a packet switched communications network configured to exchange packets in order to enable round trip packet loss measurements according to an embodiment of the present invention;

[0049] - Figure 2 Schematically shows an embodiment of the present invention. Figure 1 The structure of packets exchanged in a communication network;

[0050] - Figure 3 shows an exemplary marker signal and reflected marker signals carried by a packet stream outgoing from a node operating in accordance with an embodiment of the present invention;

[0051] - Figure 4 schematically illustrates exemplary marker signals and reflected marker signals exchanged by two nodes operating in accordance with an embodiment of the present invention;

[0052] - Figure 5 shows that it can be placed in Figure 4 Exemplary measurements performed by an intermediate observer between two nodes;

[0053] - Figure 6 is a flow chart of a reflection mechanism performed by each node according to an advantageous embodiment of the present invention; and

[0054] - Figure 7(a) and 7(b) Shown by Figure 6 Flowchart of the operation of the nodes of the present invention provides exemplary incoming marking signals and outgoing marking signals. DETAILED DESCRIPTION

[0055] Detailed description of the preferred embodiment of the present invention

[0056] Figure 1 There is schematically shown a packet switched communication network 100 configured to exchange packets in order to enable packet loss measurements according to an embodiment of the invention.

[0057] The communication network 100 includes a plurality of nodes interconnected with each other by physical links according to any known topology, including Figure 1 The two nodes 1 and 2 are shown in FIG. 1 . The nodes 1 and 2 may be connected via a single physical link or via a cascade of several physical links and intermediate nodes (not shown in the figure). The communication network 100 may be, for example, an IP network.

[0058] Node 1 is configured to transmit packets Pk to node 2, and node 2 is configured to transmit packets Pk' to node 1. Packets Pk may belong to the same packet flow (i.e., they may both have the same source address and the same destination address), or may belong to different packet flows whose paths overlap between nodes 1 and 2. Similarly, packets Pk' may belong to the same packet flow or to different packet flows whose paths overlap between nodes 2 and 1.

[0059] The packets Pk, Pk' are formatted according to a specific network protocol. As non-limiting examples, the network protocol may be the QUIC protocol mentioned above or TCP (Transmission Control Protocol).

[0060] like Figure 2 As schematically depicted in FIG. 1 , each packet Pk, Pk' comprises a payload PL and a header H (or more headers belonging to different network layers, such as an IP header and a QUIC header). The payload PL comprises user data. The header H comprises packet forwarding information ( Figure 2 ), ie information that allows the nodes of the network 100 to properly handle the packets Pk, Pk' so that they reach their destination node.

[0061] According to an embodiment of the present invention, each packet Pk, Pk' preferably includes a first field (also referred to herein as a "marker field") MF and a second field (also referred to herein as a "reflected mark field") RMF. The mark field MF and the reflected mark field RMF may be part of the same header H including packet forwarding information, such as Figure 2 As shown in . According to other embodiments, the marking field MF and the reflected marking field RMF may be located in other parts of the packets Pk', Pk'. For example, if the packets Pk, Pk' include other headers in addition to the header H, then the marking field MF and the reflected marking field RMF may be included in any such additional headers. For example, assuming that the packets Pk, Pk' include an IP header H—including packet forwarding information—and a QUIC header, the marking field MF and the reflected marking field RMF may be included in the QUIC header.

[0062] Each field MF, RMF includes one or more bits. Each field MF, RMF can be set to any one of a first tag value VA (e.g., 0 in the case of a one-bit field MF, RMF) and a second tag value VB (e.g., 1 in the case of a one-bit field MF, RMF).

[0063] According to an embodiment of the present invention, each node 1, 2 preferably sets the value of the marking field MF of each outgoing packet Pk, Pk' to be transmitted to the other node 2, 1 to implement an alternating marking technique. In particular, each node 1, 2 preferably switches the value of the marking field MF between the first and second marking values ​​VA, VB every N outgoing packets Pk, Pk'. Preferably, the number N is a predefined integer. More preferably, the number N is a power of 2, such as 2 6 = 64. The number N may be fixed or may be variable. In particular, according to a variant, the nodes 1, 2 may vary the number N in a set of discrete integers, each integer of the set being a power of 2.

[0064] The alternating marking mechanisms implemented by the two nodes 1 and 2 are not necessarily synchronized with each other, i.e. each node 1, 2 autonomously decides when it should switch the marking value applicable to its outgoing packets Pk, Pk', independently of the behavior of the other nodes 2, 1 and based solely on the number of its outgoing packets.

[0065] The flow of outgoing packets Pk, Pk' transmitted by each node 1, 2 is then in the form of a sequence of consecutive blocks of N packets, the sequence comprising blocks of N packets Pk, Pk' with a marker field MF equal to VA, alternating in time with blocks of N packets Pk, Pk' with a marker field MF equal to VB.

[0066] Since both nodes 1, 2 behave in the same way, each node 1, 2 also receives from the other node 2, 1 a stream of incoming packets Pk', Pk in the form of a sequence of consecutive blocks of packets Pk', Pk, the sequence comprising blocks of packets Pk', Pk having a marker field MF equal to VA alternating in time with blocks of packets Pk', Pk having a marker field MF equal to VB. The number of incoming packets Pk', Pk in that block is equal to N if there are no packet losses for a block, and less than N otherwise.

[0067] While a node 1, 2 implements the above-described alternating marking on its outgoing packets Pk, Pk' and receives a flow of incoming packets Pk', Pk from another node 2, 1, it also preferably implements a reflection mechanism whereby it sets the value of the reflected marking field RMF of its outgoing packets Pk, Pk' according to the value of the marking field MF of the incoming packets Pk, Pk'. Assuming the same packet transmission rate in both directions, the reflection performed by each node 1, 2 may provide, for example, copying the value VA or VB of the marking field MF of each incoming packet Pk', Pk into the reflected marking field RMF of the corresponding outgoing packet Pk, Pk'.

[0068] As mentioned above, this reflection is preferably performed in parallel with the continuous alternating setting of the marking field MF performed by the nodes 1, 2 on the outgoing packets Pk, Pk'. Therefore, the values ​​of the marking field MF and the reflected marking field RMF in each outgoing packet Pk, Pk' are unrelated, the first being solely dictated by the alternating marking mechanism, while the second is instead solely dictated by the reflection mechanism.

[0069] The flow of outgoing packets Pk, Pk' transmitted by each node 1, 2 therefore carries essentially two unrelated signals, namely the marker signal MS consisting of the sequence of values ​​of its marker field MF and the reflected marker signal RMS consisting of the sequence of values ​​of its reflected marker field RMF. Both signals are in the form of square waves switching between VA and VB.

[0070] Figure 3is a diagram showing exemplary marker signals carried by a flow of outgoing packets Pk, Pk' transmitted by nodes 1, 2 and reflected marker signals.

[0071] Since nodes 1, 2 switch the value of the marker field MF every N outgoing packets, the marker signal MS is a periodic square wave, where the width of each period is equal to N, which is expressed as the number of packets per period. Therefore, the period of the marker signal MS is 2N and the duty cycle is 50%.

[0072] Instead, since the reflected marker field RMF is set according to the marker field MF of the incoming packets Pk', Pk, its characteristics depend on the packet losses experienced by such packets Pk', Pk during transmission from the other node 2, 1 to the node 1, 2. In general, the reflected marker signal RMS is not periodic and the width of each cycle (expressed as the number of packets per cycle) depends on the number of packets Pk', Pk lost in the block corresponding to that cycle as transmitted from the other node 2, 1. For example, Figure 3 The case is shown where some of the cycles have a width equal to N (meaning no packets Pk', Pk are lost in the corresponding block) and some other cycles have a width lower than N (meaning one or more packets Pk', Pk are lost in the corresponding block).

[0073] Figure 4 Exemplary marker signals transmitted by each of nodes 1 and 2 (ie, carried by the packet streams it transmits) and reflected marker signals are shown.

[0074] like Figure 4 As depicted in FIG, node 1 transmits a square wave marker signal MS1 with a period of 2N and a duty cycle of 50%. Due to packet loss during the transmission of packet Pk from node 1 to node 2, the marker signal MS1 may lose its periodicity, making the marker signal MS1' received by node 2 appear distorted. Node 2 reflects this marker signal MS1' into a reflected marker signal RMS2 with the same shape (assuming that the packet rate in both directions is equal), and then transmits it to node 1.

[0075] At the same time, node 2 also transmits a square wave marker signal MS2 with a period of 2N and a duty cycle of 50% to node 1. Since packet loss occurs during the transmission of packet Pk' from node 2 to node 1, the marker signal MS2 may lose its periodicity, making the marker signal MS2' received by node 1 appear distorted. Node 1 reflects this marker signal MS2' into a reflected marker signal RMS1 with the same shape (assuming that the packet rates in both directions are equal), and then transmits it to node 2.

[0076] The observer 10 placed between nodes 1 and 2 can then detect the flow of packets Pk transmitted from node 1 to node 2 and / or the flow of packets Pk' transmitted from node 2 to node 1, and provide different types of packet loss measurements based on the marker signal MS carried by such flows and the reflected marker signal RMS, including unidirectional measurements such as end-to-end measurements, upstream measurements or downstream measurements.

[0077] Figure 5 Three exemplary cases are shown, in which three different types of viewers are used.

[0078] Figure 5 (a) shows the case where the observer 10 is able to detect only the flow of packets Pk transmitted from node 1 to node 2, and then detect only the marker signal MS1 and the reflected marker signal RMS1. Thus, the observer 10 can provide a packet loss measurement based only on such signals.

[0079] For example, the observer 10 may implement two pairs of counters, namely a first pair of counters CA for counting the number of packets Pk having the marking field MF equal to VA and VB, respectively. M CB M and a second pair of counters CA for counting the number of packets Pk for which the reflected tag field RMF is equal to VA and VB, respectively RM CB RM .

[0080] By knowing the value of N a priori and using a first pair of counters CA M CB M , the observer 10 may provide a packet loss measurement PL(1→10) from the transmission node 1 to the observer 10, as described, for example, in WO 2010 / 072251 in the name of the same applicant.

[0081] In addition, by knowing the value of N a priori and using a second pair of counters CA RM CB RM , the observer 10 can provide a packet loss measurement PL(2→1→10) from node 2 to node 1 and further to the observer 10. Since the reflected marker field RMF in the outgoing packet Pk is obtained by reflecting the marker field MF in the incoming packet Pk' from node 2 at node 1, the counter CA RM CB RM Consideration is in fact given to possible packet losses occurring during transmission from node 2 to node 1 regarding packet Pk' and possible packet losses occurring during transmission of packet Pk from node 1 to observer 10. Also in this case, measurements may be performed according to WO 2010 / 072251 in the name of the same applicant.

[0082] Figure 5 (b) shows the case where the observer 10 is able to detect only the flow of packets Pk' transmitted from node 2 to node 1, and then detect only marker signal MS2 and reflected marker signal RMS2. Thus, the observer 10 can provide a packet loss measurement based only on such signals.

[0083] The operation may be the same as described above, i.e. the observer 10 may implement two pairs of counters (a first pair of counters counting the number of packets Pk' whose marker field MF is equal to VA and VB, respectively; and a second pair of counters counting the number Pk' of packets whose reflected marker field RMF is equal to VA and VB, respectively), and use the values ​​of such counters, for example, as described in WO 2010 / 072251 in the name of the same applicant, to provide a packet loss measurement PL(2→10) from the transmission node 2 to the observer 10 and / or a packet loss measurement PL(1→2→10) from the node 1 to the node 2 and further to the observer 10.

[0084] Figure 5 (c) shows a case where the observer 10 can detect both the flow of the packet Pk and the flow of the packet Pk'. The observer 10 can accordingly provide the above combination Figure 5 All packet loss measurements described in (a) and 5(b). Such measurements can be combined to provide further packet loss measurements.

[0085] For example, a half round trip measurement on both sides of the observer 10 (i.e., from the observer 10 to the node 1 or 2 and back to the observer 10) can be calculated by the observer 10 as:

[0086] -PL(10→1→10)=PL(2→1→10)-PL(2→10); and

[0087] -PL(10→2→10)=PL(1→2→10)-PL(1→10).

[0088] Furthermore, the end-to-end measurement (i.e., from node 1 to node 2 and vice versa) can be calculated by the observer 10 as:

[0089] -PL(2→1)=PL(2→1→10)-PL(1→10); and

[0090] -PL(1→2)=PL(1→2→10)-PL(2→10).

[0091] Furthermore, the downstream measurement (ie, from the observer 10 to the receiving node 1 or 2) may be calculated by the observer 10 as:

[0092] -PL(10→2)=PL(1→2)-PL(1→10); and

[0093] -PL(10→1)=PL(2→1)-PL(2→10).

[0094] It can be appreciated that the observer 10 can be implemented at either node 1 or node 2. In this case, the above packet loss measurement based only on the marker field MF is an end-to-end measurement, while the packet loss measurement based only on the reflected marker field RMF is a round-trip measurement.

[0095] In any case, it will be appreciated that these packet loss measurements - particularly those based on or involving the value of the reflected marker field RMF in packets Pk and / or Pk' - are advantageously very accurate.

[0096] Since the value of the reflected marker field RMF in the outgoing packets Pk, Pk' at each node 1, 2 is set according to the value of the marker field MF in the incoming packets Pk', Pk from the other node 2, 1 as discussed above, the number of packets Pk, Pk' detected by the observer 10 with the reflected marker field RMF equal to VA or VB indicates the number of packets Pk', Pk that were actually lost as transmitted by the other node 2, 1. Other events that cause nodes 1 and 2 to declare packet loss (such as the loss of an acknowledgment from another node or the receipt of an acknowledgment after a time threshold) have no effect on the value of the reflected marker field RMF in the outgoing packets Pk, Pk'.

[0097] Furthermore, the value of the reflected marking field RMF is independent of any mechanism by which nodes 1 and 2 declare that packets Pk or Pk' are lost (e.g., the mechanism provided by the QUIC protocol described above) or any other internal mechanism of nodes 1 and 2 provided for purposes other than enabling the observer 10 to perform the measurements described above. Thus, advantageously, the entity performing the packet loss measurement has greater control over the measurement implementation, since it relies only on dedicated mechanisms implemented at nodes 1, 2 (i.e., alternating marking and reflection) for the specific purpose of supporting such measurements.

[0098] In the above description it is assumed that the packet transmission rate in both directions is substantially the same. This allows implementing a reflection mechanism, for example by copying the value of the marker field MF of each incoming packet Pk', Pk into the reflected marker field RMF of the corresponding outgoing packet Pk', Pk'.

[0099] However, in general, the packet transmission rates in both directions can be very different from each other. For example, on a QUIC-based connection, the packet transmission rate of the server can be much higher than the packet transmission rate of the client (for example, from 2:1 to 10:1). In this case, the reflection mechanism at both nodes becomes crucial. In fact, at the node with the lower packet transmission rate, the incoming packets to be reflected will accumulate indefinitely. In addition to this, at the node with the higher packet transmission rate, the reflection mechanism will be discontinuous. In any case, the reflection mechanism will not operate properly and packet loss measurements will not be performed.

[0100] The applicant has realized that Figure 6 The flowchart shown in implements a reflection step at each node 1, 2 to solve the above disadvantages.

[0101] like Figure 6 As depicted in , when packet loss measurement starts, nodes 1, 2 initially set the value of the reflected marker field RMF in outgoing packets Pk, Pk' to an initial value (eg, VA) (step 600).

[0102] When the measurement session starts, both nodes 1 and 2 start setting the marking field MF of their outgoing packets Pk, Pk' according to the alternating marking technique, as described above. Thus, each node 1, 2 starts receiving a flow of incoming packets Pk', Pk' from the other node, which carry a marking signal MS'.

[0103] When node 1, 2 completes reception of the first cycle of the marking signal MS' (step 601), it preferably sets the average cycle width AV (the width is expressed as the number of packets per cycle) to be equal to the width of the last cycle of the marking signal MS' whose reception has been completed (the width is expressed as the number of packets per cycle) (i.e., at the first iteration, the width N0 of the first cycle) (step 602).

[0104] With the first change of value in the marker field MF of an incoming packet Pk', Pk after starting the measurement session, the nodes 1, 2 may determine completion of reception of the first period of the marker signal MS' at step 601. Further details on the implementation of step 601 will be provided below.

[0105] The nodes 1, 2 then switch the marking value of the reflected marking field RMF applicable to the outgoing packets Pk, Pk' that have to be transmitted from that moment on (step 603).

[0106] Then, for each outgoing packet Pk, Pk' to be transmitted (step 604), the node 1, 2 preferably sets its reflected marker field RMF to the currently applicable marker value determined at step 603 (step 605). As long as the number of outgoing packets Pk, Pk' for which the reflected marker field RMF has been set equal to the applicable marker value since the last switching step 603 is below the average cycle width AV (step 606), the node 1, 2 preferably continues to use the currently applicable marker value, thereby transmitting the period of the reflected marker signal RMS.

[0107] Instead, when the node 1, 2 determines that such a number has become equal to the average cycle width AV, the node 1, 2 preferably terminates the current cycle of the reflected marker signal RMS and starts a new cycle. Starting a new cycle of the reflected marker signal RMS includes returning to steps 602 and 603, i.e., setting the value of the average cycle width AV equal to the width of the last cycle of the marker signal MS' whose reception has been completed and switching the applicable marker value again. The node 1, 2 then repeats the cycle 604-605-606 to transmit a new cycle of the reflected marker signal RMS.

[0108] The above steps are repeated cyclically until the measurement session ends (step 609).

[0109] While performing the above steps, the node 1, 2 continues to receive the marking signal MS' from the other node 2, 1. During each iteration of the loop 604-605-606, the node 1, 2 preferably updates the value of the average cycle width AV used at the check 606 respectively.

[0110] In particular, as described above, nodes 1, 2 initially set the value of AV equal to the width of the last cycle of the marker signal MS' for which reception was completed when initiating a new cycle of the reflected marker signal RMS at each iteration of step 602. If nodes 1, 2 complete reception of an additional cycle of the marker signal MS' during corresponding iterations of loops 604-605-606 (step 607), nodes 1, 2 preferably update the value of the average cycle width AV based on the width of the additional cycle (step 608).

[0111] As the value in the marker field MF of the incoming packet Pk', Pk changes, the nodes 1, 2 may determine the completion of the reception of the period of the marker signal MS' at step 607. Further details on the implementation of step 607 will be provided below.

[0112] Step 608 preferably comprises updating the value of the average cycle width AV by averaging the widths of further received cycles with its initial value set at step 602. This averaging is preferably repeated recursively each time a further cycle of the marker signal MS' is received before the end of an iteration of the loop 604-605-606.

[0113] Thus, if at step 602 the average cycle width AV is initially set to the width N0 of the last cycle of the marking signal MS' whose reception has been completed, and during the corresponding iteration of the loop 604-605-606, the nodes 1, 2 complete the reception of the first additional cycle of width N1 (step 607), then AV is updated to AV=(N0+N1) / 2 (step 608). If during the same iteration of the loop 604-605-606, the nodes 1, 2 complete the reception of the second additional cycle of width N2 (step 607), then AV is updated to AV=(N0+N1+N2) / 3 (step 608). And so on, until the iteration of the loop 604-605-606 terminates (the check of step 606 is negative) and the nodes 1, 2 return to step 602, thereby setting a new value for AV.

[0114] This mechanism for setting the AV at the beginning of each iteration of the loop 604-605-606 and updating its value during the same iteration advantageously allows providing a reflection mechanism that also operates correctly when the packet transmission rates in the two directions are different.

[0115] Figure 7(a) and 7(b) shows two different scenarios as an implementation Figure 6 Result of the flow chart of FIG. 1 , an exemplary marker signal MS′ received at the nodes 1 , 2 and a reflected marker signal RMS provided by the nodes 1 , 2. The abscissa axis indicates time.

[0116] FIG7(a) illustrates a case where the packet transmission rate of nodes 1, 2 is lower than the packet transmission rate of other nodes 2, 1. For example, assume that the packet transmission rate of nodes 1, 2 is half the packet transmission rate of the other node 2, 1. Therefore, the rate of incoming packets Pk', Pk at nodes 1, 2 is twice the rate of outgoing packets Pk, Pk' from nodes 1, 2. The incoming packets Pk', Pk carry the marker signal MS'. The width of the period of the marker signal MS' (expressed in the number of incoming packets) is indicated as N0, N1, N2, N3, etc.

[0117] When the measurement session starts, the node 1, 2 initially sets the value of the reflected marker field RMF in the outgoing packet Pk, Pk' to an initial value VA or VB (step 600). When the node 1, 2 completes the reception of the first cycle of the marker signal MS' (step 601), it sets the average cycle width AV equal to its width N0 (step 602) and switches the applicable marker value (step 603).

[0118] Then, nodes 1, 2 start setting the reflected tag field RMF to the currently applicable tag value (loop 604-605-606). Since the incoming packet rate is higher than the outgoing packet rate, the reception of the second period of the tag signal MS' of width N1 is completed before the number of outgoing packets for which the reflected tag field RMF has been set equal to the currently applicable tag value reaches AV=N0 (step 607). Therefore, the node updates the value of AV to AV=(N0+N1) / 2 (step 608).

[0119] When the number of outgoing packets for which the reflected mark field RMF has been set equal to the currently applicable mark value reaches AV=(N0+N1) / 2, nodes 1, 2 terminate the first cycle of the reflected mark signal RMS and start the second cycle by setting the average cycle width AV equal to the width N2 of the last cycle of the mark signal MS' whose reception has been completed (step 602) and switching the applicable mark value (step 603).

[0120] Then, nodes 1, 2 start setting the reflected tag field RMF to the currently applicable tag value (loop 604-605-606). Since the incoming packet rate is again higher than the outgoing packet rate, the reception of the fourth period of the tag signal MS' of width N3 is completed before the number of outgoing packets for which the reflected tag field RMF has been set equal to the currently applicable tag value reaches AV=N2 (step 607). Therefore, the node updates the value of AV to AV=(N2+N3) / 2 (step 608).

[0121] When the number of outgoing packets for which the reflected marker field RMF has been set equal to the currently applicable marker value reaches AV=(N2+N3) / 2, the nodes 1, 2 terminate the second period of the reflected marker signal RMS as described above. And so on.

[0122] Therefore, nodes 1, 2 essentially "merge" R consecutive cycles of the received marker signal MS' into a single cycle of the reflected marker signal RMS, whose width is the average of the widths of the R merged cycles. The number R of merged cycles is essentially an integer, approximating the ratio between the packet rates in the two directions (R=2 in the exemplary case of Figure 7(a)). Therefore, the duration of a cycle of the reflected marker signal RMS is essentially equal to the cumulative duration of the R merged cycles of the received marker signal MS'. This advantageously allows avoiding an infinite accumulation of the value of the reflected marker field RMF in the incoming packets Pk', Pk to be reflected at nodes 1, 2, which would congest the reflection mechanism at nodes 1, 2 and prevent the normal operation of the packet loss measurement.

[0123] FIG7( b) illustrates a case where the packet transmission rate of nodes 1, 2 is higher than the packet transmission rate of other nodes 2, 1. For example, assume that the packet transmission rate of nodes 1, 2 is twice the packet transmission rate of the other node 2, 1. Therefore, the rate of incoming packets Pk', Pk at nodes 1, 2 is half the rate of outgoing packets Pk, Pk' from nodes 1, 2. The incoming packets Pk', Pk carry the marker signal MS'. The width of the period of the marker signal MS' (expressed in the number of incoming packets) is indicated as N0, N1, N2, N3, etc.

[0124] When the measurement session starts, the node 1, 2 initially sets the value of the reflected marker field RMF in the outgoing packet Pk, Pk' to an initial value VA or VB (step 600). When the node 1, 2 completes the reception of the first cycle of the marker signal MS' (step 601), it sets the average cycle width AV equal to its width N0 (step 602) and switches the applicable marker value (step 603).

[0125] Then, the nodes 1, 2 start setting the reflected marker field RMF to the currently applicable marker value (loop 604-605-606). Since the incoming packet rate is lower than the outgoing packet rate, the nodes 1, 2 have not completed reception of any cycle of the marker signal MS' before the number of outgoing packets for which the reflected marker field RMF has been set equal to the currently applicable marker value reaches AV=N0. Therefore, the nodes 1, 2 do not perform any update of the value of AV=N0 during the transmission of the first cycle of the reflected marker signal RMS.

[0126] When the number of outgoing packets whose reflected tag field RMF has been set equal to the currently applicable tag value reaches AV=N0, nodes 1, 2 terminate the first cycle of the reflected tag signal RMS and start the second cycle by setting the average cycle width AV equal to the width of the last cycle of the tag signal MS' whose reception has been completed, and the average cycle width AV is still N0 (step 602), and switch the applicable tag value (step 603).

[0127] The nodes 1, 2 then start setting the reflected marker field RMF to the currently applicable marker value (loop 604-605-606). Again, the nodes 1, 2 have not completed reception of any cycle of the marker signal MS' before the number of outgoing packets for which the reflected marker field RMF has been set equal to the currently applicable marker value reaches AV=N0. Therefore, the nodes 1, 2 also do not perform any update of the value of AV=N0 during the transmission of the second cycle of the reflected marker signal RMS.

[0128] Nodes 1, 2 then terminate the second cycle of the reflected marker signal RMS and start the third cycle by setting the average cycle width AV equal to the width N1 of the last cycle of the marker signal MS' whose reception was completed (step 602) and switching the applicable marker value (step 603).

[0129] Also during the third period, the nodes 1, 2 have not completed reception of any period of the marker signal MS' before the number of outgoing packets for which the reflected marker field RMF has been set equal to the currently applicable marker value reaches AV = N1. Therefore, the nodes 1, 2 also do not perform any update of the value of AV = N1 during the transmission of the third period of the reflected marker signal RMS.

[0130] Nodes 1, 2 then terminate the third cycle of the reflected marker signal RMS and start the fourth cycle by setting the average cycle width AV equal to the width of the last cycle of the marker signal MS' whose reception has been completed, which is again N1 (step 602), and switching the applicable marker value (step 603).

[0131] Also during the fourth cycle, the nodes 1, 2 have not completed reception of any cycle of the marker signal MS' before the number of outgoing packets for which the reflected marker field RMF has been set equal to the currently applicable marker value reaches AV = N1. Therefore, the nodes 1, 2 also do not perform any updating of the value of AV = N1 during the transmission of the fourth cycle of the reflected marker signal RMS.

[0132] Nodes 1, 2 then terminate the fourth cycle of the reflected marker signal RMS as described above (steps 602, 603). And so on.

[0133] Thus, nodes 1, 2 substantially "copy" each period of the marker signal MS' into R periods of the reflected marker signal RMS, whose widths (in terms of the number of packets) are all equal to the width of the received period. The number R of repeated periods is substantially an integer, approximating the ratio between the packet rates in the two directions (R=2 in the example case of FIG. 7( b)). Thus, the cumulative duration of the R repeated periods of the reflected marker signal RMS is substantially equal to the duration of the corresponding period of the received marker signal MS'. This advantageously allows to avoid interruptions in the reflection mechanism, which would prevent the packet loss measurement from operating correctly.

[0134] Moreover, according to Figure 6 When the algorithm realizes reflection, the observer 10 can be based on the reflection in any direction (see Figure 5 (a) and 5(b)) or in both directions (see Figure 5 (c)) The packet loss measurement is performed by using the transmitted marker signal MS and the reflected marker signal RMS. Since the packet transmission rates in the two directions are different, in order to combine the packet loss measurements derived from the signals transmitted in opposite directions, the observer 10 preferably calculates their average before combining them. This ensures that the combined packet loss measurement is consistent, but the packet transmission rates in the two directions are different.

[0135] It should also be noted that the packet loss measurement based on the value of the reflected marker field RMF exhibits a delay relative to the packet loss measurement based on the value of the marker field MF, due to the delay introduced by the reflection mechanism itself, which starts only after the reception of the first period of the marker signal RM' is completed. In the case of packet loss measurements obtained using these two signals, this delay introduces an inaccuracy in the measurement result related to the first measurement period. Therefore, such measurements are preferably discarded.

[0136] At steps 601 and 607, nodes 1, 2 complete the reception of the period of the marker signal MS'. In these two steps, nodes 1, 2 can conclude that the reception of the period of the marker signal MS' is completed when they detect a change in the marker value included in the marker field MF of the incoming packet Pk', Pk.

[0137] However, reception sequence errors involving incoming packets Pk', Pk at the boundary between two periods of the marker signal MS may occur.

[0138] To manage such errors, the nodes 1, 2 may wait to receive a predefined number of incoming packets Pk', Pk (eg, 5 incoming packets) with the next tag value before concluding that reception of the previous cycle is complete.

[0139] Otherwise, the nodes 1, 2 can conclude that the reception of a period of the marker signal MS' is complete when the first incoming packet Pk', Pk with the next marker value is received. This allows to avoid introducing delays in the reflection mechanism. However, on the other hand, a late incoming packet Pk', Pk with a previous marker value will be counted as part of the next period of the marker signal MS', whose width will then increase and may also be higher than N. If N is not fixed (for example, it can vary within a discrete range including different powers of 2), the observer 10 is preferably provided with a threshold mechanism allowing it to determine - for each period of the reflected marker signal RMS whose width is higher than N - whether a late reflection of a packet Pk', Pk is included, or whether the number N has increased, for example, to a higher power of 2.

[0140] according to Figure 6 In a first variant of the algorithm in step 606, the check at step 606 provides for comparing the number of outgoing packets for which the reflected marker field RMF has been set equal to the applicable marker value with the value of AV rounded to the nearest integer. Due to the updating of the value of AV (steps 607-608), the averaging of the width of the correlation period of the marker signal MS' performed at step 608 may in fact provide a non-integer.

[0141] Additionally, since rounding the value of AV may introduce errors, according to another variation, such errors may be minimized as follows.

[0142] The variable RE is initialized to 0 and at the end of each iteration of the loop 604-605-606, it is updated to RE=round(AV+RE)-AV. This value of RE will then be used in a check 606 of the next iteration of the loop 604-605-606, where the number of outgoing packets for which the reflected marking field RMF has been set equal to the applicable marking value will be compared with the rounded value of AV plus the value of RE.

[0143] According to yet another variant, the step 608 of updating the value of the average period width AV each time the reception of a period of the marker signal MS' is completed preferably provides for excluding from the average calculation the width of the last period of the marker signal MS' whose reception was completed during the transmission of the last period of the reflected marker signal RMS, if that width has been used to update the value of AV during the transmission of the last period of the reflected marker signal RMS. This is preferably performed for each iteration of the loop 604-605-606, except the first one.

[0144] Also according to the above variant, the packet loss measurement based on or involving the value of the reflected marker field RMF in the outgoing packets Pk, Pk' is advantageously very accurate.

[0145] Also according to these variants, in practice, at each node 1, 2, the value of the reflected marker field RMF in the outgoing packets Pk, Pk' is set according to the value of the marker field MF in the packets Pk', Pk incoming from the other nodes 2, 1. Therefore, the number of packets Pk, Pk' detected by the observer 10 with a reflected marker field RMF equal to VA or VB indicates the number of packets Pk', Pk transmitted by the other node 2, 1 that were actually lost. Other events that lead to the nodes 1 and 2 declaring a packet lost, such as the loss of an acknowledgment from another node or the receipt of an acknowledgment after a time threshold has exceeded, have no effect on the value of the reflected marker field RMF in the outgoing packets Pk, Pk'.

[0146] Furthermore, the value of the reflected marker field RMF is independent of any mechanism by which nodes 1 and 2 declare packets Pk or Pk' lost (e.g., the mechanism provided by the QUIC protocol described above) or any other internal mechanism of nodes 1 and 2 that is under the exclusive control of their management entity. Thus, advantageously, the entity managing the observer 10 has full control over the packet loss measurement implementation.

[0147] Applicants estimate that the described method provides accurate packet loss measurements, especially when the packet loss rate does not exceed 20-25% in both directions. Since in most types of connections (especially in QUIC-based connections), the transmission of packets is stopped when the packet loss rate exceeds 30%, the described method is able to operate correctly in all environments where packet loss measurement is feasible and meaningful.

Claims

1. A method for exchanging packets between two nodes (1, 2) of a packet-switched communication network (100), each packet (Pk, Pk') comprising a marker field (MF) and a reflected marker field (RMF), the method comprising: a) each of the two nodes (1, 2) sets a value of the tag field (MF) in an outgoing packet (Pk, Pk') to be transmitted to the other node (2, 1), the value switching alternately between a first tag value (VA) and a second tag value (VB) every N outgoing packets, N being a predefined integer; as well as b) at least one of the two nodes (1, 2), upon receiving an incoming packet (Pk', Pk) from the other node (2, 1), sets the value of the reflected marker field (RMF) of the outgoing packet (Pk, Pk') according to the value of the marker field (MF) of the incoming packet (Pk', Pk).

2. A method according to claim 1, wherein the incoming packet (Pk', Pk) carries a marker signal (MS') consisting of a sequence of values ​​of the marker field (MF) in the incoming packet (Pk', Pk), and the outgoing packet (Pk, Pk') carries a reflected marker signal (RMS) consisting of a sequence of values ​​of the reflected marker field (RMF) in the outgoing packet (Pk, Pk'), step b) comprises transmitting at least one cycle of the reflected marker signal (RMS) in response to the reception of at least one cycle of the marker signal (MS').

3. The method according to claim 2, wherein step b) comprises setting the width of the period of the reflected marker signal (RMS) according to the width of the period of the marker signal (MS'), the width being expressed as the number of packets per period.

4. The method according to claim 3, wherein at step b), the period of transmitting the reflected marker signal (RMS) comprises: b1) when initiating the transmission of said period of said reflected marker signal (RMS), setting the width of said period of said reflected marker signal (RMS) equal to the width of the last period of said marker signal (MS') whose reception has been completed; as well as b2) If reception of at least one additional cycle of the marker signal (MS') is completed before transmission of the cycle of the reflected marker signal (RMS) is terminated, then the width of the cycle of the reflected marker signal (RMS) is updated based on the width of the at least one additional cycle of the marker signal (MS').

5. The method according to claim 4, wherein at step b2) the updating comprises setting the width of the period of the reflected marker signal (RMS) equal to the average width of the at least one further period of the marker signal (MS').

6. A method according to claim 4 or 5, wherein step b) comprises transmitting R cycles of the reflected marker signal (RMS) in response to reception of one cycle of the marker signal (MS'), R being an integer approximating the ratio between the transmission rate of the outgoing packets (Pk, Pk') and the reception rate of the incoming packets (Pk', Pk).

7. The method according to claim 6, wherein the width of each of the R cycles of the reflected marker signal (RMS) is equal to the width of the one cycle of the marker signal (MS'), the width being expressed as the number of packets per cycle.

8. A method according to claim 4 or 5, wherein step b) comprises transmitting one cycle of the reflected marker signal (RMS) in response to the reception of R cycles of the marker signal (MS'), R being an integer approximating the ratio between the reception rate of the incoming packets (Pk', Pk) and the transmission rate of the outgoing packets (Pk, Pk').

9. The method according to claim 8, wherein the width of the one period of the reflected marker signal (RMS) is equal to the average value of the widths of the R periods of the marker signal (MS'), the width being expressed as the number of packets per period.

10. A method for performing packet loss measurements on packets exchanged between two nodes (1, 2) of a packet-switched communication network (100), each packet (Pk, Pk') comprising a marker field (MF) and a reflected marker field (RMF), the method comprising the steps of the method according to any one of claims 1 to 8, and: c) by an observer device (10), detecting the outgoing packet (Pk, Pk'), providing a first parameter based on the value of the marker field (MF) in the detected outgoing packet (Pk, Pk'), and providing a second parameter based on the value of the reflected marker field (RMF) in the detected outgoing packet (Pk, Pk'), and performing the packet loss measurement based on the number N in combination with the first parameter and / or the second parameter.

11. A packet-switched communication network (100) comprising two nodes (1, 2) for switching packets (Pk, Pk'), each packet (Pk, Pk') comprising a marker field (MF) and a reflected marker field (RMF), wherein: - each of the two nodes (1, 2) is configured to set a value of the marking field (MF) in an outgoing packet (Pk, Pk') to be transmitted to the other node (2, 1), the value switching alternately between a first marking value (VA) and a second marking value (VB) every N outgoing packets, N being a predefined integer; and -At least one of the two nodes (1, 2) is further configured to set the value of the reflected marker field (RMF) of the outgoing packet (Pk, Pk') according to the value of the marker field (MF) of the incoming packet (Pk', Pk) while receiving the incoming packet (Pk', Pk) from the other node (2, 1).

12. The packet-switched communication network (100) according to claim 11 further comprises an observer device (10), the observer device (10) being configured to detect the outgoing packets (Pk, Pk'), provide a first parameter based on the value of the marker field (MF) in the detected outgoing packets (Pk, Pk'), provide a second parameter based on the value of the reflected marker field (RMF) in the detected outgoing packets (Pk, Pk'), and perform the packet loss measurement based on the number N in combination with the first parameter and / or the second parameter.

Images