Probe for measuring quality-of-service parameters in a telecommunication network

0001The present invention relates to the measurement of characteristic parameters of the equipment traversed by a data flow within a data network, in particular of telecommunications. It applies particularly well to the measurement of quality of service parameters rendered for data streams passing over this telecommunication network, but could apply to other characteristics of the equipment such as their load, their temperature, the state of their queues etc., located on the path of these data flows.

0002It is indeed important to have measurements of certain parameters in order to verify the proper functioning of its network, and in particular to know if the quality of service requested by customers is actually provided.

0003To do this, there are various devices known from the state of the art.    For example, the company Ipanéma markets measurement probes which can be placed at the access of the telecommunications network, as indicated in FIG. 1: the S probes₁ and S₂ are connected to the telecommunications network N. When data flows pass through its probes S₁ and S₂, they measure certain parameters and supply these parameters to a measuring device M.    The measuring device M transmits to the probes information on the parameters which they have to measure. It can thus configure the data stream (s) to be measured, the frequency of the measurements, etc.

0004However, such a device suffers from a major problem when the telecommunications network comprises several domains, each domain being able to be administered by a different telecommunications operator. The probes can only be installed at the ends of the domain administered by the telecommunications operator. Once you are in a real environment, that is to say made up of several domains, it is no longer possible to obtain end-to-end measurements, since the operator of a domain does not will generally be able to access only equipment in its domain, to the exclusion of other domains.    Furthermore, it may be advantageous to have a measurement not between the ends of the network or of the domain, but between the telecommunication terminals themselves, or even on the different domains or equipment through which a data flow passes.    This is particularly desirable in the case of telephony terminals over IP (Internet Protocol). In this situation, it does not seem clear how to install and / or configure the probes at the customer or within the networks crossed.    Thus, the entity wishing to carry out the measurements must to do this, discover or configure the various measurement probes set up in the various areas of the telecommunications network. The solution of the state of the art is silent on this problem.

0005The aim of the present invention is to propose a solution for the measurement of parameters, in particular of quality of service, which is easily configurable and which does not suffer from the problems of the solutions of the prior art.

0006To do this, the invention relates to a measurement probe, having means for accessing the data streams consisting of packets, transmitted along a path formed by a plurality of equipment of a telecommunication network, and measuring means for carrying out measurements, in accordance with configuration information. This probe is characterized in that it also has<ul id="ul0001" list-style="bullet" compact="compact"><li>determination means for determining that one or more packets transmitted along this path form a signaling message and</li><li>signaling means for determining from this signaling message, the configuration information.    Preferably, these measurements relate to said data flows.</li></ul>

0007According to one embodiment of the invention, the measurement means are capable of transmitting measurement reports, containing the measurements, to a measurement device determined by an identifier contained in the configuration information.

0008According to one embodiment of the invention, the measurements are transmitted to the measurement device via a mediator, the data transmitted to the mediator containing this identifier.

0009According to one embodiment of the invention, the determination means are capable of reading a specific mark, contained in the message received, and of determining whether this received message is a signaling message from this specific mark.

0010According to one embodiment of the invention, the configuration base contains a set of records, each record corresponding to a measurement task and containing in particular:<ul id="ul0002" list-style="bullet" compact="compact"><li>a filter determining the packets to which the measurements must relate,</li><li>parameters relating to the measurement method</li></ul>

0011The parameters can in particular be chosen from a set comprising:<ul id="ul0003" list-style="bullet" compact="compact"><li>the period during which the measurements must be carried out,</li><li>Sampling information, including a hash function,</li><li>a parameter triggering the time stamp of the packets to be measured,</li><li>a parameter triggering the identification of the packets to be measured, in particular by means of a hash function.</li><li>A parameter triggering the counting of packets,</li><li>the method for transmitting the measurements to the measuring device.</li></ul>

0012According to one embodiment of the invention, the transmissions with the measuring device are secure. These security means can in particular be transmitted by a signaling message.

0013According to one embodiment of the invention, the measurement probe further comprises means for deciding the creation of a new measurement task, by the signaling means, in particular as a function of a sensitivity indicator associated with this measuring probe.

0014According to one embodiment of the invention, the decision is also a function of a priority contained in the message received.

0015The invention also relates to a network element, in particular a router, comprising a measurement probe as described above, as well as a communication network comprising such measurement probes, and possibly a measurement device.

0016Thus, by the use of a signaling protocol “in the path” to indicate to the measurement probes to establish measurement tasks or to modify or delete these, the invention makes it possible to have no prior knowledge of the location of the measurement probes, and of circumventing the problem of the measurement probes arranged in a domain administered by an operator different from that of the measurement device.

0017The invention and its advantages will appear more clearly in the description of implementations which will follow in conjunction with the appended figures.<ul id="ul0004" list-style="none" compact="compact"><li>FIG. 1, already commented on, illustrates a solution of the state of the art.</li><li>FIG. 2 represents the functional architecture of a probe according to the invention.</li><li>FIG. 3 diagrams the communications between the probes according to the invention and a measurement device.</li><li>FIG. 4 illustrates the probe of the invention in an implementation context.</li></ul>

0018According to different implementations of the invention, the measurement probe can be incorporated in a specific device, such as those of the state of the art of the company Ipanéma, or else in network equipment such as a switch, a router. IP etc.    In the latter case, the measurement probe may in particular be a software module which can be executed by the operating system of the network equipment. This software module can be installed during the commissioning of the network equipment or later as part of an update of the software of this equipment, and / or dynamically by downloading through the network, by example from a dedicated server. This software module can for example be developed in Java ™ language, in order to facilitate its dynamic implementation on network equipment.

0019By network equipment is meant in particular, within the framework of a network based on an IPv4 or IPv6 protocol stack (<i>Internet Protocol, version 4</i>/<i>6</i>), a router.

0020Subsequently, the implementation of the invention for measuring quality of service parameters will be more particularly detailed, although the invention may also be applied to other parameters.

0021FIG. 2 illustrates the functional architecture of a measurement probe S, in accordance with the invention.

0022This measurement probe firstly comprises means of determination S_Of The role of these determination means is to determine if one or more incoming data packets form a signaling message or if they belong to a data stream. In the typical case of a data network based on an IPv4 protocol stack<i>(Internet Protocol version 4</i>) or IPv6 (<i>Internet Protocol, version 6</i>), signaling messages can in fact consist of several data packets.    The determination can be made by a specific brand. This specific brand can be a dedicated port number, a dedicated DSCP (<i>DiffServ Code Point</i>), an IP header protocol number etc.

0023If the group of received data packets forms a signaling message, it (or its content) is transmitted to signaling means S_S, whose role is to interpret the content of this signaling message. Depending on the content of this message, the signaling means can modify a configuration base B_VS. Configuration base B_VS contains the configuration of the probe. It can include a set of records, each record corresponding to a measurement task.

0024In general, all or part of the records in configuration database B_VS determines which data flow should be measured by the corresponding measurement task, how often measurements should be made, what parameter they should relate to etc. (We will see later that, according to an embodiment of the invention, some of these records may not correspond to a measurement task).    According to IETF terminology, these records correspond to a state of the probe. These states can be of the type defined for example, for the RSVP protocol (<i>ReSerVation Protocol</i>) defined by IETF RFC 2205.

0025The content of the recordings will be detailed later, but it is important to note here that the signaling messages can trigger:<ul id="ul0005" list-style="bullet" compact="compact"><li>Establishing a new measurement task. This establishment causes the insertion of a new record, in the configuration database B_VS, and therefore the creation of a new state, within the measurement probe. This state can preferably be of the “soft state” type, that is to say that it will be automatically deleted at the end of a certain period.</li><li>Refreshing a state. In the embodiment where the states are so-called "soft states", refresh messages make it possible to extend this period, for example by resetting a counter to an initial value. Of course, if the states are so-called "hard states", no refresh message is necessary, because the state will remain installed until a delete message is received, concerning this state.</li><li>Editing a measurement task. This type of message can aim to modify part of the parameters associated with a previously established measurement task (for example, to change a sampling rate of the measurements, dynamically, to adapt to the network load, or near a critical threshold). The corresponding record in configuration database B_VS can be changed to reflect this change.</li><li>The deletion of a measurement task. This deletion can cause the deletion of the corresponding record in configuration database B_VS. In the situation where the states are of the “hard states” type, deletion messages are transmitted to complete the measurement task and delete the corresponding state.</li></ul>

0026Furthermore, according to one embodiment, the groups of normal packets are transmitted by the determination means to measurement means S_M. Normal packets are understood to mean packets whose content is not interpreted by routers as are the contents of packets from different network protocols, such as signaling packets, routing packets, ICMP packets etc. However, the invention can also be applied to the measurement of “non-normal” packet flows, such as, for example, OSPF signaling flows (<i>Open Shortest Path First).</i>

0027The role of these measurement means S_M is actually to carry out the measurement on the received packets, according to the configuration memorized in the configuration base B_VS. More specifically, the role of the measurement means is to process the various tasks that have been set up on the measurement probe; the configuration of each task being determined by the content of the corresponding record in the configuration database B_VS.

0028As said before, this configuration can determine several things, for each of the measurement tasks.

0029First, it can determine what the measurements should relate to, i.e. the data flows to be measured, for example by means of a list of data flow identifiers.    To do this, filters can be set up, in order, very generally, to select a subset of packets by applying deterministic functions to parts of the contents of the packet such as header fields or parts payload. A filter can also consist in applying a pseudo-probabilistic law to select the subset. The concept of filter can be in conformity with that for example defined in the draft IETF "draft-ietf-psam-sample-tech-00.txt".    These filters can in particular be used to select the packets belonging to one or more data streams, on the basis of an identifier list.    Typically, these identifiers can be, in the case of an IP network, a 5-tuple composed of the addresses and port numbers of the sender and of the recipient of the stream, and the protocol number. In the case of an IPv6 network (Internet Protocol version 6), it can be added to this 5-tuple, the "Flow Label" field.

0030The configuration can also specify how the measurements should be made. More specifically, it can possibly indicate:<ul id="ul0006" list-style="bullet" compact="compact"><li>the duration during which the flow must be the subject of measurements. Alternatively, it is possible not to specify a duration, the stopping of the measurements then having to be indicated by the emission of another signaling message or by the expiration of a delay in the absence of a refresh message. According to this embodiment, there is a state mechanism (soft state) similar to that implemented for the RSVP protocol (<i>ReSerVation Protocol</i>).</li><li>if these measurements must relate to the whole of the packets, or on the contrary if a sampling must be carried out. In the case of sampling, the configuration can also contain the frequency of the measurements (one packet out of n; 1 packet every n milli-seconds ...), a hash function with a constraint on the result, etc.</li><li>a parameter triggering the time stamp of the packet,</li><li>a parameter triggering the identification of the packet thanks to a hash function.</li><li>A parameter triggering the counting of packets,</li><li>the method for transmitting the measurements to the measurement device M, in particular if these measurements must be transmitted for each measurement carried out, or else if they must be grouped in a single message in order to limit communications. In the latter case, the configuration can contain the transmission frequency (a transmission for n measurements, a transmission every n milli-seconds ...), etc.</li><li>...</li></ul>

0031As will be seen later, the configuration may also indicate an identifier of the measurement device, and of the security means.

0032The choice of parameters contained in the signaling message may in particular depend on the type of measurement to be carried out. Thus, the parameters may be different if it is a question of measuring an average transmission delay or a packet loss rate.

0033In the case of a measurement of an average transmission delay, one embodiment of the invention consists in carrying out the following steps:<ul id="ul0007" list-style="none"><li>1) sampling: it is important not to select all the packets of the sub-assembly treated, so as not to clutter the network and the collector M, but, at the same time, a minimum number is necessary. An additional difficulty is that the same packages must be selected by all of the measurement probes so that a correlation can be possible by the collector (s) M.    A deterministic sampling process is therefore implemented, for example by means of a hash function. A hash function can be a mathematical, non-bijective application, which associates with an invariant packet content (ie not modified by network elements), such as the payload of the packet, a value which is tested for determined. whether the packet should be part of the sample or not: as this function is an application, and as it is based on an invariant, two probes will result in the same value, and therefore will make the same decision.    In practice, this function can be chosen as a function of the desired probability of sampling, the speed of the data flow, and the entropy of the content of the packets.</li><li>2) Then, we associate a date with the selected package. At this stage, all the probes should have synchronized clocks. To do this, synchronization techniques known from the prior art may be used, in particular the use of a GPS (Global Positioning System) or else of the NTP protocol (Network Time Protocol), defined by RFC 1305 of the IETF<i>(Internet Engineering Task Force</i>).</li><li>3) Thirdly, we "identify" the selected package. That is to say that it is associated with a value making it possible to uniquely identify it among the other packets of the same flow and those of the other data flows. Again, identification can be accomplished using a hash function. The result of the hash function, which forms the packet identifier, must be long enough to prevent two different packets from having identical identifiers. The hash function must be identical for all the probes, so that the same packet is associated with the same identifier and allow the measuring device M (or collector) to be able to correlate the reports coming from it from the probes.</li><li>4) Finally, the fourth step consists in transmitting a measurement report to the measurement device or collector M.    Thus, for the same sampled packet, the collector M receives several measurement reports from different probes. Thanks to the unique property of identifiers, it can easily correlate its measurement reports and, by comparing the dates inserted into it by the probes, it can determine the delay of the sampled packet between each probe.    In this example, the signaling message therefore comprises the following elements: a filter, a hash function for sampling, a parameter triggering the time stamping of the packets, a hash function for identification.</li></ul>

0034In the case of a measurement of a packet loss rate, the principle is substantially the same as in the previous example. According to one embodiment, the difference lies in the fact that instead of the date of reception of the packet, the measurement report contains the serial number of the packet, given by a counter contained in the probe.    In this example, the signaling message therefore includes the following elements: a filter, a hash function for sampling, a parameter triggering the counting of packets.

0035As mentioned above, the measurements made by the measurement means S_M can then be transmitted to a measurement device, not shown in FIG. 2, the aim of which may be to consolidate the measurements received from several measurement probes. These measurement devices can also be called "collectors" (or "collector", according to the terminology in English).    An identifier of this measurement device can for example be indicated in the configuration base B_VS. This measurement device can in particular be different depending on the data flows measured. This identifier can be provided by the signaling messages and can be inserted into the configuration base by the signaling means S_S, like any other configuration information.    This identifier can be an IP address (<i>Internet protocol</i>), or a more abstract protocol number or address, such as a URL (<i>Unified Resource Locator</i>) described by RFC 2396.    In addition, the measurements can be sent to the measurement device via mediators (or proxies, according to the terminology in English), as shown in FIG. 3. The telecommunications network N is composed of a set of network equipment divided into a plurality of groups. To each group G₁, G₂, G₃... G_not is associated with a mediator (or <i>proxy</i>), respectively P₁, P₂, P₃... P_not. The measurements taken by the measurement probes of a network equipment are transmitted to the mediator associated with the corresponding group. This mediator can then transmit the measurements to the measurement device M. According to one embodiment of the invention, an identifier (the address, for example) of the measurement device M is inserted in the measurement reports transmitted to the mediators, in order to that they are able to transmit the measurement reports to the appropriate measurement device.

0036Possibly, the mediators can carry out a pre-treatment before sending it to the measurement device M. This pre-treatment can for example simply consist in aggregating the measurements received from the probes, in order to send more synthetic reports to the measure M and limit traffic.

0037This embodiment is advantageous in the case of large telecommunications networks, since it makes it possible to better distribute communications between network elements and measurement device (s), as well as to limit inter-operator communications in the event of measurements on different networks.

0038According to an embodiment of the invention, the measurements can be transmitted to the measurement device in a secure form, for example in a coded manner by a public key.

0039One of the advantages of the invention is to easily establish and determine a large number of measurement probes. These measurement probes can be redundant, that is to say more numerous than would be necessary. For example, to measure quality of service parameters between 2 points A and B, two probes would be necessary, but one can choose to establish 2 in the vicinity of point A and 2 in the vicinity of point B. The advantage of such redundancy is to minimize the risks of measurement errors or deficiency of a measurement probe.    Another advantage of the invention is to easily find and configure measurement probes. State-of-the-art architectures with 2 probes require determining which probes can be used and accessing them. In a multi-domain situation, a measurement can hardly be requested by an operator, on a probe from another operator. In addition to solving these problems, the invention makes it possible to make several measurements along the path of a flow to better locate a dysfunction (congestion, a quality of service problem, etc.).    Another advantage of the invention is that the measurements are carried out by the measurement probes without the latter being aware of the presence of the others, and <i>a fortiori</i>, measurements made by other measurement probes. Also, any intentionally erroneous measurement provided by a measurement probe can be easily detected by comparison with measurements provided by nearby measurement probes.

0040According to one embodiment of the invention, the signaling means S_S also have the means to decide whether or not to create a new measurement task.    It can be chosen not to insert in the configuration database B_VS that the records associated with created measurement tasks, or the records associated with any signaling message requiring the creation of a measurement task, whether or not this is accepted by the signaling means S_S. This second implementation is especially interesting when states of the “soft states” type have been chosen: in this implementation, refresh messages can be regularly received. Keeping track of "refused" signaling messages helps to maintain consistency in the decisions taken.

0041In order to make these decisions, the measurement probe is associated with a sensitivity indicator.    This sensitivity identifier can for example represent a probability that the signaling probe decides to process the signaling message. For example, when it receives a signaling message, the measurement probe can trigger the drawing of a random number. By comparison with the sensitivity indicator, it easily determines whether the signaling message should be treated or not.    According to one implementation, this mechanism only acts for signaling messages containing information relating to the addition of a measurement task. Conversely, signaling messages modifying a measurement task or deleting a previously existing measurement task can always be processed, i.e. imply a modification of the configuration base B_VS, by signaling means S_S.

0042According to an implementation of the invention, the signaling messages can contain a priority. The decision whether or not to process the signaling message can be weighted by the value of this priority.    For example, “routine” measurement (monitoring) work may be assigned a low priority. If an anomaly has been identified at a given time, a management system may decide to transmit a signaling message with a higher priority in order to trigger measurements by a greater number of measurement probes, and therefore to allow a more local specifies the problem.    According to a particular embodiment of the invention, the signaling messages can be stored in another database, not shown in the figure, even if the signaling means decide not to accept the creation of a new task for measures and does not modify the configuration base B_VS.

0043FIG. 4 illustrates an implementation of the invention.    A telecommunications network includes 5 measurement probes A, B, C, D and E. A signaling message is transmitted to A and then successively to B, C, D and E. This signaling message contains measurement information relating to the implementation of a measurement task. The measurement probes have different sensitivity indicators: the probes A, C, D and E decide to insert this measurement information in their respective configuration bases. The measurement probe B decides to ignore the signaling message and does not modify its configuration base.    When messages belonging to the data flow corresponding to this measurement information, the probes A, C, D and E take measurements in accordance with this measurement information, as indicated above. These measurements are transmitted to a measurement device M.    Insofar as the measurement probes C and D are juxtaposed, if the measurements transmitted by these two probes differ beyond an acceptable margin of error, the measurement device M will be able to determine that one of the probes measurement at least is deficient.    If the measurements from measurement probes A and C differ beyond a certain level, the measurement device can determine that an anomaly exists between these two probes.    In order to specify the location of the anomaly, the measuring device M can cause a new signaling message to be sent to the measuring probe A, with a higher priority. This time, the measurement probe B decides to establish a measurement task and to insert the measurement information into its configuration base. Upon receipt of a data flow message, the measurement probe B will also transmit measurements to the measurement device M.    By comparing the measurements received on the one hand from probes A and B and on the other hand from probes B and C, the measuring device can determine whether the anomaly is between A and B or between B and C (or if it is distributed between A and C).

0044An additional advantage of the invention is that the measurement probes A, C, D and E transmit their measurements independently of each other. Likewise, each of the measurement probes cannot be informed of the content of the measurements of the other probes, and even of the existence of these measurements and of the measurement probes themselves.    This results in a high level of security / reliability of the invention.    In the case where the network is multi-domain, that is to say managed by several operators, these operators can thus have the assurance that the measurement information cannot be known by the measurement probes belonging to a managed domain. by another operator.

0045According to one embodiment of the invention, the signaling messages conform to the following grammar, defined in the form of Backus-Naur (BNF, for <i>Backus-Naur Form,</i> in English):<img file="EP1453242A2_D0001.tif" />

0046This grammar indicates that a signaling message according to the invention indicates, for the creation of a measurement task:<ul id="ul0008" list-style="bullet" compact="compact"><li>an identifier of the measurement task,</li><li>an “acceptance factor” priority which, in collaboration with the sensitivity indicator, makes it possible to decide whether or not to create the measurement task,</li><li>a filter, "FIOW_FILTER" allowing to select a subset of packets, on which the measurements must be made,</li><li>“METERING_ACTIONS” measurement actions, which make it possible to specify which type of processing must be carried out by the measurement means S_M, especially if it is counting ("COUNTER"), sampling ("SAMPLING"), identification ("IDENTIFICATION") ...</li><li>the identification of a collector M, in particular its address and optionally, the frequency and safety parameters.</li></ul>