METHOD AND GATEWAY FOR DETECTING AND DIAGNOSING SLOWNESS IN A WIRELESS LOCAL COMMUNICATION NETWORK

Description

technical field

[0001] The present invention relates to a method for detecting and diagnosing slowdowns in a wireless local area network. It also relates to a gateway for detecting and diagnosing slowdowns in a wireless local area network.

[0002] Such a device allows the internet service provider to respond more quickly to user requests when slowdowns occur on the wireless local area network. The invention's field of application is specifically that of wireless networks. Prior art

[0003] Currently, the complexity of home networks is rapidly increasing in terms of topology and the services provided. The home local area network (LAN) topology is evolving from a classic star topology to a more complex tree or mesh topology due to the growing number of devices connected via various and heterogeneous connectivity technologies. Furthermore, the home network provides diverse services such as telephony, IPTV, video on demand, home automation, online gaming, e-health, and more. These services compete with numerous web applications used by the client (Facebook, Skype, YouTube, etc.), potentially leading to quality of service issues. In addition, only a few devices on the home network (the home gateway, the TV set-top box, and a few HNIDSs) are controlled by the Internet Service Provider (ISP), making management and monitoring more challenging for the operator.

[0004] In this context, when a user experiences performance issues, the internet service provider can hardly detect and diagnose the problem. This negatively impacts customer satisfaction and generates costs for the provider (high call rates from dissatisfied customers).

[0005] End-to-end frame transmission delay is a key component of the user experience. Several studies have demonstrated the impact of this delay on the end-user experience. Studies have shown that a delay exceeding 100 milliseconds significantly degrades the experience for online gamers. This observation is confirmed for both Voice over IP services and web browsing.

[0006] Furthermore, in wireless configurations, local area network (LAN) latency varies depending on several factors. Studies on the relationship between certain radio factors and actively measured LAN latency have demonstrated a non-linear relationship between these factors and the variation in latency. This non-linearity makes classical modeling of this relationship impossible.

[0007] Document WO2016 / 115406 discloses a method for diagnosing slowness in a WiFi network which consists of: passively measuring TCP connection statistics, determining connection metrics, processing said measurements by grouping them and using an AI model to predict the slowness state of the network and classify the link quality as 'good' or 'bad'.

[0008] Therefore, the aim of the present invention is to resolve at least one of the aforementioned drawbacks with a more complex modeling based on so-called artificial intelligence techniques. Description of the invention

[0009] At least one objective of the invention is achieved with a method for detecting and diagnosing slowdowns in a local wireless communication network, this local network comprising wireless stations and a gateway for communicating with an external network, this method comprising the following steps: passive measurement of session establishment delays between each wireless station and the external network, passive measurement of metrics relating to a wireless module of the gateway and metrics relating to the wireless connection of each station, processing of said measurements by a collection system to form at least one unique vector grouping said measurements at each given time t, use of an explainable artificial intelligence model to predict the state of the local network from the metrics contained in the unique vector in order to define whether the network state is slow or not, comparison between the prediction of the state of the local network and an actual state obtained from the passive delay measurements, in the case where the network state is slow, an identification of the metrics responsible for the slowness is carried out from the explainable artificial intelligence model.

[0010] Advantageously, the process operates in real time, allowing the internet service provider to take corrective action as quickly as possible. Even more advantageously, the process is autonomous.

[0011] Explainable artificial intelligence (AI) refers to a form of artificial intelligence designed to describe its object, logic, and decision-making process in an intelligible manner. This AI technique is used to perform supervised statistical learning, enabling binary classification of different cases based on the Wi-Fi metrics contained within the vector. Using an explainable AI model also allows for automatic diagnosis of slowdowns observed on the station, identifying the metrics responsible for the slowdown.

[0012] A "passive probe" is defined as a hardware or software device capable of detecting frames passing through the gateway. This detection is based solely on observing the frames that cross the gateway and does not generate any additional measurement traffic (active probe).

[0013] A wireless module can, for example, be a Wifi interface in the gateway for wireless communication with wireless stations.

[0014] Advantageously, the measured delays can correspond to a local network delay and an external network delay. The local and external network delays correspond to the time difference between frames based on their direction of transit. The time difference measured between an outgoing frame to the external network and an incoming frame from that network corresponds to the delay observed on that external network. Similarly, the time difference between an incoming frame from an external network and an outgoing frame to it corresponds to the local network delay. An example of implementation in the case of the TCP transport protocol is: the local network delay corresponds to the time difference between the SYN / ACK and ACK frames, and the external network delay corresponds to the time difference between SYN and SYN / ACK frames. Delay measurements are taken passively on the connections that pass through the gateway.A summary of these measurements can be retrieved periodically, with a period of 300 seconds and / or between 1 second and 360 seconds.

[0015] Relying solely on TCP establishment frames allows the process to function even with limited frame observability (hardware acceleration of frame transmission). Furthermore, the sending and receiving of these types of frames are handled at the network stack level of the workstations. Consequently, the delay caused by the application running on the workstation is eliminated.

[0016] Preferably, unique vectors can be stored over a rolling period of between 1 day and 3 months. These stored unique vectors can be used as the model's training data during its initial use. Training the model allows for improved prediction accuracy when presented with new incoming cases.

[0017] Advantageously, the local network state can be classified into two categories based on a defined threshold. These two categories are: degraded and undegraded. The defined threshold can be 100 milliseconds. Other categories can be defined depending on the specific issues to be detected on the network.

[0018] The comparison between the prediction of the local network state and the actual state obtained from passive delay measurements can be carried out continuously in order to measure the performance of the model and define its evolution.

[0019] A preliminary threshold can be defined, and this threshold is compared to the model's performance metrics. The process can then be automatically restarted on a new dataset if the model's performance metric falls below the threshold. The defined threshold can be 75% and / or between 50% and 90%.

[0020] Advantageously, passive measurements can be performed by probes located in the gateway, with the model and model-related processing able to be implemented in the data collection system or in the gateway. Measurements are performed passively by the probes to avoid introducing additional measurement traffic.

[0021] Preferably, the collection system can be remote.

[0022] The metrics for a wireless module of the gateway may include at least: signal strength on the wireless module, number of time slots occupied due to the presence of wireless stations, frequency band used, number of time slots occupied due to the presence of wired stations, identifier of a channel in use, number of frames retransmitted, synchronization rate, uplink and downlink transmission rate, number of slots available for sending, number of wireless access points scanned on a channel in use, number of wireless access points scanned on channels adjacent to the channel in use, number of frames and / or bytes sent, radar detection information, transmission strength per antenna of the wireless module.

[0023] The term "radar detection" refers to military radars scanned by WIFI gateways, with the scan being carried out in such a way as not to interfere with military radars.

[0024] The metrics relating to the wireless connection of each station may include at least: a manufacturer code of the wireless card of the wireless station, the type of the wireless station, an upstream and / or downstream rate negotiated with the wireless module, a measured radio noise strength, a frequency band used, a number of byte frames sent and / or sent, a number of frames retransmitted, a signal strength measured from the wireless station, a timestamp of the radio inactivity of the wireless station.

[0025] Passive delay measurements can be exported to the collection system using an export format without aggregation or by using statistical aggregation of measurements per wireless station. When passive delay measurements are exported to the collection system using an export format without aggregation, the latency measurement is sent in its raw form. Advantageously, passive delay measurements are exported using aggregation. Using statistical aggregation helps to limit the data volume on the collection system side.

[0026] Aggregated export can be performed over a given time period, ranging from 10 seconds to 30 minutes. Aggregation can be defined relative to an interval T and for each station that generated a set of flows, from a single flow to the total number of flows measured during the interval T. Configuring T allows for managing the desired process responsiveness.

[0027] Advantageously, the measurements can be periodic. Measurements such as passive metric measurements can therefore be periodic, with a period of 300 seconds and / or between 1 second and 360 seconds.

[0028] According to a second aspect of the invention, a gateway is proposed for detecting and diagnosing slowness in a local wireless communication network, the gateway being equipped with a processing unit to implement the steps of the process.

[0029] According to a third aspect of the invention, a computer program product is proposed comprising instructions which, when the program is executed by a computer, lead the computer to implement the steps of the process. Description of the figures and methods of implementation

[0030] Other advantages and features of the invention will become apparent upon reading the detailed description of implementations and embodiments, which are by no means limiting, and the following attached drawings: [ Fig.1 ] illustrates a home wireless local area network, [ Fig. 2 ] illustrates the time measurements for each TCP establishment according to one embodiment of the invention, [ Fig.3 ] illustrates a flowchart of a process according to an embodiment of the invention.

[0031] These embodiments are not exhaustive; in particular, variants of the invention may be considered that comprise only a selection of features described or illustrated hereafter, isolated from the other described or illustrated features (even if this selection is isolated within a sentence including these other features), provided that this selection of features is sufficient to confer a technical advantage or to differentiate the invention from the prior art. This selection includes at least one preferably functional feature without structural details, and / or with only a portion of the structural details if this portion alone is sufficient to confer a technical advantage or to differentiate the invention from the prior art.

[0032] We will first describe, with reference to the [ Fig.1[ ], a wireless home local area network 100. The network center is a home gateway 1 distributing an internet connection to the various stations 2 connected throughout the home. The home gateway 1 is located in the home, for example, in the living room. The home gateway 1 is the central point for all data streams: images, music, videos, etc. Each station 2 is connected to the gateway via a wireless cable using Wi-Fi or an Ethernet cable. The home gateway 1 is connected, on one side, to the home local area network 4 (LAN) and, on the other side, to the external network 5 (WAN). The local area network 4 refers to any interconnected computer equipment or station 2 within the user's home, such as a television, tablet, mobile phone, or game console. The present invention focuses on the local area network 4.The external network 5 is connected to a remote collection system 3. The home local area network context is used as an example. The method described by the present invention is applicable to other types of wireless communication networks, for example, university networks, corporate networks, etc.

[0033] We will describe, with reference to figures 2 And 3 , a method for detecting and diagnosing slowdowns in the local wireless communication network.

[0034] According to the [ Fig.3 ], the process is broken down into five steps. Initially, a passive measurement of session establishment delays 6 between each wireless station 2 and the external network 5 is performed. According to the [ Fig.1 ], three time-limiting measures are therefore implemented. The [ Fig. 2 ] illustrates the time measurements for each TCP establishment according to one embodiment of the invention. The [ Fig. 2] presents the links established between a station 2, the gateway 1 and the collection system 3.

[0035] Communication between gateway 1, application server 3 and user station 2 is initiated by transmission of frames: SYN, SYN / ACK and ACK.

[0036] Home gateway 1 is equipped with a set of passive probes including a network delay probe based on user traffic sent by a station 2 and a probe collecting WIFI performance metrics from each associated wireless station 2 and a WIFI interface present on home gateway 1. The delay probe passively observes the establishment of TCP sessions from each station 2. For each observed TCP session, a WAN delay corresponding to the time difference between the SYN and SYN / ACK frames is measured, as well as a LAN delay corresponding to the time difference between the SYN / ACK and ACK frames (see [ Fig. 2 The method is therefore capable of measuring the LAN delay and dissociating it from the WAN delay. Each delay measurement is exported and recorded for each connected station 2 in the remote collection system 3. The passive delay measurements are exported to the collection system 3 using an export format that employs statistical aggregation of the measurements per wireless station 2 over a given time period. A histogram including the minimum, maximum, mean, and variance of the latency measurements observed over the period is sent. An aggregation can be composed, for example, as follows: A key to uniquely identify the generating station (MAC address, IP address), Minimum LAN delay measured for any Fi stream expressed in milliseconds (milliseconds), Maximum LAN delay measured for any Fi stream, expressed in milliseconds (milliseconds), Average LAN delay measured for over all Fi streams (milliseconds), Variance of LAN delay measured for over all Fi streams (milliseconds), Standard deviation of LAN delay measured for over all Fi streams (milliseconds), Median LAN delay measured for over all Fi streams (milliseconds), Timestamp of the first observed stream in period T, Timestamp of the last observed stream in period T.

[0037] According to the [ Fig.3], the delay measurements are followed by the passive measurement of metrics 7 relating to a wireless module of gateway 1 and metrics relating to the wireless connection of each station 2. The probe allowing the measurement of the WIFI stations is located in gateway 1. The metrics reported belong to two categories: measurements by radio interface present on the home gateway 1 and measurements by station 2 connected to the home gateway 1.

[0038] The measurements, which were taken periodically, are then processed by the data collection system and grouped to form at least one unique vector at each given time t. These unique vectors are stored in the data collection system. A unique vector therefore consists of at least one time measurement and one metric measurement.

[0039] An explainable artificial intelligence model 9 is then used to predict the state of the local network 4 from the metrics contained in the unique vector. The model and model-related processing are implemented in the collection system 3. The unique vectors stored in the collection system 3 are used as the model's training data.

[0040] A measurement of the actual local network is performed. The state of the actual local network is classified into two categories: "degraded", in the case where the measurement is greater than one hundred milliseconds, or otherwise, "non-degraded".

[0041] A comparison is performed between the predicted state of the local network (LAN) and its actual state, obtained from passive latency measurements. If the network state is considered "degraded," and the predicted LAN state matches the measured actual state, the internet service provider (ISP) then uses the model's explanation for diagnosis. The metrics responsible for the slowness on LAN are identified, and the ISP can intervene with the user more quickly. In the case of a station connected via an Ethernet cable, the LAN latency tends to be low, and potential degradation is generally explained by simple factors such as cable length and quality.

[0042] Typically at least one of the means of the device according to the invention previously described, preferably each of the means of the device according to the invention previously described, are technical means.

[0043] Typically, each of the means of the device according to the invention described above may include at least one computer, a central processing unit or computing unit, an analog electronic circuit (preferably dedicated), a digital electronic circuit (preferably dedicated), and / or a microprocessor (preferably dedicated), and / or software means.

[0044] Of course, the invention is not limited to the examples just described and many modifications can be made to these examples without departing from the scope of the invention.

[0045] Of course, the various features, forms, variants, and embodiments of the invention can be combined in various ways, provided they are not incompatible or mutually exclusive. In particular, all the variants and embodiments described above are combinable.

Claims

1. Method for detecting and diagnosing slowness in a local network (4) for wireless communication, said local network (4) comprising wireless stations (2) and a gateway (1) for communicating with an external network (5), said method comprising the following steps: - passive measurement of session set-up delays (6) between each wireless station (2) and the external network (5) by a probe in the gateway (1), characterized in that the method further comprises the following steps: - passive measurement of metrics (7) relating to a gateway wireless module and metrics relating to the wireless connection of each station (2) by the probe in the gateway (1), - processing of said measurements (8) by a collection system (3) to form at least one unique vector grouping said measurements at each given time t, - use of an explainable artificial intelligence model (9) to predict the state of the local network (4) from the metrics contained in the unique vector, so as to define whether the state of the network is slow or not, - comparison (10) between the predicted state of the local network (4) and an actual state obtained from passive delay measurements by the collection system (3), - if the network state is slow and the prediction of the state of the local network (4) corresponds to the actual measured state, the metrics responsible for the slowness are identified using the explainable artificial intelligence model, the model and model-related processing being implemented in the collection system (3).

2. Method according to claim 1, wherein the measured delays correspond to a local network delay and an external network delay, the local and external network delay corresponding to the time difference between frames depending on their transit direction.

3. Method according to either of the preceding claims, wherein the unique vectors are stored over a sliding period of between 1 day and 3 months.

4. Method according to claim 3, wherein the stored unique vectors are used as a basis for training the model when it is first used.

5. Method according to any of the preceding claims, wherein the state of the local network (4) is classified into two categories according to a defined threshold.

6. Method according to claim 5, wherein the two categories of the state of the local network (4) are: degraded and non-degraded.

7. Method according to any of the preceding claims, wherein the comparison (10) between the predicted state of the local network (4) and the actual state obtained from passive delay measurements is carried out continuously in order to measure a performance of the model and define its evolution.

8. Method according to claim 7, wherein a prior threshold is defined, the defined threshold being compared with model performance measurements, the method being automatically restarted on a new data set if the model performance measurement is below the threshold.

9. Method according to any of the preceding claims, wherein the collection system (3) is remote.

10. Method according to any of the preceding claims, wherein the metrics relating to a wireless module of the gateway (1) comprise at least: - a signal strength on the wireless module, - a number of busy time intervals due to the presence of wireless stations (2), - a frequency band used, - a number of busy time intervals due to the presence of wired stations, - a channel identifier used, - a number of retransmitted frames, - a synchronization rate, - a transmission rate in upstream and downstream directions, - a number of intervals available for sending, - a number of wireless access points scanned on a channel in use, - a number of wireless access points scanned on channels adjacent to the channel in use, - a number of frames and / or bytes sent, - information on radar detection, - an antenna transmitting force of the wireless module.

11. Method according to any of the preceding claims, wherein the metrics relating to the wireless connection of each station (2) comprise at least: - a manufacturer code for the wireless card in the wireless station (2), - the type of the wireless station (2), - an upload and / or download rate negotiated with the wireless module, - a strength of the measured radio noise, - a frequency band used, - a number of byte frames and / or frames sent, - a number of retransmitted frames, - a signal strength measured from the wireless station (2), - a time stamp of the radio inactivity of the wireless station (2).

12. Method according to any of the preceding claims, wherein passive delay measurements are exported to the collection system (3) in an export format without aggregation or using statistical aggregation of the measurements per wireless station (2).

13. Method according to any of the preceding claims, wherein the measurements are periodic.

14. System for detecting and diagnosing slowness in a local network (4) for wireless communication, comprising a gateway, stations and a collection system, the gateway and collection system being configured to implement the method steps according to any of claims 1 to 13.