A method and system for application layer target state awareness and intelligent offloading

Abstract

This invention discloses a method and system for application-layer target state perception and intelligent traffic splitting. The method includes: constructing a reachability state mapping for each target device at the application layer and jointly evaluating the target communication state; performing multi-level traffic splitting control on the data stream corresponding to each target based on the reachability state mapping, switching transmission strategies between different levels to prioritize continuous transmission of active targets when reachability decreases; proactively maintaining the address resolution mapping cache of target devices before data transmission and / or according to a preset strategy; dynamically coordinating the priority and bandwidth of targets based on the reachability state mapping when transmission queue resource occupancy exceeds a threshold, reducing the bandwidth proportion of unreachable or suspected unreachable targets; and performing timing smoothing and jitter tolerance processing on the reachability determination results. This invention aims to solve the technical problem of poor continuity and real-time performance in audio transmission.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of network audio transmission, in particular to a method and system for application layer target state awareness and intelligent flow splitting. BACKGROUND

[0002] In a distributed network audio system, for the consideration of resources and real-time performance, UDP is often used to carry and share the same socket by multiple audio streams; however, such implementation generally relies on the passive reachability maintenance of ARP / neighbor table of the operating system, once the receiving end is offline or the network is abnormal, the misalignment of ARP aging and sending timeout will cause a continuous window: the ARP entry has become obsolete but is not identified in time, the sender still sends packets and triggers ARP probes until its own sending timeout, during which the socket sending buffer is occupied by invalid data, and valid audio cannot be enqueued, thus causing the receiving end to play and the sending side to waste resources, and the broadcast occupation and table entry oscillation caused by a large number of ARP requests further aggravate the network and kernel burden; traditional remedies are to adjust ARP parameters or shorten the sending timeout, but the former may damage the existing optimized configuration of the device, and the latter is easy to misjudge when the network is temporarily fluctuating, thus reducing the system robustness and service continuity. SUMMARY

[0003] The purpose of the embodiments of the present application is to provide a method and system for application layer target state awareness and intelligent flow splitting, to solve the technical problem of poor continuity and real-time performance of audio transmission.

[0004] To solve the above technical problems, the embodiments of the present application provide a method for application layer target state awareness and intelligent flow splitting, which adopts the following technical solution:

[0005] A method for application layer target state awareness and intelligent flow splitting, applied to distributed network audio transmission, comprising the following steps:

[0006] A reachability state mapping is constructed for each target device at the application layer, and a joint evaluation is performed on the target communication state, the joint evaluation being based on state indicators from multiple sources;

[0007] According to the reachability state mapping, multi-level flow control is performed on the data stream corresponding to each target, and the sending strategy is switched between different levels, so as to preferentially guarantee the continuous transmission of active targets when the reachability is reduced;

[0008] Before data sending and / or according to a preset strategy, the address resolution mapping cache of the target device is actively maintained, so as to reduce the sending blockage caused by cache aging or failure;

[0009] When the sending queue resource occupation exceeds a threshold, dynamic coordination of priority and bandwidth is implemented for the target according to the reachability state mapping, and the bandwidth occupation ratio of unreachable or suspected unreachable target is reduced;

[0010] Timing smoothing and jitter tolerance processing is performed on the reachability determination result, so as to reduce the influence of short-time network fluctuation on the splitting decision.

[0011] In a possible implementation, the joint evaluation is based on at least two types of state indicators of different sources, and the state indicators are selected from link layer / network layer activity signals, active probe feedback and service layer quality feedback.

[0012] In a possible implementation, the multi-level splitting control includes at least two levels, and the measures are selected from the following:

[0013] The sending priority of the target data flow is lowered, and the sending is mainly in the form of probe;

[0014] The information amount or code rate of the target data flow is degraded to reduce the bandwidth occupation;

[0015] When it is determined that the target is unreachable in a continuous period, the sending of service data of the target is suspended, and periodic probe is reserved.

[0016] In a possible implementation, the multi-source state indicators include:

[0017] Active probe feedback, a connectionless probe carrying a time identifier is sent to the target device, and link abnormality is determined based on the number of consecutive non-responses, and the probe data packet related to the target is sent in a low-priority queue and with a probe identifier;

[0018] Link layer / network layer activity signal, a neighbor / address resolution table event interface of an operating system is listened to to identify the addition, invalidation or change state of a table item corresponding to the target device;

[0019] Service layer quality feedback, real-time transport control packets are analyzed to obtain indicators including at least packet loss rate, round-trip delay and delay jitter, and a target communication quality score is generated according to the indicators.

[0020] In a possible implementation, the multi-level splitting control includes:

[0021] When the reachability is reduced, the target data packet is sent in a low-priority queue as a probe packet;

[0022] When there is no valid response to continuous probe, code rate degradation is performed on the data flow of the target to send narrow-band placeholder / silence audio, and the original code rate is restored when the reachability score is restored to a recovery threshold;

[0023] When there is still no effective recovery within the continuous period, the target is added to the black hole list and the transmission of business data is suspended, while periodic probes are reserved at a fixed or adaptive frequency, and the entry condition, keep-alive frequency and exit condition of the black hole list are set.

[0024] In a possible implementation, the actively maintaining the address resolution mapping cache of the target device comprises:

[0025] Before sending data to the target device, and / or sending a gratuitous address resolution packet or a neighbor advertisement periodically to refresh or extend the validity of the target entry, and setting an upper limit for the sending frequency of such packets, and setting a priority maintenance strategy for key targets.

[0026] In a possible implementation, when the sending queue resource occupancy exceeds a threshold, the step of implementing dynamic coordination of priority and bandwidth for the target according to the reachability state mapping, and reducing the bandwidth occupancy ratio of the unreachable or suspected unreachable target, triggers priority rearrangement and bandwidth quota adjustment when the sending queue utilization rate reaches a preset percentage threshold, and reduces the bandwidth occupancy ratio of the target in the black hole list to ensure the transmission continuity of the target in the reachable state.

[0027] In a possible implementation, in the step of performing timing smoothing and jitter tolerance processing on the reachability determination result to reduce the influence of short-term network fluctuations on the splitting decision, the timing smoothing updates the reachability score using exponential weighted average, and a jitter tolerance time window is set to avoid misjudgment caused by short-term probe interruption.

[0028] In a possible implementation, the message authentication code integrity check is performed on the probe or its confirmation feedback to prevent misjudgment caused by a fake probe, and the sending frequency of the packet for actively maintaining the address resolution mapping cache is set to not exceed a preset upper limit to reduce the risk of being misjudged as an attack.

[0029] To solve the above technical problems, the embodiment of the application further provides an application layer target state awareness and intelligent splitting system, which adopts the technical scheme as follows:

[0030] An application layer target state awareness and intelligent splitting system comprises:

[0031] A state mapping module is configured to construct a reachability state mapping for each target device at the application layer, and jointly evaluate the target communication state, wherein the joint evaluation is based on state indicators from multiple sources.

[0032] A splitting control module is configured to perform multi-level splitting control on data streams corresponding to each target according to the reachability state mapping, and switch the sending strategy between different levels to preferentially guarantee the continuous transmission of active targets when the reachability is reduced.

[0033] an address resolution maintenance module, configured to actively maintain the address resolution mapping cache of the target device before data transmission and / or according to a preset strategy, so as to reduce the transmission blockage caused by cache aging or invalidation;

[0034] a congestion coordination module, configured to perform dynamic coordination of priority and bandwidth on the target according to the reachability state mapping when the resource occupation of the transmission queue exceeds a threshold, so as to reduce the bandwidth proportion of the unreachable or suspected unreachable target;

[0035] a robust decision module, configured to perform timing smoothing and jitter tolerance processing on the reachability determination result, so as to reduce the influence of short-time network fluctuation on the splitting decision.

[0036] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0037] The application layer target state awareness and intelligent splitting method disclosed in the present application can early identify the reachability decline through the front-end with multiple source evidences, pre-load and release the buffer / bandwidth through the middle section with hierarchical splitting, avoid retry blockage caused by ARP aging misplacement through active cache maintenance of the side road, and at the same time, tilt the resources to the active flow through congestion coordination of the rear section, and inhibit frequent switching through smoothing / jitter tolerance, so as to maintain the continuity and low latency of the end-to-end media link without changing the underlying protocol parameters, thereby guaranteeing the continuity and real-time performance of the audio transmission. BRIEF DESCRIPTION OF DRAWINGS

[0038] In order to more clearly illustrate the schemes in the present application, the drawings needed in the description of the embodiments of the present application will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0039] Figure 1 is a flow chart of an embodiment of the application layer target state awareness and intelligent splitting method according to the present application;

[0040] Figure 2 is a structural schematic diagram of an embodiment of the application layer target state awareness and intelligent splitting system according to the present application. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical scheme and advantages of the present application more clear, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0042] REFERENCE Figure 1, shows a flow chart of one embodiment of the application layer target state awareness and intelligent offloading method according to the application. The application layer target state awareness and intelligent offloading method comprises the following steps:

[0043] Step S101, constructing reachability state mapping for each target device at the application layer, jointly evaluating the target communication state, and the joint evaluation is based on state indicators from multiple sources.

[0044] The method in this embodiment runs in a distributed network audio transmission scenario. The application layer first establishes reachability state mapping for each target device (such as the receiving end IP / session ID). When implemented, a "target state table" can be maintained in memory, recording the latest service layer feedback timestamp, the latest several active probe round-trip delay and loss, the active state of the neighbor / ARP table item, the impact degree of the sending queue on the target, and a "reachability score" updated over time for each target. The information sources are diverse: first, active probe feedback, which can be achieved by periodically sending UDP lightweight probes carrying time identifiers and random nonces to the target, recording ACK or echo times to estimate RTT, and accumulating the number of consecutive non-responses; second, link layer / network layer activity signals, which can be achieved by subscribing to operating system neighbor / ARP event interfaces (such as Linux netlink / ndisc) to perceive the addition, change and invalidation of target MAC mapping, thereby judging whether address resolution is in an available state; third, service layer quality feedback, which can be achieved by analyzing the ReceiverReport returned by the receiving end if the media bearer uses RTP / RTCP, and extracting packet loss rate, jitter, round-trip delay and other indicators as side notes of communication quality. To achieve "joint evaluation", the above indicators can be normalized to the [0, 1] interval first, and then fused according to pre-set or adaptive weights to obtain a "reachability score"; when some indicators are temporarily missing, the method of "using the last valid value and applying time decay" can be used to avoid misjudgment caused by temporary blank, and the score is updated every time new evidence arrives, and the timestamp and evidence source are retained for subsequent decision tracing.

[0045] Step S102, performing multi-level offloading control on the data stream corresponding to each target according to the reachability state mapping, switching the sending strategy between different levels, to preferentially guarantee the continuous transmission of active targets when the reachability decreases.

[0046] In this embodiment, after the state mapping is completed, the application layer performs multi-level shunt control on the data stream of each target according to the score and evidence type, so as to reduce the occupation of resources in advance when the target accessibility decreases, give way to other active targets, and maintain the overall audio continuity as much as possible. Four working levels can be set in engineering: normal level (both the sending path and the code rate are the business default configuration), probe priority level (the business load is reduced to secondary, and the probe packet is sent intermittently in the low priority queue to verify whether the target is restored), degraded transmission level (the coding parameters are reduced to narrowband / silence placeholder, typically from 64-128 kbps to 8-16 kbps, or comfort noise / silence frame is sent to ensure that the rendering at the receiving end does not interrupt but the bandwidth is very low), and black hole keep-alive level (suspend the business load and only keep the low-frequency probe to release the socket and link resources). The switching between levels is based on explicit entry and exit conditions, for example: when the probe times out continuously for N times and the neighbor / ARP reports that it is invalid, the probe priority level is switched from the normal level; if it is still not restored within the Tdowngrade time window, the degraded transmission level is entered; if there is still no positive evidence for Tblackhole, the black hole keep-alive level is entered. In order to avoid “back and forth jitter”, it is recommended to introduce hysteresis: the threshold when recovering is higher than that when degrading, and combined with “evidence freshness” (for example, requiring at least K positive evidences within the last M seconds). The shunt control can be implemented through multi-queue scheduling (putting packets into queues according to priority), configuring token bucket / weighted round robin (DRR / WFQ) according to destination, or dynamically adjusting the bit rate on the codec side; the probe packet can be marked in the custom header, so that the receiving end can selectively echo or respond with an application ACK, forming a light closed loop.

[0047] Step S103, before data transmission and / or according to a preset strategy, actively maintaining the address resolution mapping cache of the target device to reduce the transmission blockage caused by cache aging or invalidation.

[0048] In the embodiment, to reduce passive waiting caused by address resolution aging misplacement, the method actively maintains address resolution mapping cache of the target device before sending service data and / or according to a predetermined strategy. Typical implementation is to use gratuitous ARP for IPv4 and neighbor advertisement for IPv6, broadcast / unicast within a reasonable upper limit of frequency to the local area network, refresh exchange / host side cache, and reduce repeated resolution and queue retention triggered by premature expiration of ARP entries. For "key targets" (such as main speakers / conference mixing ends), a more aggressive maintenance period can be set; at the same time, frequency limiting and backoff strategies are used to ensure that the mechanism can improve cache availability without being misjudged as abnormal flooding by network devices. It should be emphasized that this maintenance is completely initiated at the application layer and does not require modification of operating system kernel parameters or system ARP aging strategy, thereby maintaining non-intrusiveness to existing deployments.

[0049] Step S104, when the sending queue resource occupation exceeds the threshold, the dynamic coordination of priority and bandwidth is implemented for the target according to the reachability state mapping, and the bandwidth proportion of unreachable or suspected unreachable target is reduced.

[0050] In the embodiment, from the perspective of the whole machine resource, when it is detected that the sending queue resource occupation exceeds the threshold (for example, estimated according to the proportion of SO_SNDBUF used / available or the depth of the sending ring queue), the method triggers congestion-aware dynamic coordination of priority and bandwidth. Typical practice is: calculate the reachability score and level state of each target, immediately down-regulate the weight or freeze the service sending quota of the target in black hole keep-alive or suspected unreachable; for targets with high score and high service criticality, the weight or minimum bandwidth guarantee is improved; weighted fair queue, token bucket + additional "anemia protection" logic, or quota redistribution within the scheduling period can be used at the system level. The coordination frequency should not be too high, and small step adjustment is usually done in tens to hundreds of milliseconds as the period; at the same time, adjustment logs and statistics are recorded to smoothly return the resources when recovery.

[0051] Step S105, time sequence smoothing and jitter tolerance processing are performed on the reachability determination result to reduce the influence of short-term network fluctuations on the split decision.

[0052] In the embodiment, to suppress the interference of network short-time fluctuation on the judgment, the method performs time smoothing and jitter tolerance processing on the reachability judgment result. For example, the exponential weighted moving average (EMA) can be used to update the score: S_t = a x_t + (1-a) S_{t-1}, where x_t is the instant score at the current time, and a can be 0.5-0.8 to bias the new evidence; at the same time, a jitter tolerance window W (for example, 1-3 seconds) is set, sporadic probe loss or transient RTT peak within W is not directly triggered condition for degradation or entering black hole, but requires continuous duration and sufficient evidence before performing degradation; similarly, K consecutive positive evidence and score exceeding the recovery threshold are required for recovery, and such setting can significantly reduce level oscillation and maintain the user side “continuous sound and stable experience”.

[0053] In summary, the workflow of the method of the embodiment can be as follows: when the target is normal, the business is given priority, and the probe is supplemented; the state table is continuously refreshed by new evidence; when the target is unreachable due to network disconnection or failure, the application layer first perceives the abnormality of the probe and the neighbor / ARP within seconds, and enters the probe priority and degradation transmission without waiting for the system to send a timeout, and then enters the black hole keep-alive to release critical resources after confirming the unreachability; during this period, the congestion coordination continuously tilts the bandwidth to the surviving target to ensure the overall audio continuity; when the target recovers online, the neighbor / ARP is recovered first, the probe is responded, the RTCP quality rises, the EMA score exceeds the recovery threshold, and the system gradually recovers from the black hole to the degradation to the normal according to the hysteresis rule, and the audio code rate is automatically restored.

[0054] The application identifies the reachability decline in advance through multiple source evidence in the front end, pre-loads and releases the buffer / bandwidth through hierarchical shunting in the middle section, avoids retry congestion caused by ARP aging misplacement through active cache maintenance in the side channel, and tilts resources to active flow through congestion coordination in the rear section, while suppressing frequent switching through smoothing / jitter tolerance, thereby maintaining the continuity and low latency of the end-to-end media link without changing the underlying protocol parameters, thereby guaranteeing the continuity and real-time performance of the audio transmission.

[0055] In some optional implementation modes of the embodiment, the joint evaluation is based on at least two types of state indicators of different sources, which are selected from link layer / network layer activity signals, active probe feedback, and service layer quality feedback.

[0056] In this embodiment, when implemented, the sub-scores of each type of index can be defined first, such as the round-trip time (RTT) obtained by active probing, the probe loss rate, the neighbor / ARP "active / failure / change" state code generated by the link / network activity signal and the time interval of the latest event, the packet loss rate, jitter, estimated bandwidth, round-trip time, etc. in the service layer quality feedback (such as RTCP RR). After each sub-score is normalized and mapped to [0, 1], the "reachability score" is obtained by weighted fusion according to the configured weight vector. In order to be compatible with missing or asynchronous indexes, the "last valid value + time decay" can be used for missing items, and a freshness time window (such as 1-3s) is set for each type of index. If the time window is exceeded, the weight is reduced. In this way, reliable joint evaluation can still be completed even if there are only two types of indexes, meeting the extension requirement of "at least two types", and a more stable judgment can be automatically obtained when three types are complete.

[0057] The present application establishes reachability judgment on at least two types of heterogeneous evidence, so that the accidental distortion of any source is balanced / calibrated by another source, and the influence of missing items is weakened by freshness and weight fusion, so that stable and consistent judgment can still be formed in a complex network fluctuation scenario.

[0058] In some optional implementation modes of the present embodiment, the above-mentioned multi-level shunt control at least includes two levels, and the measures are selected from the following items:

[0059] The sending priority of the target data stream is adjusted downward, and the sending is mainly based on the probe;

[0060] The information amount or code rate of the target data stream is degraded to reduce the bandwidth occupation;

[0061] When it is determined that the target is unreachable within a continuous period, the sending of service data of the target is suspended while the periodic probe is retained.

[0062] In this embodiment, a shunt state machine can be defined in the implementation: normal level→probe priority→degraded transmission level→black hole keep-alive level, but it is not mandatory to go through all levels. For example, when the link is degraded but there are occasional responses, only the "priority down probe + code rate degradation" two levels can meet the requirements; if the target is unreachable for a long time, it can directly jump from the probe priority to the black hole keep-alive level. In order to avoid oscillation, the entry / exit hysteresis and the minimum residence time (such as N consecutive negative evidence for entering the degradation, K positive evidence for exiting and the score exceeding the recovery threshold) should be set, and different two-level combinations can be configured for different targets according to the importance of the service, so that flexible scheduling can be realized under the framework of "at least two levels".

[0063] The application designs a shunting strategy with at least two levels of gradual regulation, so that the system can gradually reduce occupation, pre-eliminate congestion in the evolution process of "suspected abnormality → confirmed as unreachable", and avoid state jitter with a hysteresis threshold, thereby protecting key business flows from being squeezed at the initial stage of abnormality.

[0064] In some optional implementations of the embodiment, the above-mentioned multi-source state indicators include:

[0065] Active probe feedback, sending a connectionless probe carrying a time identifier to the target device and determining link abnormality based on the number of consecutive non-responses, and the probe data packet related to the target is sent in a low-priority queue and carries a probe identifier;

[0066] Link layer / network layer activity signal, listening to the operating system neighbor / address resolution table event interface to identify the addition, invalidation or change state of the target device corresponding table item;

[0067] Service layer quality feedback, analyzing the real-time transport control message to obtain indicators including at least packet loss rate, round-trip delay and delay jitter, and generating a target communication quality score based thereon.

[0068] In the embodiment, first, the active probe feedback can use a UDP lightweight probe, the payload contains a timestamp and a random nonce, which is used to calculate the RTT and match with the response; the probe can be sent in a low-priority queue, and a probe mark can be added to the application layer header, and the receiving end can return the mark in the form of application ACK or echo when online. The count of consecutive non-responses reaches a threshold (such as N=3-5) to generate negative evidence. Second, the link / network activity signal can obtain neighbor item addition, invalidation, change and other events by subscribing to the operating system neighbor / ARP event interface (such as Linux netlink, BSD routing socket, Windows NDI), and the event type and the latest occurrence time are used as activity clues. Third, the service layer quality feedback can analyze the RTCP reception report (RR / XR), extract the packet loss rate, jitter, round-trip delay, and reception rate, and map them to quality sub-scores. In engineering, the three types of indicators provide three dimensions of "fast response, bottom layer connectivity, end-to-end quality", and when used together, they can complement each other and significantly reduce single-dimensional misjudgment.

[0069] The application provides fast connectivity evidence through active probe, provides bottom layer reachability endorsement through neighbor / ARP event, and provides end-to-end experience quantification through service layer quality feedback, which are complementary in time scale and semantics. When any evidence chain lags behind, the other two can still support the judgment, thereby significantly shortening the time from abnormality occurrence to system identification.

[0070] In some optional implementations of the embodiment, the above-mentioned multi-level shunting control includes:

[0071] sending the target data packet as a probe packet to a low priority queue when reachability is reduced;

[0072] performing code rate degradation on the data stream of the target to send narrowband placeholder / silence audio when there is no valid response to the continuous probe, and restoring the original code rate when the reachability score recovers to a recovery threshold;

[0073] adding the target to a black hole list and suspending the sending of service data when there is still no valid recovery within a sustained period, while retaining periodic probes at a fixed or adaptive frequency, and setting the entry condition, keep-alive frequency and exit condition of the black hole list.

[0074] In the embodiment, first, when reachability is reduced but not yet confirmed to be unreachable, the service packet is converted to probe priority: the service load is reduced, and the probe is sent intermittently at low priority to reduce the occupation of the queue and bandwidth, while maintaining "liveness detection" of the target. Second, when there is no valid response to the continuous probe, code rate degradation is performed: the bit rate is reduced to narrowband (such as from 64-128 kbps to 8-16 kbps) or replaced by silence / comfort noise placeholder frames on the encoder side to maintain playback continuity and session survival, while releasing link and buffer resources; when the reachability score recovers to the recovery threshold and "continuous positive evidence" is met, the code rate is restored linearly or hierarchically.

[0075] The application minimizes invalid occupation while maintaining session reversibility by forming a "decrease negative pressure first, stop sending, and recoverable and restore" resource recycling link when reachability is reduced, first probe priority to reduce service pressure, code rate degradation to maintain minimum continuity when there is no response for a long time, and entering the black hole keep-alive to release resources but retain the wake-up channel when the target is unreachable for a long time.

[0076] In some optional implementations of the embodiment, the above-mentioned active maintenance of the address resolution mapping cache of the target device includes:

[0077] sending free address resolution packets or neighbor advertisements to refresh or extend the validity of the target table item before sending data to the target device and / or periodically, and setting an upper limit on the sending frequency of such packets, and setting a priority maintenance strategy for key targets.

[0078] In the embodiment, IPv4 Gratuitous ARP can be sent, and IPv6 Neighbor Advertisement (NA) can be sent to refresh the neighbor cache of the local area network device and the host in a unicast or broadcast / multicast manner at a reasonable frequency, so as to reduce the first sending failure or repeated resolution caused by cache aging. In order to avoid being misjudged as abnormal traffic, the sending frequency upper limit and the backoff need to be set: for example, the normal target is greater than or equal to 30 s once, the key target can be greater than or equal to 15-30 s once, and the sending frequency is automatically slowed down when the network is congested or the device CPU is high. At the same time, a key target priority maintenance strategy (such as a conference main mixer / broadcast terminal priority refresh) is provided. The maintenance mechanism is triggered completely in the application layer and does not depend on modifying the system ARP aging parameter, which is convenient for cross-platform deployment.

[0079] The application makes the address resolution item active in advance on the service side, reduces ARP retries / broadcast caused by the first sending and aging misplacement, and reduces the congestion inducement caused by link idle and sending failure by performing Gratuitous ARP / neighbor advertisement on the target before sending and setting the frequency upper limit and the key target priority strategy.

[0080] In some optional implementation manners of the embodiment, when the sending queue resource utilization exceeds the threshold, the priority rearrangement and bandwidth quota adjustment are triggered when the sending queue utilization rate reaches the preset percentage threshold, and the bandwidth proportion of the target in the black hole list is reduced to ensure the transmission continuity of the target in the reachable state.

[0081] In the embodiment, the socket sending buffer usage, the queue depth or the driver layer ring queue occupancy can be periodically read to calculate the utilization rate. When the threshold (such as greater than or equal to 70-85%) is exceeded, the priority rearrangement and bandwidth quota adjustment are triggered. The scheduling can adopt a combination of weighted fair queue (WFQ), deficit round robin (DRR) or “token bucket + minimum quota”: for the target in the black hole list or suspected to be unreachable, the queue weight or service quota is immediately reduced, and only the keep-alive probe is reserved; for the active target with a high score, the minimum quota is guaranteed and the weight can be temporarily increased; when the congestion is removed, the quota is slowly restored according to a smoothing strategy to avoid queue shock caused by instantaneous rebound.

[0082] The application directly suppresses the queuing delay and jitter of the key flow at the queuing level by rearranging the queue weight and bandwidth quota according to the target state when the sending queue high occupancy is triggered, and transferring the resources from the unreachable / suspected to be unreachable flow to the active flow, and smoothly returning after the congestion subsides.

[0083] In some optional embodiments of the present embodiment, in the step of performing timing smoothing and jitter tolerance processing on the reachability determination result to reduce the influence of short-term network fluctuations on the splitting decision, the timing smoothing updates the reachability score using exponential weighted average, and a jitter tolerance time window is set to avoid misjudgment caused by short-term probe interruption.

[0084] In the present embodiment, the score update can be S_t = α·x_t + (1−α)·S_{t−1}, where x_t is the instantaneous score after the latest evidence fusion, and α is taken as 0.5–0.8 to emphasize the current; at the same time, a jitter window W (such as 1–3 s) is set, and sporadic lost probes, short RTT peaks or single neighbor item jitter within W do not immediately trigger degradation or black hole, but require continuous negative evidence to reach the threshold before action. Similarly, when recovering, continuous positive evidence and score exceeding the recovery threshold are required to form the hysteresis of entering / leaving, and this combination can significantly reduce the false switching caused by network transient disturbance, avoid frequent jitter of the state machine, and improve the stability of the user side listening.

[0085] The present application performs low-pass filtering on the score by exponential weighted average, and restricts the sensitivity of “single abnormality triggering” by a jitter tolerance window, so that the state transition must meet the condition of “continuous negative evidence / positive evidence + threshold”, thereby avoiding frequent oscillation of the level / code rate and the perceived jitter caused thereby.

[0086] In some optional embodiments of the present embodiment, message authentication code integrity check is performed on the probe or its confirmation feedback to prevent misjudgment caused by fake probes, and the sending frequency of the packet for actively maintaining the address resolution mapping cache is set to not exceed a preset upper limit to reduce the risk of being misjudged as an attack.

[0087] In the present embodiment, to prevent fake probes / confirmation from interfering with the judgment, a message authentication code (MAC) based on a pre-shared key, such as HMAC-SHA256, is carried in the probe and ACK, which covers the fields of timestamp, nonce, flow identifier, etc.; the receiving end only adopts it as positive evidence after verifying the MAC, and discards the response exceeding the delay upper limit or failing the MAC check. The key can be periodically rotated according to the session / device to reduce the risk of replay. On the other hand, the sending frequency and rate (token bucket or leaky bucket rate limiting) of the maintenance packet (free ARP / NA) and probe traffic for “actively maintaining the address resolution cache” are set to not exceed a preset upper limit, and when the switch / gateway explicitly throttles or alarms, the speed is reduced to avoid being misjudged as an attack or triggering the device side security policy. The above measures ensure that the state judgment is reliable and friendly to the network device.

[0088] The application resists forgery / replay by adding a message authentication code to the probe / confirmation message, and applies a frequency upper limit and backoff to the maintenance and detection messages, so as to ensure that the judgment basis is reliable and avoid being misjudged as abnormal by the network device, thereby stabilizing the input quality of the closed-loop decision and reducing the disturbance caused by external attacks or risk control triggers.

[0089] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by computer readable instructions instructing related hardware, and the computer readable instructions can be stored in a computer readable storage medium. When the program is executed, it can include the processes of the above-mentioned embodiment methods. The storage medium can be a non-volatile storage medium such as a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0090] It should be understood that, although each step in the flowchart of the accompanying drawings is shown in sequence according to the direction of the arrow, these steps are not necessarily executed in sequence according to the direction of the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and they can be executed in other sequences. Moreover, at least part of the steps in the flowchart of the accompanying drawings can include multiple sub-steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence is not necessarily sequential, but can be executed alternately or alternately with at least part of other steps or sub-steps or stages of other steps.

[0091] Further referring to Figure 2 , as an implementation of the method shown in Figure 1 , the application provides an embodiment of an application layer target state awareness and intelligent shunting system. The system embodiment corresponds to the method embodiment shown in Figure 1 , and the system can be applied to various electronic devices.

[0092] As shown in Figure 2 , the application layer target state awareness and intelligent shunting system 200 described in the embodiment includes a state mapping module 201, a shunting control module 202, an address resolution maintenance module 203, a congestion coordination module 204, and a robust decision module 205. Among them:

[0093] The state mapping module 201 is configured to construct reachability state mapping for each target device at the application layer, and jointly evaluate the target communication state. The joint evaluation is based on state indicators from multiple sources.

[0094] The shunt control module 202 is configured to perform multi-stage shunt control on the data flow corresponding to each target according to the reachability state mapping, and switch the sending strategy between different stages, so as to preferentially guarantee the continuous transmission of the active target when the reachability is reduced.

[0095] The address resolution maintenance module 203 is configured to actively maintain the address resolution mapping cache of the target device before data sending and / or according to a preset strategy, so as to reduce the sending blockage caused by cache aging or invalidation.

[0096] The congestion coordination module 204 is configured to implement dynamic coordination of the priority and bandwidth of the target according to the reachability state mapping when the sending queue resource occupation exceeds a threshold, so as to reduce the bandwidth occupation ratio of the unreachable or suspected unreachable target.

[0097] The robust decision module 205 is configured to perform timing smoothing and jitter tolerance processing on the reachability determination result, so as to reduce the influence of short-time network fluctuation on the shunt decision.

[0098] The application layer target state perception and intelligent shunt system provided by the embodiment of the application can realize all processes of the application layer target state perception and intelligent shunt method of the above-mentioned embodiment, the functions of each module in the device and the technical effects achieved, which are the same as the functions of the application layer target state perception and intelligent shunt method of the above-mentioned embodiment and the technical effects achieved, and will not be repeated here.

[0099] The above is only a preferred embodiment of the present application, and is not used to limit the present application, and any modification, equivalent replacement and improvement made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. An application-layer target state perception and intelligent traffic splitting method, applied to distributed network audio transmission, characterized in that, Includes the following steps: At the application layer, an reachability state mapping is constructed for each target device, and the target communication state is jointly evaluated. The joint evaluation is based on state indicators from multiple sources. Based on the reachability state mapping, multi-level traffic splitting control is performed on the data stream corresponding to each target, and the transmission strategy is switched between different levels to prioritize the continuous transmission of active targets when reachability decreases. Before data transmission and / or according to a preset strategy, proactively maintain the address resolution mapping cache of the target device to reduce transmission blockage caused by cache aging or failure; When the resource usage of the sending queue exceeds the threshold, the priority and bandwidth of the target are dynamically coordinated according to the reachability state mapping to reduce the bandwidth ratio of unreachable or suspected unreachable targets. The reachability determination results are processed with time-series smoothing and jitter tolerance to reduce the impact of short-term network fluctuations on traffic splitting decisions.

2. The application-layer target state perception and intelligent traffic diversion method according to claim 1, characterized in that, The joint assessment is based on at least two different types of status indicators, selected from link layer / network layer activity signals, active probe feedback, and service layer quality feedback.

3. The application-layer target state perception and intelligent traffic triage method according to claim 1, characterized in that, The multi-level shunting control includes at least two levels, and its measures are selected from the following: The sending priority of the target data stream is reduced and it is sent primarily in a probing manner; Degrading the amount of information or bit rate of the target data stream to reduce bandwidth usage; If a target is determined to be unreachable for a given period of time, the transmission of business data to that target is suspended while periodic probes are maintained.

4. The application-layer target state perception and intelligent traffic triage method according to any one of claims 1 to 3, characterized in that, In the step of constructing reachability state mappings for each target device at the application layer and jointly evaluating the target communication state, wherein the joint evaluation is based on multi-source state indicators, the multi-source state indicators include: Actively detect and respond by sending a connectionless probe with a time stamp to the target device and determining link anomalies based on the number of consecutive unresponding events. Probe data packets related to the target are sent in a low-priority queue with a detection identifier. Link layer / network layer activity signals; monitor the operating system's neighbor / address resolution table event interface to identify the addition, failure, or change status of the corresponding table entries of the target device. The service layer provides quality feedback by parsing real-time transmission control messages to obtain indicators including at least packet loss rate, round-trip time, and latency jitter, and generates a target communication quality score accordingly.

5. The application-layer target state perception and intelligent traffic triage method according to any one of claims 1 to 3, characterized in that, The multi-level current splitting control includes: When reachability is reduced, the target data packet is sent as a probe packet in a low-priority queue. When there is no effective response during continuous probing, the bit rate of the data stream to the target is downgraded to send narrowband placeholder / silent audio, and then restored to the original bit rate when the reachability score recovers to the recovery threshold; If no effective recovery is achieved within the specified period, the target will be added to the black hole list and service data transmission will be suspended. At the same time, periodic probes will be retained at a fixed or adaptive frequency, and the entry conditions, keep-alive frequency and exit conditions for the black hole list will be set.

6. The application-layer target state perception and intelligent traffic triage method according to claim 1, characterized in that, The active maintenance of the target device's address resolution mapping cache includes: Before sending data to the target device, and / or periodically send free address resolution messages or neighbor advertisements to refresh or extend the validity of target entries, set an upper limit on the sending frequency of such messages, and set priority maintenance policies for critical targets.

7. The application-layer target state perception and intelligent traffic splitting method according to claim 1, characterized in that, In the step of dynamically coordinating the priority and bandwidth of targets based on the reachability state mapping when the resource occupancy of the sending queue exceeds the threshold, and reducing the bandwidth ratio of unreachable or suspected unreachable targets, priority reordering and bandwidth quota adjustment are triggered when the utilization rate of the sending queue reaches a preset percentage threshold, and the bandwidth ratio of targets in the black hole list is reduced to ensure the transmission continuity of targets in the reachable state.

8. The application-layer target state perception and intelligent traffic splitting method according to claim 1, characterized in that, In the step of performing time-series smoothing and jitter tolerance processing on the reachability determination results to reduce the impact of short-term network fluctuations on the traffic splitting decision, the time-series smoothing uses an exponentially weighted average to update the reachability score and sets a jitter tolerance time window to avoid misjudgment caused by short-term probe interruption.

9. The application-layer target state perception and intelligent traffic splitting method according to claim 8, characterized in that, The integrity of the message authentication code is verified on the probe or its confirmation feedback to prevent misjudgment caused by forged probes, and a strategy is set to limit the message sending frequency of actively maintaining the address resolution mapping cache to no more than a preset upper limit to reduce the risk of being misjudged as an attack.

10. An application-layer target state perception and intelligent traffic diversion system, characterized in that, include: The state mapping module is used to construct reachability state mappings for each target device at the application layer and to jointly evaluate the target communication state, which is based on state indicators from multiple sources. The traffic splitting control module is used to perform multi-level traffic splitting control on the data stream corresponding to each target according to the reachability state mapping, and switch the transmission strategy between different levels to prioritize the continuous transmission of active targets when reachability is reduced. The address resolution maintenance module is used to proactively maintain the address resolution mapping cache of the target device before data transmission and / or according to a preset strategy, so as to reduce transmission blockage caused by cache aging or failure. The congestion coordination module is used to dynamically coordinate the priority and bandwidth of targets based on the reachability state mapping when the resource occupancy of the sending queue exceeds the threshold, thereby reducing the bandwidth ratio of unreachable or suspected unreachable targets. The robust decision module is used to perform time-series smoothing and jitter tolerance processing on the reachability determination results, so as to reduce the impact of short-term network fluctuations on the traffic splitting decision.

Images