SD-WAN-based intelligent hybrid link management method and system

Abstract

The invention discloses an SD-WAN-based intelligent hybrid link management method and system, and belongs to the field of information transmission, and the SD-WAN-based intelligent hybrid link management method comprises the following steps: monitoring the quality of each link, and determining the link with the optimal quality, the link quality indexes comprising delay, packet loss rate and jitter. The method has the advantages that when the game is started, the flow is classified, different links are distributed to different types of flow, game operation is guaranteed preferentially, and network delay in the game process is reduced; when it is detected that high-priority flow transmission does not reach a set value, bandwidth threshold limitation is carried out on non-game flow and low-priority game flow, the bandwidth of the high-priority game flow is improved, and the delay, jitter and packet loss rates of games are guaranteed; when the game is abused, voice and characters are automatically shielded, the game environment is purified, and the game experience of players is improved.

Description

Technical Field

[0001] The present invention belongs to the field of information transmission, and in particular, relates to an intelligent hybrid link management method and system based on SD-WAN. Background Art

[0002] SD-WAN (Software Defined Wide Area Network) is a wide area network management technology based on Software Defined Network (SDN). Through intelligent routing and traffic optimization technologies, SD-WAN can significantly reduce network latency during games, enabling players to enjoy more immediate game feedback and interactive experiences.

[0003] During games, non-game traffic (background updates, advertisement pushes) and low-priority game traffic (game update file downloads) will occupy bandwidth, which will affect the transmission of high-priority game traffic (real-time operation of game character movements, real-time voice, real-time text, etc.), resulting in game lag and needing improvement. Summary of the Invention

[0004] Based on this, it is necessary to provide an intelligent hybrid link management method and system based on SD-WAN for the above problems.

[0005] An embodiment of the present invention is implemented as follows. An intelligent hybrid link management method based on SD-WAN includes the following steps:

[0006] Monitor the quality of each link, determine the link with the best quality. The link quality indicators include delay, packet loss rate, and jitter, and allocate high-priority game traffic to the link with the best quality;

[0007] When high-priority game traffic is monitored to be transmitted, allocate low-priority game traffic, including game update file downloads, to the link with the second-best quality; allocate non-game traffic, including background updates and advertisement pushes, to other links except the ones with the best and second-best quality;

[0008] If the transmission of high-priority game traffic on the link with the best quality does not reach the set value (the set value can be a delay ≤ 50 ms, a jitter ≤ 10 ms, and a packet loss rate less than 1%), impose bandwidth threshold restrictions on non-game traffic and low-priority game traffic (such as limiting the speed to 1 Mbps and 5 Mbps respectively), release the occupied physical link bandwidth resources, and increase the available bandwidth of the link with the best quality.

[0009] In one embodiment, the present invention provides an intelligent hybrid link management method based on SD-WAN. In the step of monitoring the quality of each link, determining the link with the best quality, where the link quality indicators include delay, packet loss rate, and jitter, and allocating high-priority game traffic to the link with the best quality, it specifically includes:

[0010] Combining active detection and passive detection to obtain link quality indicator data. The link quality indicators include latency, packet loss rate, and jitter. Active detection is to send test traffic to actively measure the link quality, and passive detection is to analyze the actual service traffic to passively measure the link quality;

[0011] Collect all link quality indicator data, score each link quality, and the link with the highest score is the link with the best quality. The scoring formula is Score = (1 - latency weight * latency ratio) + (1 - jitter weight * jitter ratio) + (1 - packet loss rate weight * packet loss rate ratio). In the game, the weight distribution is 50% for latency, 30% for packet loss rate, and 20% for jitter data;

[0012] Allocate high-priority game traffic to the link with the best quality.

[0013] In one embodiment, the present invention provides an intelligent hybrid link management method based on SD-WAN. When high-priority game traffic is monitored and transmitted, low-priority game traffic is allocated to the link with the second-best quality, and low-priority game traffic includes game update file downloads; non-game traffic is allocated to other links except the best-quality and second-best-quality links. Non-game traffic includes background updates and advertisement pushes. The specific steps include:

[0014] Classify traffic types into high-priority game traffic, low-priority game traffic, and non-game traffic through deep packet inspection (DPI) or port identification; high-priority game traffic includes real-time operation of game character movement, real-time voice, and real-time text, low-priority game traffic includes game update file downloads, and non-game traffic includes background updates and advertisement pushes;

[0015] When it is detected that high-priority game traffic is transmitted through the link with the best quality, allocate low-priority game traffic to the link with the second-best quality for transmission, and allocate non-game traffic to other links except the best-quality and second-best-quality links for transmission (set a first mark on high-priority game traffic, a second mark on low-priority game traffic, and a third mark on non-game traffic; different marks select different links).

[0016] In one embodiment, the present invention provides an intelligent hybrid link management method based on SD-WAN. If the transmission of high-priority game traffic on the link with the best quality does not reach the set value, after implementing bandwidth threshold limits on non-game traffic and low-priority game traffic, releasing the occupied physical link bandwidth resources, and increasing the available bandwidth of the link with the best quality, the following steps are further included:

[0017] Capture voice data packets and text data packets in high-priority game traffic, convert voice data into raw audio, and convert the raw audio into text data;

[0018] Pre-define a list of abusive keywords (such as racial discrimination, swear words), and quickly match the abusive keywords in the text data packets and the text data converted from voice data through regular expressions;

[0019] When the abusive keywords are not successfully matched, further use a pre-trained model (such as BERT or RoBERTa model) to understand the implicit abuse (such as sarcasm, variant spellings) in combination with the context of the text data packets and the text data converted from voice data;

[0020] When it is determined that there are abusive keywords or implicit abuse in the voice data packets or text data packets, after identifying the abusive user, discard their voice data packets and text data packets.

[0021] In one embodiment, the present invention provides an intelligent hybrid link management method based on SD-WAN. When the transmission of high-priority game traffic on the link with the best quality does not reach the set value, after implementing bandwidth threshold restrictions on non-game traffic and low-priority game traffic, releasing the occupied physical link bandwidth resources, and increasing the available bandwidth of the link with the best quality, the following steps are further included:

[0022] When logging in to the game, use a traffic generation tool to create simulated high-priority game traffic, simulated low-priority game traffic, and simulated non-game traffic;

[0023] Conduct a simulated fault test, actively disconnect the link with the best quality, and after a delay, simulate the recovery of the fault and reactivate the link with the best quality;

[0024] During the process of disconnecting and reactivating the link with the best quality, detect whether the simulated high-priority game traffic, simulated low-priority game traffic, and simulated non-game traffic automatically change the link and the time of changing the link, and detect the packet loss rate, and give a feedback prompt when it is abnormal.

[0025] In one embodiment, the present invention provides an intelligent hybrid link management system based on SD-WAN, including:

[0026] A link determination module, used to monitor the quality of each link, determine the link with the best quality, the link quality indicators include delay, packet loss rate, and jitter, and allocate high-priority game traffic to the link with the best quality;

[0027] A link allocation module, which is used to allocate low-priority game traffic to a link with sub-optimal quality when high-priority game traffic transmission is detected. The low-priority game traffic includes game update file downloads; and allocate non-game traffic to other links except the ones with the best and sub-optimal quality. The non-game traffic includes background updates and advertisement pushes.

[0028] A bandwidth allocation module, which is used to impose bandwidth threshold limits on non-game traffic and low-priority game traffic (such as limiting the speed to 1 Mbps and 5 Mbps respectively) when the transmission of high-priority game traffic on the link with the best quality does not reach the set value (the set value can be a delay ≤ 50 ms, a jitter ≤ 10 ms, and a packet loss rate less than 1%), release the occupied physical link bandwidth resources, and increase the available bandwidth of the link with the best quality.

[0029] In one embodiment, the present invention provides an intelligent hybrid link management system based on SD-WAN. The link determination module includes:

[0030] A link quality measurement unit, which is used to obtain link quality index data by combining active detection and passive detection. The link quality indexes include delay, packet loss rate, and jitter. Active detection is to send test traffic to actively measure the link quality, and passive detection is to analyze actual service traffic to passively measure the link quality.

[0031] A link quality scoring unit, which is used to collect all link quality index data, score the quality of each link, and the link with the highest score is the link with the best quality. The scoring formula is Score = (1 - delay weight * delay ratio) + (1 - jitter weight * jitter ratio) + (1 - packet loss rate weight * packet loss rate ratio). In the game, the weight distribution is that the delay accounts for 50%, the packet loss rate accounts for 30%, and the jitter data accounts for 20%.

[0032] An optimal link allocation unit, which is used to allocate high-priority game traffic to the link with the best quality.

[0033] In one embodiment, the present invention provides an intelligent hybrid link management system based on SD-WAN. The link allocation module includes:

[0034] A traffic type identification unit, which is used to classify traffic types into high-priority game traffic, low-priority game traffic, and non-game traffic through deep packet inspection (DPI) or port identification. High-priority game traffic includes real-time operation of game character movement, real-time voice, and real-time text. Low-priority game traffic includes game update file downloads. Non-game traffic includes background updates and advertisement pushes.

[0035] A traffic corresponding link determination unit, which is used to, when detecting that high-priority game traffic is transmitted through the link with the optimal quality, allocate low-priority game traffic to the link with the sub-optimal quality for transmission, and allocate non-game traffic to other links except the links with the optimal and sub-optimal quality (set a first mark on the high-priority game traffic, set a second mark on the low-priority game traffic, and set a third mark on the non-game traffic; different marks select different links).

[0036] In one embodiment, the present invention provides an intelligent hybrid link management system based on SD-WAN, further including:

[0037] A data acquisition module, which is used to capture voice data packets and text data packets in high-priority game traffic, convert the voice data into raw audio, and convert the raw audio into text data;

[0038] An abusive keyword matching module, which is used to pre-define a list of abusive keywords (such as racial discrimination, swear words), and the text data packets and the text data converted from the voice data quickly match the abusive keywords through regular expressions;

[0039] An implicit abuse understanding module, which is used to, when the abusive keywords are not successfully matched, further use a pre-trained model (such as BERT or RoBERTa model), and combine the context of the text data packets and the text data converted from the voice data to understand implicit abuse (such as sarcasm, variant spellings);

[0040] A data packet discarding module, which is used to, when it is determined that there are abusive keywords or implicit abuse in the voice data packets or text data packets, after identifying the abusive user, discard its voice data packets and text data packets.

[0041] In one embodiment, the present invention provides an intelligent hybrid link management system based on SD-WAN, further including:

[0042] A traffic simulation module, which is used to, when a game is logged in, use a traffic generation tool to create simulated high-priority game traffic, simulated low-priority game traffic, and simulated non-game traffic;

[0043] A fault simulation module, which is used to simulate a fault test, actively disconnect the link with the optimal quality, and after a delay, simulate the recovery of the fault and reactivate the link with the optimal quality;

[0044] A link switching detection module, which is used to, during the process of disconnecting and reactivating the link with the optimal quality, detect whether the simulated high-priority game traffic, simulated low-priority game traffic, and simulated non-game traffic automatically change links and the time of changing links, and detect the packet loss rate, and give a feedback prompt when it is abnormal.

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows: When the game is started, the present invention classifies traffic, allocates different links to different types of traffic, preferentially guarantees the operation of the game, and reduces network latency during the game; when it is detected that the transmission of high-priority traffic does not reach the set value, bandwidth threshold limits are imposed on non-game traffic and low-priority game traffic, the bandwidth of high-priority game traffic is increased, and the delay, jitter, and packet loss rate of the game are guaranteed; when there is abuse during the game, the voice and text are automatically blocked to purify the game environment and improve the game experience of players. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 FIG. 6 is a schematic diagram of the first part of the process of an intelligent hybrid link management method based on SD-WAN provided by an embodiment of the present invention.

[0047] Figure 2 FIG. 10 is a schematic diagram of the process of link quality judgment provided by an embodiment of the present invention.

[0048] Figure 3 FIG. 14 is a schematic diagram of the process of allocating traffic to links provided by an embodiment of the present invention.

[0049] Figure 4 FIG. 18 is a schematic diagram of the second part of the process of an intelligent hybrid link management method based on SD-WAN provided by an embodiment of the present invention.

[0050] Figure 5 FIG. 22 is a schematic diagram of the third part of the process of an intelligent hybrid link management method based on SD-WAN provided by an embodiment of the present invention.

[0051] Figure 6 FIG. 26 is a schematic diagram of the first part of an intelligent hybrid link management system based on SD-WAN provided by an embodiment of the present invention.

[0052] Figure 7 FIG. 30 is a schematic diagram of a link determination module provided by an embodiment of the present invention.

[0053] Figure 8 FIG. 34 is a schematic diagram of a link allocation module provided by an embodiment of the present invention.

[0054] Figure 9 FIG. 38 is a schematic diagram of the second part of an intelligent hybrid link management system based on SD-WAN provided by an embodiment of the present invention.

[0055] Figure 10 FIG. 42 is a schematic diagram of the third part of an intelligent hybrid link management system based on SD-WAN provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0056] To make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0057] It can be understood that the terms "first", "second", etc. used in this application may be used herein to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are only used to distinguish one element from another. For example, without departing from the scope of this application, the first xx script may be referred to as the second xx script, and similarly, the second xx script may be referred to as the first xx script.

[0058] In one embodiment, as Figure 1 shown, an intelligent hybrid link management method based on SD-WAN includes the following steps:

[0059] Step S1: Monitor the quality of each link, determine the link with the best quality, and the link quality indicators include delay, packet loss rate, and jitter. Allocate high-priority game traffic to the link with the best quality.

[0060] Step S2: When high-priority game traffic is monitored, allocate low-priority game traffic to the link with the second-best quality. Low-priority game traffic includes game update file downloads; allocate non-game traffic to other links except the best-quality and second-best-quality links. Non-game traffic includes background updates and advertisement pushes.

[0061] Step S3: If the transmission of high-priority game traffic on the link with the best quality does not reach the set value (the set value can be a delay ≤ 50 ms, a jitter ≤ 10 ms, and a packet loss rate less than 1%), impose bandwidth threshold limits on non-game traffic and low-priority game traffic (such as throttling them to 1 Mbps and 5 Mbps respectively), release the occupied physical link bandwidth resources, and increase the available bandwidth of the link with the best quality.

[0062] To ensure the gaming experience of players, high-priority game traffic is allocated to the link with the best quality, and indicators such as the delay (round-trip time of data packets), packet loss rate (proportion of lost data packets), and jitter (amplitude of delay fluctuations) of each link are monitored. For example:

[0063] Link A: Delay 45 ms, packet loss rate 0.5%, jitter 8 ms;

[0064] Link B: Delay 80 ms, packet loss rate 2%, jitter 20 ms;

[0065] Link C: Delay 120 ms, packet loss rate 5%, jitter 50 ms;

[0066] Link D: latency 110ms, packet loss rate 3%, jitter 60ms;

[0067] According to the preset thresholds (such as latency ≤ 50ms, packet loss rate ≤ 1%, jitter ≤ 10ms), the system automatically determines that Link A is "optimal in quality", Link B is "sub - optimal", and Links C and D are "other links". At this time, high - priority game traffic (such as real - time battle commands, character action synchronization, real - time voice, real - time text, etc.) is preferentially allocated to Link A, low - priority game traffic (such as a 3GB version update package of a certain game) is automatically allocated to Link B, and non - game traffic (such as background updates, advertisement pushes) uses Links C and D, etc., ensuring the transmission quality of high - priority game traffic and providing players with a good gaming experience. If the latency of Link A is 52ms, the packet loss rate is 0.8%, and the jitter is 9ms, at this time the latency has exceeded the threshold of 50ms. To ensure the gaming experience of players, bandwidth threshold restrictions are imposed on non - game traffic and low - priority game traffic, releasing the occupied physical link bandwidth resources and increasing the available bandwidth of the link with the optimal quality. For example, during peak hours of a certain game, through this strategy, 80% of the bandwidth is ensured for high - priority game traffic, 15% for low - priority game traffic, and 5% for non - game traffic.

[0068] In one embodiment, as Figure 2 shown, the present invention provides an intelligent hybrid link management method based on SD - WAN. In step S1 of monitoring the quality of each link and determining the link with the optimal quality, the link quality metrics include latency, packet loss rate, and jitter. In the step of allocating high - priority game traffic to the link with the optimal quality, it specifically includes:

[0069] Step S11, combining active probing and passive probing to obtain link quality metric data. The link quality metrics include latency, packet loss rate, and jitter. Active probing is to send test traffic and actively measure the link quality, and passive probing is to analyze actual service traffic and passively measure the link quality;

[0070] Step S12, collecting all link quality metric data, scoring the quality of each link, and the link with the highest score is the link with the optimal quality. The scoring formula is Score=(1 - latency weight * latency ratio)+(1 - jitter weight * jitter ratio)+(1 - packet loss rate weight * packet loss rate ratio). In the game, the weight distribution is 50% for latency, 30% for packet loss rate, and 20% for jitter data;

[0071] Step S13, allocating high - priority game traffic to the link with the optimal quality.

[0072] Active detection: Deploy active detection tools (such as NQA systems or ICMP probes) in the network, set test parameters, select protocols such as ICMP (Ping), TCP / UDP, etc. to send test packets. The test frequency is, for example, to send a set of probe packets every 5 seconds to evaluate the real-time link quality. Clearly define the start and end points of the detection path (such as cross-region servers), generate test data packets (such as ICMP echo request messages), and simulate real traffic; complete data collection and metric calculation. For latency, record the round-trip time (RTT) of the data packets and take the average value; for packet loss rate, count the proportion of probe packets that do not receive responses; for jitter, calculate the standard deviation of the differences in RTT of consecutive data packets.

[0073] Passive detection: Deploy traffic mirroring or deep packet inspection (DPI) tools at key link nodes to capture traffic. For latency calculation: Deduce the one-way latency through TCP handshake timestamps or application layer protocols (such as game heartbeat packets); for packet loss detection: Analyze the TCP retransmission rate or UDP sequence number continuity; for jitter evaluation: Statistically analyze the arrival time fluctuations of consecutive data packets in the service traffic.

[0074] By combining active and passive detection, a comprehensive evaluation and dynamic optimization of network quality can be achieved, providing stable transmission guarantees for high-real-time services such as games.

[0075] In one embodiment, as Figure 3 shown, the present invention provides an intelligent hybrid link management method based on SD-WAN. In step S2, when high-priority game traffic transmission is detected, low-priority game traffic, including game update file downloads, is allocated to the link with sub-optimal quality, and non-game traffic, including background updates and advertisement pushes, is allocated to other links except the ones with optimal and sub-optimal quality. Specifically, it includes:

[0076] Step S21: Classify traffic types into high-priority game traffic, low-priority game traffic, and non-game traffic through deep packet inspection (DPI) or port identification. High-priority game traffic includes real-time operation of game character movement, real-time voice, and real-time text. Low-priority game traffic includes game update file downloads. Non-game traffic includes background updates and advertisement pushes.

[0077] Step S22: When it is detected that high-priority game traffic is transmitted through the link with optimal quality, allocate low-priority game traffic to the link with sub-optimal quality for transmission, and allocate non-game traffic to other links except the ones with optimal and sub-optimal quality for transmission (set a first mark on high-priority game traffic, a second mark on low-priority game traffic, and a third mark on non-game traffic; different marks select different links).

[0078] The classification of traffic types is as shown in Table 1, the traffic classification table:

[0079] Traffic type Identification feature Priority policy High-priority game traffic The message length of the real-time interaction protocol (such as the RTP protocol of Riot Games) is ≤ 200 bytes and the sending interval is ≤ 20 ms DSCP mark EF (the first mark) Low-priority game traffic The HTTP Range request feature of the large file transfer mode (such as the resource pack download of League of Legends) AF21 mark (the second mark) Non-game traffic The video streaming media TS shard transmission mode of the advertisement SDK feature code (such as the API path of Google AdMob) BE mark (the third mark)

[0080] ‌

[0081] In one embodiment, as Figure 4 shown, the present invention provides an intelligent hybrid link management method based on SD-WAN. After the step S3, when the transmission of high-priority game traffic on the link with the best quality does not reach the set value, bandwidth threshold limits are imposed on non-game traffic and low-priority game traffic, the occupied physical link bandwidth resources are released, and the available bandwidth of the link with the best quality is increased. The following steps are further included:

[0082] Step S4, capture the voice data packets and text data packets in the high-priority game traffic, convert the voice data into raw audio, and convert the raw audio into text data;

[0083] Step S5, predefined a list of abusive keywords (such as racial discrimination, swear words), and the text data packets and the text data converted from the voice data are quickly matched with the abusive keywords through regular expressions;

[0084] Step S6, when the abusive keywords are not successfully matched, further use a pre-trained model (such as BERT or RoBERTa model), and combine the context of the text data packets and the text data converted from the voice data to understand implicit abuse (such as sarcasm, variant spellings);

[0085] Step S7, when it is determined that there are abusive keywords or implicit abuse in the voice data packets or text data packets, after identifying the abusive user, discard its voice data packets and text data packets.

[0086] Extract Mel-frequency cepstral coefficients (MFCC) or spectral features from the audio signal to characterize the acoustic characteristics of the voice; call a pre-trained speech recognition model (such as Baidu Speech API, Tencent Cloud ASR, or CMU Sphinx), match the audio features with the acoustic model and the language model‌; generate a preliminary text result according to the model output, and usually save it as a text file encoded in UTF-8 (such as TXT, SRT).

[0087] In one embodiment, as Figure 5 shown, the present invention provides an intelligent hybrid link management method based on SD-WAN. After the step S3, when the transmission of high-priority game traffic on the link with the best quality does not reach the set value, bandwidth threshold limits are imposed on non-game traffic and low-priority game traffic, the occupied physical link bandwidth resources are released, and the available bandwidth of the link with the best quality is increased. The following steps are further included:

[0088] Step S8, when logging in to the game, use a traffic generation tool to create simulated high-priority game traffic, simulated low-priority game traffic, and simulated non-game traffic;

[0089] Step S9, perform a simulated fault test. Actively disconnect the link with the best quality. After a delay, simulate the recovery from the fault and reactivate the link with the best quality;

[0090] Step S10, during the process of disconnecting and reactivating the link with the best quality, detect whether the simulated high-priority game traffic, simulated low-priority game traffic, and simulated non-game traffic automatically switch links and the time for switching links, and detect the packet loss rate. Provide a feedback prompt when an abnormality occurs.

[0091] Taking MOBA games as an example, according to the real-time interaction requirements of League of Legends and DOTA 2, the network needs to meet the following key indicators:

[0092] The switching time of high-priority game traffic needs to complete link switching within ≤ 50 ms to ensure that operations such as character control and skill release have no perceivable delay. Detection can be carried out by simulating a link interruption with an IXIA tester to trigger the route request (RREQ) and reply (RREP) of a routing protocol (such as AODV), and recording the time window from link disconnection to new link activation.

[0093] The packet loss rate of high-priority game traffic needs to be ≤ 1%, otherwise it will cause fatal problems such as character teleportation and skill delay. Detection can inject traffic with a marked DSCP value of 46 (accelerated forwarding class) in an IXIA tester and count the number of lost packets during the switching period.

[0094] In one embodiment, as Figure 6 shown, the present invention provides an intelligent hybrid link management system based on SD-WAN, including:

[0095] A link determination module 1, which is used to monitor the quality of each link, determine the link with the best quality. The link quality indicators include delay, packet loss rate, and jitter, and allocate high-priority game traffic to the link with the best quality;

[0096] A link allocation module 2, which is used to allocate low-priority game traffic to the link with the second-best quality when high-priority game traffic is detected in transmission. The low-priority game traffic includes game update file downloads; allocate non-game traffic to other links except the links with the best and second-best quality. The non-game traffic includes background updates and advertisement pushes;

[0097] The bandwidth allocation module 3 is used to impose bandwidth threshold limits on non-game traffic and low-priority game traffic (such as limiting the speed to 1 Mbps and 5 Mbps respectively) when the transmission of high-priority game traffic on the link with the best quality does not reach the set value (the set value can be a delay ≤ 50 ms, jitter ≤ 10 ms, and packet loss rate less than 1%), release the occupied physical link bandwidth resources, and increase the available bandwidth of the link with the best quality.

[0098] The delay of Link A (the link with the best quality) is 52 ms, the packet loss rate is 0.8%, and the jitter is 9 ms, which does not reach the set value (delay ≤ 50 ms, jitter ≤ 10 ms, and packet loss rate less than 1%). A speed limit instruction is sent through the SD-WAN controller to limit the speed of non-game traffic to 1 Mbps and the speed of low-priority game traffic to 5 Mbps; the released bandwidth resources after speed limiting (such as releasing 9 Mbps from low-priority game traffic and non-game traffic) are automatically allocated for Link A to use; the available bandwidth value of Link A is updated through a dynamic routing protocol (such as BGP) to ensure that high-priority game traffic obtains additional resources.

[0099] In one embodiment, as Figure 7 shown, the present invention provides an intelligent hybrid link management system based on SD-WAN. The link determination module 1 includes:

[0100] The link quality measurement unit 11 is used to obtain link quality index data by combining active detection and passive detection. The link quality indexes include delay, packet loss rate, and jitter. Active detection is to send test traffic to actively measure the link quality, and passive detection is to analyze the actual service traffic to passively measure the link quality.

[0101] The link quality scoring unit 12 is used to collect all link quality index data, score the quality of each link, and the link with the highest score is the link with the best quality. The scoring formula is Score = (1 - delay weight * delay ratio) + (1 - jitter weight * jitter ratio) + (1 - packet loss rate weight * packet loss rate ratio). In the game, the weight distribution is 50% for delay, 30% for packet loss rate, and 20% for jitter data.

[0102] The optimal link allocation unit 13 is used to allocate high-priority game traffic to the link with the best quality.

[0103] The weight allocation logic is adapted according to the game traffic characteristics. Delay takes precedence (50%): For real-time operation instructions (such as the movement of MOBA game characters), the end-to-end delay needs to be ensured to be ≤50 ms. Excessive delay will cause the operation to be out of sync with the screen. Packet loss rate comes second (30%): The TCP retransmission mechanism is sensitive to packet loss, but the UDP protocol (commonly used in game voice / video) allows partial packet loss, and the reliability and real-time performance need to be balanced. Jitter weight is the lowest (20%): The impact of jitter can be mitigated through buffering techniques (such as FIFO queues), but there are still strict requirements for audio / video synchronization (such as jitter ≤20 ms). For traffic such as video conferencing and file downloads, the requirements for delay, packet loss rate, and jitter are different from those of games.

[0104] In one embodiment, as Figure 8 shown, the present invention provides an intelligent hybrid link management system based on SD-WAN. The link allocation module 2 includes:

[0105] A traffic type recognition unit 21, configured to classify traffic types into high-priority game traffic, low-priority game traffic, and non-game traffic through deep packet inspection (DPI) or port recognition; high-priority game traffic includes real-time operation of game character movement, real-time voice, and real-time text, low-priority game traffic includes game update file downloads, and non-game traffic includes background updates and advertisement pushes;

[0106] A traffic corresponding link determination unit 22, configured to, when detecting that high-priority game traffic is transmitted through the link with the best quality, allocate low-priority game traffic to the link with the second-best quality for transmission, and allocate non-game traffic to other links except the best-quality and second-best-quality links (set a first mark on high-priority game traffic, set a second mark on low-priority game traffic, and set a third mark on non-game traffic; different marks select different links).

[0107] Port recognition. Standard ports quickly classify traffic types based on the IANA registered port library (such as 80 = HTTP, 443 = HTTPS, 27015 = Steam game). For non-standard ports (such as P2P applications with random high ports), DPI needs to be combined for secondary verification, and the TCP / UDP port range (such as UDP 27015 - 27030 is common in games) is used to narrow the detection range.

[0108] Deep Packet Inspection (DPI) technology: 1. Protocol feature parsing, which parses the application layer protocol headers (such as HTTP request paths, DNS query domain names) and specific patterns in the payloads (such as video stream TS shard features); 2. Behavioral pattern analysis, which detects high-frequency short packets (such as MOBA game operation instructions) or long connections (such as real-time video conference transmissions), and combines traffic statistical features (such as packet size distribution, transmission interval) to distinguish real-time services from download traffic for analysis.

[0109] In one embodiment, as Figure 9 shown, the present invention provides an intelligent hybrid link management system based on SD-WAN, further including:

[0110] A data acquisition module 4, which is used to capture voice data packets and text data packets in high-priority game traffic, convert the voice data into raw audio, and convert the raw audio into text data;

[0111] An abusive keyword matching module 5, which is used to pre-define a list of abusive keywords (such as racial discrimination, swear words), and the text data packets and the text data converted from the voice data quickly match the abusive keywords through regular expressions;

[0112] An implicit abuse understanding module 6, which is used when the abusive keywords are not successfully matched, and further uses a pre-trained model (such as BERT or RoBERTa model), combined with the context of the text data packets and the text data converted from the voice data to understand implicit abuse (such as sarcasm, variant spellings);

[0113] A data packet discarding module 7, which is used to determine that when there are abusive keywords or implicit abuse in the voice data packets or text data packets, after identifying the abusive user, discard its voice data packets and text data packets.

[0114] The specific implementation process of identifying the abusive user is as follows:

[0115] Unify the encoding (UTF-8) of the text data packets and the text data generated by voice recognition, and add source tags to distinguish the data modalities. Clean the noise data, filter ASR errors, and non-verbal symbols (such as garbled emojis).

[0116] Construct a labeled data set containing explicit abuse (such as "stupid *") and implicit abuse (such as ironic "You are so smart"), and it is necessary to label single sentences and context windows (such as the first 3 sentences + the last 3 sentences).

[0117] Select models, RoBERTa model: Adopt a dynamic masking mechanism, which has a stronger ability to capture long text context and is suitable for detecting context-dependent implicit abuse. BERT model: Parse the relationship between words through a bidirectional attention mechanism to identify variant abuse.

[0118] Modify the output layer, add a BiLSTM / BiGRU layer on top of the pre-trained model to enhance the sequence feature extraction ability; for multi-task learning, jointly train the abuse classification (main task) and sentiment polarity judgment (auxiliary task) to improve the model's sensitivity to negative emotions.

[0119] Segment long texts by a fixed window (such as 512 tokens), retain the context association for the overlapping parts to avoid information fragmentation, extract the attention matrix of the Transformer layer to locate sensitive words and their modifiers; use the confidence score output by ASR as an additional feature to input into the model to assist in judging semantic deviations caused by recognition errors.

[0120] Add common ASR error samples (such as homophone substitution) to the training data to improve the model's robustness to speech-to-text noise; use model distillation techniques (such as DistilBERT) to compress the model size to meet the real-time detection requirements (real-time game chat scenario).

[0121] In the first layer, use regular expressions to quickly filter explicit abuse; in the second layer, use the pre-trained model to deeply analyze implicit abuse to reduce the consumption of computing resources; screen boundary samples through the model prediction confidence (such as 0.4 < confidence < 0.6), and add them to the training set after manual review; monitor emerging abuse variants, regularly update the fine-tuning dataset and model parameters to determine abuse keywords or implicit abuse.

[0122] In one embodiment, as Figure 10 shown, the present invention provides an intelligent hybrid link management system based on SD-WAN, further including:

[0123] A traffic simulation module 8, used to create simulated high-priority game traffic, simulated low-priority game traffic, and simulated non-game traffic using a traffic generation tool when logging in to the game.

[0124] A fault simulation module 9, used to simulate fault testing, actively disconnect the link with the best quality, and after a delay, simulate fault recovery and reactivate the link with the best quality.

[0125] A switched link detection module 10, used to detect whether the simulated high-priority game traffic, simulated low-priority game traffic, and simulated non-game traffic automatically change the link and the time of changing the link during the process of disconnecting and reactivating the link with the best quality, and detect the packet loss rate, and give a feedback prompt when abnormal.

[0126] When the time for simulating the switching link of high-priority game traffic > 50 ms or the packet loss rate > 1%, the system needs to push an alarm through logs or APIs for timely handling to ensure the gaming experience of players. The requirements for the time and packet loss rate of simulating the switching links of low-priority game traffic and non-game traffic are relatively low, for example, the time ≤ 50 ms or the packet loss rate ≤ 5%.

[0127] It should be understood that although the steps in the flowcharts of the embodiments of the present invention are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise clearly stated in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in each embodiment may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same moment, but can be executed at different moments. The execution order of these sub-steps or stages is not necessarily sequential either, but can be executed alternately or in turn with at least a part of other steps or sub-steps or stages of other steps.

[0128] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0129] The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as falling within the scope described in this specification.

[0130] The above-described embodiments merely represent several implementation manners of the present invention. The description thereof is relatively specific and detailed, but it should not be construed as a limitation to the scope of the patent of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the appended claims.

[0131] The above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

[0132] In addition, it should be understood that although this specification is described in terms of embodiments, not every embodiment only contains an independent technical solution. This narrative manner of the specification is only for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. An intelligent hybrid link management method based on SD-WAN, characterized in that, The intelligent hybrid link management method based on SD-WAN includes the following steps: Monitor the quality of each link, determine the link with the best quality. The link quality indicators include latency, packet loss rate, and jitter. Allocate high-priority game traffic to the link with the best quality; When high-priority game traffic is detected, allocate low-priority game traffic to the link with the second-best quality. Low-priority game traffic includes game update file downloads; allocate non-game traffic to other links except the ones with the best and second-best quality. Non-game traffic includes background updates and advertisement pushes; If the transmission of high-priority game traffic on the link with the best quality does not reach the set value, impose bandwidth threshold limits on non-game traffic and low-priority game traffic, release the occupied physical link bandwidth resources, and increase the available bandwidth of the link with the best quality.

2. The intelligent hybrid link management method based on SD-WAN according to claim 1, wherein, In the step of monitoring the quality of each link, determining the link with the best quality, where the link quality indicators include latency, packet loss rate, and jitter, and allocating high-priority game traffic to the link with the best quality, it specifically includes: Combining active detection and passive detection to obtain link quality indicator data. The link quality indicators include latency, packet loss rate, and jitter. Active detection is to send test traffic and actively measure the link quality. Passive detection is to analyze actual service traffic and passively measure the link quality; Collect all link quality indicator data, score the quality of each link. The link with the highest score is the link with the best quality. The scoring formula is Score=(1 - latency weight * latency ratio)+(1 - jitter weight * jitter ratio)+(1 - packet loss rate weight * packet loss rate ratio). In the game, the weight distribution is 50% for latency, 30% for packet loss rate, and 20% for jitter data; Allocate high-priority game traffic to the link with the best quality.

3. The intelligent hybrid link management method based on SD-WAN according to claim 1, wherein, In the step of when high-priority game traffic is detected, allocate low-priority game traffic to the link with the second-best quality. Low-priority game traffic includes game update file downloads; allocate non-game traffic to other links except the ones with the best and second-best quality. Non-game traffic includes background updates and advertisement pushes, it specifically includes: Classify traffic types into high-priority game traffic, low-priority game traffic, and non-game traffic through deep packet inspection or port identification. High-priority game traffic includes real-time operation of game character movement, real-time voice, and real-time text. Low-priority game traffic includes game update file downloads. Non-game traffic includes background updates and advertisement pushes; When it is detected that high-priority game traffic is transmitted through the link with the best quality, allocate low-priority game traffic to the link with the second-best quality for transmission, and allocate non-game traffic to other links except the ones with the best and second-best quality for transmission.

4. The intelligent hybrid link management method based on SD-WAN according to any one of claims 1 to 3, characterized in that, After the step of if the transmission of high-priority game traffic on the link with the best quality does not reach the set value, impose bandwidth threshold limits on non-game traffic and low-priority game traffic, release the occupied physical link bandwidth resources, and increase the available bandwidth of the link with the best quality, the following steps are also included: Capture voice data packets and text data packets in high-priority game traffic, convert the voice data into raw audio, and convert the raw audio into text data; Pre-define a list of abusive keywords, and quickly match the abusive keywords in the text data packet and the text data converted from the voice data through regular expressions; When the abusive keywords are not successfully matched, further use a pre-trained model to understand the implied abuse in combination with the context of the text data packet and the text data converted from the voice data; When it is determined that there are abusive keywords or implied abuse in the voice data packet or text data packet, after identifying the abusive user, discard their voice data packet and text data packet.

5. The intelligent hybrid link management method based on SD-WAN according to any one of claims 1 to 3, characterized in that After the step of implementing bandwidth threshold limit on non-game traffic and low-priority game traffic when the high-priority game traffic does not reach the set value when transmitted on the link with the best quality, release the occupied physical link bandwidth resources, and increase the available bandwidth of the link with the best quality, the following steps are also included: When logging in to the game, use a traffic generation tool to create simulated high-priority game traffic, simulated low-priority game traffic, and simulated non-game traffic; Conduct simulated fault testing, actively disconnect the link with the best quality, and after a delay, simulate fault recovery and reactivate the link with the best quality; During the process of disconnecting and reactivating the link with the best quality, detect whether the simulated high-priority game traffic, simulated low-priority game traffic, and simulated non-game traffic automatically change the link and the time of changing the link, and detect the packet loss rate, and give a feedback prompt when abnormal.

6. An intelligent hybrid link management system based on SD-WAN, characterized in that, Include: A link determination module, which is used to monitor the quality of each link, judge the link with the best quality, the link quality indicators include delay, packet loss rate, and jitter, and allocate high-priority game traffic to the link with the best quality; A link allocation module, which is used to allocate low-priority game traffic to the link with the second-best quality when high-priority game traffic is detected, and the low-priority game traffic includes game update file downloads; allocate non-game traffic to other links except the best-quality and second-best-quality links, and the non-game traffic includes background updates and advertisement pushes; A bandwidth allocation module, which is used to implement bandwidth threshold limit on non-game traffic and low-priority game traffic when the high-priority game traffic does not reach the set value when transmitted on the link with the best quality, release the occupied physical link bandwidth resources, and increase the available bandwidth of the link with the best quality.

7. The intelligent hybrid link management system based on SD-WAN according to claim 6, characterized in that The link determination module includes: A link quality measurement unit, which is used to obtain link quality indicator data by combining active detection and passive detection. The link quality indicators include delay, packet loss rate, and jitter. Active detection is to send test traffic and actively measure the link quality, and passive detection is to analyze the actual service traffic and passively measure the link quality; A link quality scoring unit, which is used to collect all link quality indicator data, score the quality of each link, and the link with the highest score is the link with the best quality. The scoring formula is Score=(1 - delay weight * delay ratio)+(1 - jitter weight * jitter ratio)+(1 - packet loss rate weight * packet loss rate ratio). In the game, the weight distribution is 50% for delay, 30% for packet loss rate, and 20% for jitter data; An optimal link allocation unit for allocating high-priority game traffic to the link with the optimal quality.

8. The intelligent hybrid link management system based on SD-WAN according to claim 6, wherein The link allocation module includes: A traffic type identification unit for classifying traffic types into high-priority game traffic, low-priority game traffic, and non-game traffic through deep packet detection or port identification; high-priority game traffic includes real-time operation of game character movement, real-time voice, and real-time text, low-priority game traffic includes game update file downloads, and non-game traffic includes background updates and advertisement pushes; A traffic corresponding link determination unit for, when detecting that high-priority game traffic is transmitted through the link with the optimal quality, allocating low-priority game traffic to the link with the sub-optimal quality for transmission and allocating non-game traffic to other links other than the optimal and sub-optimal quality links for transmission.

9. The intelligent hybrid link management system based on SD-WAN according to any one of claims 6 to 8, characterized in that, The SD-WAN-based intelligent hybrid link management system further includes: A data acquisition module for capturing voice data packets and text data packets in high-priority game traffic, converting voice data into raw audio, and converting the raw audio into text data; An abusive keyword matching module for pre-defining a list of abusive keywords and quickly matching the text data packets and the text data converted from voice data with the abusive keywords through regular expressions; An implicit abuse understanding module for, when the abusive keywords are not successfully matched, further using a pre-trained model to understand implicit abuse in combination with the context of the text data packets and the text data converted from voice data; A data packet discarding module for, when it is determined that there are abusive keywords or implicit abuse in the voice data packets or text data packets, identifying the abusive user and then discarding their voice data packets and text data packets.

10. The intelligent hybrid link management system based on SD-WAN according to any one of claims 6 to 8, characterized in that, The SD-WAN-based intelligent hybrid link management system further includes: A traffic simulation module for, when a game logs in, using a traffic generation tool to create simulated high-priority game traffic, simulated low-priority game traffic, and simulated non-game traffic; A fault simulation module for simulating a fault test, actively disconnecting the link with the optimal quality, and after a delay, simulating fault recovery and reactivating the link with the optimal quality; A switched link detection module for, during the process of disconnecting and reactivating the link with the optimal quality, detecting whether the simulated high-priority game traffic, simulated low-priority game traffic, and simulated non-game traffic automatically change links and the time of changing links, and detecting the packet loss rate, and giving a feedback prompt when abnormal.

Images