Network optimization method and device, storage medium and computer program product

Abstract

The embodiment of the invention discloses a network optimization method and device, a storage medium and a computer program product. The method comprises the following steps: detecting a network congestion state of a current network; when it is detected that the network congestion state is that the current network is in network congestion, obtaining an adjustment factor; determining a first value and a second value based on the adjustment factor; the first numerical value represents the sending rate of a target data packet, and the second numerical value represents the size of a target congestion window; based on the first numerical value and the second numerical value, optimizing the current network to obtain an optimized target network; the target network is used for realizing data migration.

Description

Technical Field

[0001] This application relates to the field of computer network technology, and in particular to a network optimization method, device, storage medium, and computer program product. Background Technology

[0002] With the rapid development of the Internet, the data generated by various applications and businesses is growing explosively. Therefore, big data migration to the cloud is an inevitable trend, and migration is a necessary means for big data migration to the cloud. Furthermore, a good congestion control algorithm is an important guarantee for efficient and stable data migration during the big data migration process.

[0003] In related technologies, packet loss-based congestion control algorithms passively judge network congestion based on packet loss events in the network. Even if the network load is high, the sending rate will not be actively reduced as long as no congestion packet loss occurs. Once network congestion is detected, the congestion window will be reduced at an exponential rate, resulting in very low bandwidth utilization. Moreover, the congestion window takes a long time to recover to its normal size, causing fluctuations in migration rate and affecting the stability of big data migration. At the same time, congestion packet loss can generally only be detected after the migration link buffer is exhausted, which has a certain lag compared to the actual network congestion situation. Summary of the Invention

[0004] To address the technical problems existing in related technologies, embodiments of this application provide a network optimization method, device, storage medium, and computer program product.

[0005] To achieve the above objectives, the technical solution of this application embodiment is implemented as follows: In a first aspect, embodiments of this application provide a network optimization method, the method comprising: Detect the current network congestion status; If the network congestion state is detected to be that the current network is experiencing network congestion, an adjustment factor is obtained; Based on the adjustment factor, a first value and a second value are determined; the first value represents the target data packet transmission rate, and the second value represents the target congestion window size. Based on the first value and the second value, the current network is optimized to obtain an optimized target network; the target network is used to realize data migration.

[0006] Secondly, embodiments of this application also provide a network optimization device, including: a processor and a memory for storing a computer program capable of running on the processor; When the processor runs the computer program, it executes the steps of the network optimization method described in the embodiments of this application.

[0007] Thirdly, embodiments of this application also provide a storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the network optimization method described in embodiments of this application.

[0008] Fourthly, embodiments of this application also provide a computer program product, including a computer program that, when executed by a processor, implements the steps of the network optimization method described in embodiments of this application.

[0009] The network optimization method, device, storage medium, and computer program product provided in this application embodiment detect the current network congestion state; when the network congestion state is detected to be that the current network is experiencing network congestion, an adjustment factor is obtained; based on the adjustment factor, a first value and a second value are determined; the first value represents the target data packet sending rate, and the second value represents the target congestion window size; based on the first value and the second value, the current network is optimized to obtain an optimized target network; the target network is used to realize data migration. By adopting the technical solution of this application embodiment, when the current network is already experiencing congestion, by introducing an adjustment factor, the data packet sending rate and the value of the congestion window can be adaptively adjusted according to different network environments, thereby mitigating network congestion and reducing the migration performance loss caused by network congestion, thereby improving link utilization and enhancing the stability and efficiency of big data cloud migration networks. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating the network optimization method according to an embodiment of this application. Figure 1 ; Figure 2 This is a schematic diagram illustrating the workflow of the BBR algorithm in an embodiment of this application; Figure 3 This is a flowchart illustrating the network optimization method according to an embodiment of this application. Figure 2 ; Figure 4 This is a flowchart illustrating the network optimization method according to an embodiment of this application. Figure 3 ; Figure 5 This is a schematic diagram of the composition structure of the network optimization device according to an embodiment of this application; Figure 6 This is a schematic diagram of the hardware structure of the network optimization device according to an embodiment of this application. Detailed Implementation

[0011] The present application will now be described in further detail with reference to the accompanying drawings and embodiments.

[0012] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application.

[0013] With the rapid development of the internet, the data generated by various applications and businesses is experiencing explosive growth. Therefore, big data migration to the cloud is an inevitable trend, and migration is a necessary means to achieve this. A good network environment is an important condition for migration. If network congestion occurs frequently during the migration process, a certain proportion of data packets will be retransmitted, resulting in a decrease in network throughput and a significant loss in migration performance. Therefore, a good congestion control algorithm is an important guarantee for efficient and stable data migration during big data cloud migration.

[0014] Among related technologies, packet loss-based congestion control algorithms, such as Reno, NewReno (an improved version of Reno), and Cubic, have exposed numerous problems. Specifically, packet loss-based congestion control algorithms passively judge network congestion based on packet loss events in the network. Even if the network load is high, they will not actively reduce their transmission rate unless congestion-related packet loss occurs. Once network congestion is detected, the congestion window decreases exponentially, resulting in very low bandwidth utilization. Furthermore, the congestion window takes a long time to recover to its normal size, causing fluctuations in migration rates and affecting the stability of big data migration. Additionally, packet loss-based congestion control algorithms typically only detect congestion-related packet loss after the migration link buffer is exhausted, which introduces a certain lag compared to the actual network congestion situation.

[0015] To address the problems of early packet loss-based congestion control algorithms, a new congestion control algorithm based on bottleneck bandwidth and round-trip time (BBR) has been proposed to improve bandwidth utilization and latency performance in networks. The BBR algorithm calculates the bandwidth-delay product (BDP) by detecting the bottleneck bandwidth and round-trip time (RTT) of network links, thereby dynamically adjusting the data transmission rate and congestion window size to achieve congestion control. Unlike earlier packet loss-based congestion control algorithms, the BBR algorithm improves the conditions for determining network congestion, more accurately reflecting the actual congestion status of the network and improving network migration efficiency. Furthermore, this algorithm allows network migration to achieve the maximum migration rate and the minimum latency possible.

[0016] While the BBR algorithm is a high-performance congestion control algorithm, it suffers from drawbacks in practical applications, including slow convergence, insufficient packet loss resilience, and limited jitter resistance. In big data cloud migration scenarios, when the network environment has limited caching capacity, its packet loss resilience falls short of expectations. The algorithm periodically sends packets to probe the maximum available bandwidth of the link; when the cache is full, this leads to a high packet loss rate, requiring extensive retransmission. Conversely, when the network environment has sufficient caching capacity, the BBR algorithm fails to effectively utilize cache resources, maintaining a low bandwidth change rate and resulting in slow convergence of multiple data streams sharing the link bandwidth to the ideal state. Therefore, the BBR algorithm cannot dynamically adjust based on the link cache size. Furthermore, its jitter resistance is limited; when new data streams join, network throughput fluctuates significantly, and the time required to converge to a state with evenly distributed link bandwidth is lengthy.

[0017] Because the network environment involved in big data cloud migration scenarios is relatively complex, higher requirements are placed on congestion control algorithms in terms of convergence performance, stability, and adaptability. The BBR algorithm in related technologies performs poorly in these aspects in big data cloud migration scenarios. Therefore, it is necessary to optimize the BBR algorithm to address its shortcomings in big data cloud migration scenarios.

[0018] Based on this, this application proposes a network optimization method. In various embodiments of this application, when the current network is already experiencing congestion, by introducing an adjustment factor, the data packet sending rate and the value of the congestion window can be adaptively adjusted according to different network environments, thereby alleviating network congestion and reducing the migration performance loss caused by network congestion, thereby improving link utilization and enhancing the stability and efficiency of big data cloud migration networks.

[0019] This application provides a network optimization method, which is applied to a network optimization device. Figure 1 This is a flowchart illustrating the network optimization method according to an embodiment of this application. Figure 1 ,like Figure 1 As shown, the network optimization method includes: Step 101: Detect the current network congestion status.

[0020] In this embodiment, the network optimization method can be applied to big data cloud migration scenarios. Accordingly, the current network can be the current migration network. In other words, the network optimization method proposed in this embodiment is a method for optimizing the migration network involved in big data cloud migration scenarios.

[0021] Here, the network optimization device can be an electronic device. This application embodiment does not limit the specific type of electronic device. For example, it can be a smartphone, tablet computer, etc. The network optimization device can also be a server. This application embodiment does not limit the specific type of server. For example, it can be a cloud server.

[0022] Here, the current network congestion state includes one of the following: the current network is experiencing network congestion; the current network is not experiencing network congestion. It should be noted that in practical applications, the current network environment may be different at different times, and correspondingly, the degree of network congestion will also be different.

[0023] Step 102: If the network congestion state is detected as the current network is experiencing network congestion, obtain the adjustment factor.

[0024] Here, the network optimization device can determine whether there is congestion in the current network by comparing the actual amount of data in the network link with the maximum capacity that the network link can store data (i.e., the bandwidth-delay product (BDP)). Specifically, if the actual amount of data in the network link is less than the BDP, it is determined that the network congestion has been completely cleared; if the actual amount of data in the network link is greater than or equal to the BDP, it is determined that there is still congestion in the current network. At this time, an adjustment factor can be introduced to reduce the data packet sending rate and the value of the congestion window, thereby alleviating network congestion and reducing the migration performance loss caused by network congestion, thereby improving link utilization and improving the stability and efficiency of the network for big data cloud migration.

[0025] In practical applications, the network optimization device has already generated the adjustment factor before obtaining it.

[0026] Accordingly, in one embodiment, the method further includes: determining the adjustment factor; Determining the adjustment factor includes: A first delay and a second delay are determined; the first delay represents the average delay detected within a detection period; the second delay represents the network delay when network congestion occurs in the current network. The adjustment factor is determined based on the first delay and the second delay.

[0027] Here, the probe period refers to each period in the bandwidth probe (ProbeBW, Probe Bandwidth) phase of the BBR algorithm when probing the maximum available bandwidth of the migration link. The migration link can be simply referred to as a link.

[0028] Here, in one embodiment, determining the adjustment factor based on the first delay and the second delay includes: determining the adjustment factor based on the ratio of the first delay to the second delay. Specifically, this can be expressed by the following formula (1): (1); in, Indicates the regulating factor; This indicates the first time delay (i.e., the average time delay detected within one detection period); This indicates the second delay (i.e., the network delay when network congestion occurs in the current network). This indicates the total number of data streams present in the current network links; Indicates the number of data streams in the link, for example At that time, it represents the second data stream in the link, where, ; Indicates the current data stream transmission delay; Indicates data stream The average delay detected within a detection period.

[0029] Here, data streams can also be called data packets. In practical applications, The larger the value, the more severe the congestion in the current network (i.e., the migrating network). In this case, the reduction in packet sending rate and congestion window size should be increased. In other words, the adjustment factor... The smaller the value, the better.

[0030] Step 103: Determine the first value and the second value based on the adjustment factor.

[0031] In this embodiment of the application, the first value represents the target data packet transmission rate, that is, the adjusted data packet transmission rate; the second value represents the target congestion window size, that is, the adjusted congestion window size.

[0032] In practical applications, network optimization devices can optimize the initial data packet sending rate (i.e., the source data packet sending rate) by using adjustment factors to obtain an improved target data packet sending rate.

[0033] Based on this, in one embodiment, determining the first value based on the adjustment factor includes: A third value is determined; the third value represents the magnitude of the source data packet transmission rate. The third value is adjusted using the adjustment factor to obtain the first value.

[0034] Here, the data packet sending rate can also be called the data packet transmission rate. The magnitude of the source data packet sending rate is mainly determined by the data packet sending rate gain factor, which is the most important parameter for reducing packet loss and improving convergence speed in the bandwidth probing phase of the BBR algorithm.

[0035] Based on this, in one embodiment, determining the third value includes: Determine a first gain factor and a first bandwidth; the first gain factor represents the data packet transmission rate gain factor, and the first bandwidth represents the maximum available bandwidth of the link detected during the bandwidth detection phase; The third value is determined based on the first gain factor and the first bandwidth.

[0036] Here, in one embodiment, determining the third value based on the first gain factor and the first bandwidth includes: determining the third value based on the product of the first gain factor and the first bandwidth. Specifically, this can be expressed by the following formula (2): (2); in, This represents the third value (i.e., the rate at which the source data packet is sent). This represents the first gain factor (i.e., the data packet transmission rate gain factor). This represents the first bandwidth (i.e., the maximum available bandwidth of the link detected during the bandwidth detection phase).

[0037] Here, after determining the third value, the network optimization device uses an adjustment factor. The third value is improved to obtain the first value (i.e., the improved data packet sending rate, which is also the target data packet sending rate). Specifically, it can be expressed by the following formula (3): (3); in, Indicates the first numerical value; This represents the first gain factor (i.e., the data packet transmission rate gain factor). This represents the first bandwidth (i.e., the maximum available bandwidth of the link detected during the bandwidth detection phase). This indicates a regulatory factor.

[0038] As can be seen, in this embodiment of the application, if the network optimization device detects that the current network is congested, in order not to further aggravate the congestion, the source data packet sending rate should be automatically reduced according to the degree of network congestion. Therefore, an adjustment factor is introduced to reduce the source data packet sending rate to alleviate congestion. That is, the improved data packet sending rate obtained by the above formula (3) can be adaptively adjusted according to different network environments, reducing the migration performance loss caused by network congestion, thereby improving link utilization and improving the stability and efficiency of big data cloud migration network.

[0039] The process of determining the first gain factor is explained in detail below.

[0040] Before describing the process of determining the first gain factor, this application will first describe the workflow of the BBR algorithm. The state machine used in the BBR algorithm is mainly divided into four stages: Startup, Drain, ProbeBW, and ProbeRTT. Figure 2 This is a schematic diagram of the workflow of the BBR algorithm in an embodiment of this application. The following is in conjunction with... Figure 2 The workflow of the BBR algorithm is explained as follows: Startup Phase: In this phase, the BBR algorithm updates the data packet transmission rate with a gain factor of 2 / ln2, rapidly increasing the network data packet transmission rate to quickly fill the network channel. If the calculated bandwidth does not increase for three consecutive times, it can be determined that the maximum bandwidth has been reached and enters the emptying state. At this time, the actual amount of data in the network link is three times the bandwidth * delay.

[0041] Emptying Phase: In this phase, the BBR algorithm rapidly reduces the network packet sending rate with a gain factor of ln2 / 2 to empty the excess packets sent to the network buffer during the startup phase. If the actual amount of data in the network link (inflight) is less than the maximum capacity that the network link can hold (i.e., the bandwidth-delay product (BDP)), it means that the network congestion has been completely cleared and the algorithm can proceed to the next state (i.e., the bandwidth probing phase). If inflight is greater than or equal to BDP, it means that the network is still congested and the algorithm cannot proceed to the next state; emptying continues.

[0042] Bandwidth Probing Phase: In this phase, the BBR algorithm probes the maximum available bandwidth of the link by periodically adjusting the packet sending rate. If the minimum RTT is not obtained within a certain time period (usually 10 seconds), it then proceeds to the next phase (i.e., the latency probing phase).

[0043] Latency Probing Phase: In this phase, the primary goal is to detect the minimum network latency (i.e., minimum RTT). During this time, the BBR algorithm maintains a fixed sending window of 4 packets, meaning it sends almost no packets, accounting for 2% of the entire process. Detecting the minimum RTT requires minimizing network congestion, reducing the congestion window, and sending relatively few packets. After detecting the minimum RTT, if the link bandwidth is fully utilized, the process enters the bandwidth probing phase; otherwise, it enters the startup phase to attempt to fill the network channel.

[0044] During the entire BBR algorithm operation, most of the time is spent in the bandwidth probing (ProbeBW) phase. Therefore, the embodiments of this application mainly optimize this phase to improve the network environment for big data cloud migration and enhance migration quality and efficiency.

[0045] Here, there are two important parameters in the bandwidth probing phase: packet transmission rate and congestion window (cwnd). The packet transmission rate is mainly determined by the packet sending rate gain factor, which is the most important parameter for reducing packet loss and improving convergence speed in the bandwidth probing phase. The size of the congestion window is mainly determined by the congestion window gain factor and BDP.

[0046] For big data migration scenarios to the cloud, the network environment is highly complex. The packet transmission rate gain factor in the BBR algorithm of related technologies is a fixed array of values, such as [2 / ln2, ln2 / 2, 1, 1, 1, 1], which cannot adaptively adjust according to differences in the network environment and the degree of network congestion. Therefore, this application analyzes the multi-flow probe bandwidth behavior in the bandwidth probing phase of the BBR algorithm, analyzes the value of the packet transmission rate gain factor, and calculates the optimal packet transmission rate gain factor that can adapt to different network environments.

[0047] In practical applications, there are often multiple data streams in a migration link. Each data stream sends data packets to perform maximum bandwidth probing during the bandwidth probing phase. The multi-stream bandwidth probing behavior is related to the buffering of the migration link in the network environment, thereby better setting the data packet sending rate gain factor.

[0048] Based on this, in one embodiment, determining the first gain factor includes: Determine the current link buffer size and the second bandwidth; the second bandwidth represents the initial bandwidth allocated to the current data stream. The first gain factor is determined based on the current link buffer size and the second bandwidth.

[0049] Here, it is assumed that there are a total of [number] in the migration link. For each data stream, the corresponding packet sending rate gain factor is an array. Each gain factor in this array can be adaptively adjusted according to the differences in network environment and network congestion level, so as to adapt to different network environments.

[0050] Here, let's assume the first value in the array of packet transmission rate gain factors is denoted as... Convergence bandwidth in the migration link and The value of is related to . During the bandwidth probing phase, data streams may over-transmit data. If the amount of over-transmitted data exceeds the capacity of the migration link buffer, the excess data will result in packet loss, leading to packet retransmission, wasting network resources, and reducing the utilization rate of the migration link. To address this issue, this application adds a constraint to the packet transmission rate gain factor, limiting the amount of over-transmitted data during the bandwidth probing phase to not exceeding the capacity of the migration link buffer. This effectively reduces the packet loss rate of the link and allows the data stream to converge to the ideal state as quickly as possible.

[0051] Here, in one embodiment, determining the first gain factor based on the current link buffer size and the second bandwidth includes: determining the first gain factor based on the ratio of the current link buffer size to the second bandwidth.

[0052] Here, the first gain factor is taken as For example, specifically, it can be represented by the following formula (4): (4); in, This indicates the current link buffer size (also known as the current link cache size). This indicates the second bandwidth, i.e., the first data stream. The corresponding initial bandwidth.

[0053] It should be noted that the process for determining the packet transmission rate gain factor for other data streams is the same as that for the first data stream. The process for determining the corresponding data packet transmission rate gain factor is similar and will not be repeated here.

[0054] In practical applications, network optimization devices can optimize the initial congestion window size (i.e., the source congestion window size) by using adjustment factors to obtain an improved target congestion window size.

[0055] Therefore, in one embodiment, determining the second value based on the adjustment factor includes: Determine the fourth value; the fourth value represents the source congestion window size; The fourth value is adjusted using the adjustment factor to obtain the second value.

[0056] In practical applications, the source congestion window size is mainly determined by the congestion window gain factor and BDP.

[0057] Based on this, in one embodiment, determining the fourth value includes: Determine a second gain factor and a fifth value; the second gain factor represents the congestion window gain factor, and the fifth value represents the bandwidth-delay product; The fourth value is determined based on the second gain factor and the fifth value.

[0058] Here, the bandwidth-delay product can be understood as the maximum capacity that data can be stored in a network link. The bandwidth-delay product is related to two parameters: the maximum available bandwidth of the link detected by the BBR algorithm and the minimum network latency.

[0059] In one embodiment, determining the fifth value includes: determining the maximum available bandwidth of the link and the minimum network latency; and determining the fifth value based on the product of the maximum available bandwidth of the link and the minimum network latency. Specifically, it can be expressed by the following formula (5): (5); in, This indicates the maximum available bandwidth of the link detected during the bandwidth detection phase (corresponding to the aforementioned first bandwidth). Indicates the minimum network latency; This represents the fifth value (i.e., the bandwidth-delay product).

[0060] Here, in one embodiment, determining the fourth value based on the second gain factor and the fifth value includes: determining the fourth value based on the product of the second gain factor and the fifth value. Specifically, this can be expressed by the following formula (6): (6); in, This represents the fourth value (i.e., the source congestion window size). This represents the second gain factor (i.e., the congestion window gain factor).

[0061] Here, after determining the fourth value, the network optimization device uses an adjustment factor. The fourth value is improved to obtain the second value (i.e., the improved congestion window size, or the target congestion window size). Specifically, it can be expressed by the following formula (7): (7); in, Indicates the second numerical value; This represents the second gain factor (i.e., the congestion window gain factor). This represents the fifth value (i.e., the bandwidth-delay product). This indicates a regulatory factor.

[0062] As can be seen, in this embodiment of the application, if the network optimization device detects that the current network is congested, in order not to further aggravate the congestion, the source congestion window size should also be automatically reduced according to the degree of network congestion. Therefore, an adjustment factor is introduced to reduce the source congestion window size to alleviate the network burden. That is, the improved congestion window size obtained by the above formula (7) can be adaptively adjusted according to different network environments, reducing the migration performance loss caused by network congestion, thereby improving link utilization and improving the stability and efficiency of big data cloud migration network.

[0063] Step 104: Based on the first value and the second value, optimize the current network to obtain the optimized target network.

[0064] In this embodiment of the application, the target network is used to realize data migration, specifically, big data migration to the cloud.

[0065] This application also provides another network optimization method, which is applied to a network optimization device. Figure 3 This is a flowchart illustrating the network optimization method according to an embodiment of this application. Figure 2 ,like Figure 3 As shown, the network optimization method includes: Step 301: Detect the current network congestion status.

[0066] Step 302: If the network congestion state is detected as the current network is experiencing network congestion, obtain the adjustment factor.

[0067] In one embodiment, the method further includes: determining the adjustment factor; Determining the adjustment factor includes: A first delay and a second delay are determined; the first delay represents the average delay detected within a detection period; the second delay represents the network delay when network congestion occurs in the current network. The adjustment factor is determined based on the first delay and the second delay.

[0068] Step 303: Determine the third value, and adjust the third value using the adjustment factor to obtain the first value.

[0069] In this embodiment of the application, the first value represents the target data packet transmission rate; the third value represents the source data packet transmission rate.

[0070] Here, in one embodiment, determining the third value includes: Determine a first gain factor and a first bandwidth; the first gain factor represents the data packet transmission rate gain factor, and the first bandwidth represents the maximum available bandwidth of the link detected during the bandwidth detection phase; The third value is determined based on the first gain factor and the first bandwidth.

[0071] In one embodiment, determining the first gain factor includes: Determine the current link buffer size and the second bandwidth; the second bandwidth represents the initial bandwidth allocated to the current data stream. The first gain factor is determined based on the current link buffer size and the second bandwidth.

[0072] Step 304: Determine the fourth value, and adjust the fourth value using the adjustment factor to obtain the second value.

[0073] In this embodiment of the application, the second value represents the target congestion window size, and the fourth value represents the source congestion window size.

[0074] Here, in one embodiment, determining the fourth value includes: Determine a second gain factor and a fifth value; the second gain factor represents the congestion window gain factor, and the fifth value represents the bandwidth-delay product; The fourth value is determined based on the second gain factor and the fifth value.

[0075] Step 305: Based on the first value and the second value, optimize the current network to obtain the optimized target network.

[0076] Step 306: Use the target network to perform data migration.

[0077] It should be noted that the specific process by which the network optimization device performs network optimization has been detailed above and will not be repeated here.

[0078] By adopting the technical solution of this application embodiment, when the current network is already experiencing congestion, by introducing an adjustment factor, the data packet sending rate and the value of the congestion window can be adaptively adjusted according to different network environments, thereby alleviating network congestion and reducing the migration performance loss caused by network congestion, thereby improving link utilization and enhancing the stability and efficiency of big data cloud migration networks.

[0079] The present application will be described below with reference to application examples.

[0080] Since the BBR algorithm in related technologies does not consider network caching resources, this application proposes an optimization method for migration networks in big data cloud migration scenarios. This method improves the BBR algorithm by optimizing the behavior of each data flow during the stage of probing the maximum available bandwidth of the link. It dynamically adjusts the initial value strategy of the data flow's transmission rate gain based on the size of the buffer and bandwidth, ensuring that each data flow converges quickly and does not exceed the buffer size. When network congestion occurs during migration, adjustment factors are introduced into the value strategy of transmission rate gain and congestion window gain. By reducing both transmission rate gain and congestion window gain, network congestion is mitigated, reducing the migration performance loss caused by network congestion, improving link utilization, and enhancing migration speed and stability.

[0081] The core of the BBR algorithm is to detect two parameters: maximum bandwidth (corresponding to the maximum available bandwidth of the link mentioned above) and minimum delay (i.e., the minimum network latency). The product of maximum bandwidth and minimum delay yields the BDP, which is the maximum capacity (i.e., bandwidth-delay product) that can be stored in the network link. Specifically, it can be expressed by the following formula (5): (5); in, Indicates the maximum bandwidth; Indicates the minimum delay; This represents the bandwidth-delay product.

[0082] The algorithm can be divided into four stages: Startup, Drain, ProbeBW (Bandwidth Probe), and ProbeRTT (Latency Probe). Its workflow includes: Startup Phase: The BBR algorithm rapidly increases the network packet transmission rate with a gain factor of 2 / ln2, quickly filling the network channel. If the calculated bandwidth does not increase for three consecutive times, it can be determined that the maximum bandwidth has been reached and enters the Drain state. At this time, the actual amount of data in the network link is three times the bandwidth * delay.

[0083] Drain Phase: The BBR algorithm rapidly reduces the network packet sending rate with a gain factor of ln2 / 2 to drain the excess packets sent to the network buffer during the startup phase. If the actual amount of data in the network link (inflight) is less than the maximum data capacity (BDP) that the network link can hold, it means that the congestion caused by BBR has been completely cleared. If inflight is greater than or equal to BDP, it means that the network is still congested and cannot proceed to the next state (i.e., the ProbeBW phase).

[0084] Bandwidth Probe Phase (ProbeBW): The BBR algorithm probes the maximum available bandwidth of the link by periodically adjusting the packet sending rate. If the minimum RTT is not obtained within a certain time period (usually 10 seconds), it then enters the next phase (i.e., the ProbeRTT phase).

[0085] Delay Probe Time To Test (ProbeRTT): This phase primarily aims to detect the minimum network latency. During this time, the BBR algorithm maintains a fixed sending window of 4 packets, meaning it sends almost no packets, accounting for 2% of the entire process. Detecting the minimum RTT requires minimizing network congestion, reducing the congestion window, and sending relatively few packets. After detecting the minimum latency, if the link bandwidth is fully utilized, the process enters the bandwidth probe phase; otherwise, it enters the startup phase to attempt to fill the network channel.

[0086] During the entire BBR algorithm operation, most of the time is spent in the ProbeBW phase. Therefore, this application focuses on optimizing this phase in order to improve the network environment for big data cloud migration and enhance migration quality and efficiency.

[0087] In the ProbeBW phase, there are two important parameters: transmission rate ( The transmission rate is mainly determined by the rate gain factor ( ) and the congestion window (cwnd). To decide, It is the most important parameter for reducing packet loss and improving convergence speed in the ProbeBW phase; the size of the congestion window is mainly determined by the window gain coefficient ( The decision is made based on ( ) and BDP. That is: (2); (6); The network environment in big data cloud migration scenarios is highly complex. The transmission rate gain factor in the BBR algorithm is a fixed array of values ​​[2 / ln2, ln2 / 2, 1, 1, 1, 1], which cannot adaptively adjust according to differences in the network environment and the degree of network congestion. This application analyzes the multi-stream probe bandwidth behavior in the bandwidth probing phase of the BBR algorithm and analyzes... The problem of determining the value of is to calculate the optimal value. It can adapt to different network environments.

[0088] The main implementation process of this application is described below.

[0089] I. Bandwidth Probing Behavior Analysis In a migration link, there are often multiple network flows. Each flow sends data packets to perform maximum bandwidth probing during the bandwidth probing phase. In order to detect the relationship between the bandwidth probing behavior of multiple flows and the network environment, so as to better set the data packet sending rate gain.

[0090] Assuming there are a total of Each flow, denoted as the maximum bandwidth in the migration link. The data flow in the link is denoted as ( The initial bandwidth obtained by each data stream in the link is Then there is .

[0091] During the bandwidth probing phase of the BBR algorithm, the data stream periodically sends probe packets to detect whether there is any spare available bandwidth in the link. The packet sending rate during bandwidth probing is multiplied by... The value can be used as the final transmission rate, and data over-transmission can be achieved by increasing the transmission rate. For simplicity, this application will... The first value is denoted as (i.e., the first transmission rate gain factor).

[0092] In a multi-stream scenario, let's first assume... Then there is The initial bandwidth of the stream is , The initial bandwidth of the stream is ,but During the bandwidth probing phase, it is assumed that... The flow first performs bandwidth detection, and then... Initial bandwidth of the stream Multiply by the gain factor ,Right now To enable the over-sending of data packets, The bandwidth of the stream remains constant. At this time... Over-sending packets in a flow will consume the migration link's buffer. Assuming the migration link's buffer is... .exist After flow probing is complete, oversent data will remain in the buffer to form a queue. At this point, the packet sending rate gain factor will be reduced to [value missing]. This allows excess data cached in the migration chain to be drained, at which point: (8); in, for The amount of data over-produced; This is the ratio of the initial maximum data volume of the link to the maximum data volume of the link after data over-sending.

[0093] After emptying the over-issued data, due to... ,therefore Therefore, we can conclude that: (9); exist After the flow detection is completed, The stream begins probing bandwidth. Similar to flow, The flow is first multiplied by the gain factor. The subsequent transmission rate probes for excess bandwidth in the migration link and occupies the migration link buffer. In this state: (10); It should be noted that the formula (10) uses... Initial bandwidth of the stream Multiply by the gain factor Instead of using the value of formula (9), that is This is because Compare The value is small, and bandwidth probing is to detect excess maximum bandwidth. If the probed value is smaller than the original value, bandwidth updates are usually not triggered. This mechanism ensures that multiple flows in the link can eventually converge to the ideal state.

[0094] There are also [something] within the buffer. The queues created by the over-sending of data in the stream still need to be emptied by reducing the sending rate. At this point, we have: (11); in, express The amount of data sent during bandwidth probing. express After the flow completes bandwidth probing, it migrates the remaining data volume on the link. Therefore, express The amount of data over-produced during the bandwidth probing phase. This represents the ratio of the initial maximum data volume of the link to the maximum data volume of the link after data over-transmission.

[0095] Similarly, after emptying the over-issued data, due to the presence of ,therefore Therefore, we can conclude that: (12); In practical applications, when flow and harmony After the streams have completed bandwidth probing, flow and harmony The bandwidth of the flow has been fully updated. As can be seen from formulas (11) and (12), the convergence bandwidth in the migration link is... The value of is related to .

[0096] This application expects flow and harmony During continuous bandwidth probing, the flow can converge as quickly as possible to a state where the bottleneck link bandwidth is evenly distributed. That is, the bandwidth obtained after bandwidth probing and emptying the queue is... Under these conditions, it is possible to maximize .

[0097] Furthermore, during bandwidth probing, data over-transmission occurs. If the over-transmitted data exceeds the link buffer capacity, the excess data will result in packet loss, leading to packet retransmission, wasting network resources, and reducing the utilization of the migrated link. To address this issue, this application introduces a packet transmission rate gain factor. Add constraints to ensure that the amount of data over-issued during the bandwidth probing phase does not exceed the migration link's cache. When When the stream performs bandwidth probing, The bandwidth occupied by the stream needs to The following constraints must be met between them: (13); Therefore, under the constraints of the above equation, the following equation can be derived: (14); According to the optimization principle, using the least squares method, the above equation (14) can be transformed into: (15); right After differentiation, we can obtain: (16); Therefore, during the bandwidth probing phase, multiple flows limit the amount of over-transmitted data to no more than the buffer capacity of the migration link. This effectively reduces the packet loss rate of the link and allows the data flow to converge to the ideal state as quickly as possible. The first value in the array It should be set to As can be seen, the value of this gain coefficient is related to the buffer of the current migration link and the bandwidth allocated to the current data stream.

[0098] II. Network Congestion Behavior Analysis In addition to considering the size of the bottleneck link buffer, the value should also be further adjusted according to the current network congestion situation. The value should also take into account the impact of network congestion. If network congestion has been detected, the sending rate should be automatically reduced according to the degree of congestion to prevent further exacerbation. The congestion window should also automatically decrease according to the degree of congestion, thereby reducing the network burden and improving the stability and efficiency of the network for big data migration to the cloud.

[0099] Based on the aforementioned behavior analysis during the bandwidth probing phase, this application introduces a regulation factor. , Let be the ratio of the average latency detected within a detection period to the network latency when network congestion occurs. Then we have: (1); in, This represents the average time delay detected within one detection period. This indicates network latency when network congestion occurs.

[0100] In practical applications, the current data packet transmission delay The larger the value, the more severe the congestion in the migration network. In this case, the packet sending rate and the reduction in the congestion window size should be increased. The value of should also be as small as possible. Combining the packet sending rate and the congestion window value with the adjustment factor, the following formula can be obtained when network congestion occurs: (3); (7); in, This indicates the improved data packet sending rate. This indicates the improved congestion window size.

[0101] As can be seen from the above formulas (3) and (7), the improved data packet sending rate and congestion window value can be adaptively adjusted according to different network environments, which increases the flexibility of the algorithm and improves the stability and efficiency of big data cloud migration network.

[0102] Figure 4 This is a flowchart illustrating the network optimization method according to an embodiment of this application. Figure 3 ,like Figure 4As shown, for big data cloud migration scenarios, this network optimization method is specifically implemented as follows: During the bandwidth probing (ProbeBW) phase, the maximum available bandwidth of the link is probed by periodically adjusting the data packet sending rate. The minimum network latency was detected during the ProbeRTT phase. Next, we calculate BDP, that is... It also sets a congestion window, the size of which can be adjusted by... The calculation is then performed to determine if network congestion is occurring. Specifically, it checks if the actual amount of data (inflight) in the network link is less than the maximum data storage capacity (BDP) of the network link. If it is not less, it indicates that network congestion still exists, and an adjustment factor is introduced. To reduce the transmission rate and congestion window size, the formula can be used. and If the value is less than 1, it means that the network congestion caused by BBR has been completely cleared. Maintaining the source transmission rate and congestion window size can be achieved using the formula... and Finally, the big data migration process is implemented based on the sending rate and the congestion window size.

[0103] This application proposes a network optimization method for big data cloud migration, which is an improved version of the BBR algorithm. Specifically designed for big data cloud migration scenarios, this method, based on the existing BBR algorithm, comprehensively considers the size of the migration network environment buffer and network congestion during the migration process to propose an improved network optimization method. This method significantly improves the performance indicators of the migration network environment, such as convergence speed, latency, packet loss, and link utilization. It can maintain a relatively stable migration speed in big data cloud migration scenarios and has good perception and adjustment capabilities to changes in the network environment. In complex network environments, it can adjust to changes in the network environment, avoiding performance loss caused by network environment changes.

[0104] Furthermore, due to the complexity of the network environment in big data cloud migration scenarios, there are high requirements for data migration speed, latency, and stability. The fixed transmission rate gain of the traditional BBR algorithm can no longer meet the requirements of this scenario. To address this, this application describes the convergence behavior of each flow in the bandwidth probing phase of the BBR algorithm by considering link bandwidth and link buffer, calculates the changes in bandwidth allocation after bandwidth probing for each flow, and, under the constraint that the amount of over-transmitted data by each flow does not exceed the link buffer, derives an adaptive transmission rate gain strategy based on the network environment. Subsequently, the impact of network congestion is further considered in the transmission rate gain strategy, and an adjustment factor is introduced. After detecting network congestion, the adjustment factor is used to further alleviate congestion by reducing the transmission rate gain and the congestion window size.

[0105] Compared with the solutions of related technologies, the solution of this application has the following advantages: 1) The BBR algorithm in related technologies has a slow bandwidth convergence speed and a low transmission rate gain. The fixed value selection strategy does not consider the differences in cache resources under complex network scenarios and cannot make adaptive adjustments based on network cache resources. Therefore, this application, in the complex network scenario of big data cloud migration, combines cache resources to... By making more reasonable adjustments, a strategy for determining the optimal transmission rate gain based on the network environment is derived. This approach can significantly improve performance metrics such as convergence speed, latency, packet loss, and link utilization when migrating to a new network environment, avoiding performance losses caused by changes in the network environment and reducing the waste of network resources.

[0106] 2) In addition to considering cache resources, this application also introduces an adjustment factor, which can be adjusted according to the current network environment. When network congestion occurs, the adjustment factor can be adaptively adjusted according to the degree of network congestion, thereby reducing the data packet sending rate, reducing the size of the congestion window, accelerating the relief of network congestion during the migration of big data to the cloud, and improving the efficiency and stability of big data migration.

[0107] To implement the network optimization method of this application embodiment, this application embodiment also provides a network optimization apparatus. Figure 5 This is a schematic diagram of the composition structure of the network optimization device according to an embodiment of this application, as shown below. Figure 5 As shown, the network optimization device includes: Detection unit 51 is used to detect the current network congestion status; The acquisition unit 52 is used to acquire an adjustment factor when the detection unit 51 detects that the network congestion state is that the current network is experiencing network congestion; The first determining unit 53 is used to determine a first value based on the adjustment factor; the first value represents the target data packet transmission rate. The second determining unit 54 is used to determine a second value based on the adjustment factor; the second value represents the target congestion window size. The optimization unit 55 is used to optimize the current network based on the first value and the second value to obtain an optimized target network; the target network is used to realize data migration.

[0108] In one embodiment, the network optimization apparatus further includes: a third determining unit; wherein, The third determining unit is used to determine the adjustment factor; The third determining unit is specifically used for: A first delay and a second delay are determined; the first delay represents the average delay detected within a detection period; the second delay represents the network delay when network congestion occurs in the current network. The adjustment factor is determined based on the first delay and the second delay.

[0109] In one embodiment, the first determining unit 53 includes: a fourth determining unit and a fifth determining unit; wherein, The fourth determining unit is used to determine the third value; the third value represents the magnitude of the source data packet transmission rate; The fifth determining unit is used to adjust the third value using the adjustment factor to obtain the first value.

[0110] In one embodiment, the fourth determining unit includes: a sixth determining unit, a seventh determining unit, and an eighth determining unit; wherein, The sixth determining unit is used to determine the first gain factor; the first gain factor characterizes the data packet transmission rate gain factor. The seventh determining unit is used to determine the first bandwidth; the first bandwidth represents the maximum available bandwidth of the link detected during the bandwidth detection phase; The eighth determining unit is used to determine the third value based on the first gain factor and the first bandwidth.

[0111] In one embodiment, the sixth determining unit is specifically used for: Determine the current link buffer size and the second bandwidth; the second bandwidth represents the initial bandwidth allocated to the current data stream. The first gain factor is determined based on the current link buffer size and the second bandwidth.

[0112] In one embodiment, the second determining unit 54 includes: a ninth determining unit and a tenth determining unit; wherein, The ninth determining unit is used to determine the fourth value; the fourth value represents the source congestion window size; The tenth determining unit is used to adjust the fourth value using the adjustment factor to obtain the second value.

[0113] In one embodiment, the ninth determining unit is specifically used for: Determine a second gain factor and a fifth value; the second gain factor represents the congestion window gain factor, and the fifth value represents the bandwidth-delay product; The fourth value is determined based on the second gain factor and the fifth value.

[0114] In practical applications, the detection unit 51, the first determination unit 53, the second determination unit 54, and the optimization unit 55 can be implemented by the processor in the network optimization device, and the acquisition unit 52 can be implemented by the communication interface in the network optimization device.

[0115] It should be noted that the network optimization device provided in the above embodiments is only illustrated by the division of the above program modules when performing network optimization. In practical applications, the above processing can be assigned to different program modules as needed, that is, the internal structure of the device can be divided into different program modules to complete all or part of the processing described above. In addition, the network optimization device and the network optimization method embodiments provided in the above embodiments belong to the same concept, and their specific implementation process can be found in the network optimization method embodiments, which will not be repeated here.

[0116] Based on the hardware implementation of the above program modules, and in order to implement the network optimization method of this application embodiment, this application embodiment also provides a network optimization device. Figure 6 This is a schematic diagram of the hardware composition structure of the network optimization device according to an embodiment of this application, such as... Figure 6 As shown, the network optimization device 60 includes: The communication interface 61 enables information exchange with other devices; The processor 62 is connected to the communication interface 61 to enable information interaction with other devices. When running a computer program, it executes the network optimization method provided above, and the computer program is stored in the memory 63.

[0117] It should be noted that the specific processing procedures of communication interface 61 and processor 62 can be understood by referring to the network optimization method described above.

[0118] Of course, in practical applications, the various components in the network optimization device 60 are coupled together via a bus system 64. It is understood that the bus system 64 is used to implement communication between these components. In addition to a data bus, the bus system 64 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 6 The general labeled all buses as Bus System 64.

[0119] The memory 63 in this embodiment is used to store various types of data to support the operation of the network optimization device 60. Examples of such data include any computer program used to operate on the network optimization device 60.

[0120] The network optimization method disclosed in the above embodiments of this application can be applied to the processor 62, or implemented by the processor 62. The processor 62 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above network optimization method can be completed by the integrated logic circuit of the hardware or by instructions in the form of software in the processor 62. The processor 62 may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 62 can implement or execute the network optimization methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the network optimization method disclosed in the embodiments of this application can be directly reflected as being executed by a hardware decoding processor, or being executed by a combination of hardware and software modules in the decoding processor. The software modules may be located in a storage medium, which is located in the memory 63. The processor 62 reads the information in the memory 63 and combines its hardware to complete the steps of the aforementioned network optimization method.

[0121] In an exemplary embodiment, the network optimization device 60 may be implemented by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers (MCUs), microprocessors, or other electronic components to perform the aforementioned network optimization method.

[0122] It is understood that the memory 63 in this embodiment can be volatile memory or non-volatile memory, or both. Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), ferromagnetic random access memory (FRAM), flash memory, magnetic surface memory, optical disc, or compact disc read-only memory (CD-ROM); magnetic surface memory can be disk storage or magnetic tape storage. Volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Synchronous Static Random Access Memory (SSRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic Random Access Memory (SDRAM), Double Data Rate Synchronous Dynamic Random Access Memory (DDRSDRAM), Enhanced Synchronous Dynamic Random Access Memory (ESDRAM), SyncLink Dynamic Random Access Memory (SLDRAM), and Direct Rambus Random Access Memory (DRRAM).The memory 63 described in the embodiments of this application is intended to include, but is not limited to, these and any other suitable types of memory.

[0123] In an exemplary embodiment, this application also provides a storage medium, namely a computer storage medium, specifically a computer-readable storage medium, such as a memory 63 storing a computer program. This computer program can be executed by a processor 62 in the network optimization device 60 to complete the steps of the network optimization method described in the foregoing embodiments of this application. The computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, Flash Memory, magnetic surface memory, optical disc, or CD-ROM.

[0124] In an exemplary embodiment, this application also provides a computer program product, including a computer program that can be executed by a processor 62 in a network optimization device 60 to perform the steps of the network optimization method described in the foregoing embodiments of this application.

[0125] It should be noted that terms such as "first," "second," and "third" are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0126] Furthermore, the technical solutions described in the embodiments of this application can be combined arbitrarily without conflict.

[0127] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A network optimization method, characterized in that, The method includes: Detect the current network congestion status; If the network congestion state is detected to be that the current network is experiencing network congestion, an adjustment factor is obtained; Based on the adjustment factor, a first value and a second value are determined; the first value represents the target data packet transmission rate, and the second value represents the target congestion window size. Based on the first value and the second value, the current network is optimized to obtain an optimized target network; the target network is used to realize data migration.

2. The method according to claim 1, characterized in that, The method further includes: Determine the adjustment factor; Determining the adjustment factor includes: A first delay and a second delay are determined; the first delay represents the average delay detected within a detection period; the second delay represents the network delay when network congestion occurs in the current network. The adjustment factor is determined based on the first delay and the second delay.

3. The method according to claim 1, characterized in that, Determining the first value based on the adjustment factor includes: A third value is determined; the third value represents the magnitude of the source data packet transmission rate. The third value is adjusted using the adjustment factor to obtain the first value.

4. The method according to claim 3, characterized in that, The determination of the third value includes: Determine a first gain factor and a first bandwidth; the first gain factor represents the data packet transmission rate gain factor, and the first bandwidth represents the maximum available bandwidth of the link detected during the bandwidth detection phase; The third value is determined based on the first gain factor and the first bandwidth.

5. The method according to claim 4, characterized in that, Determining the first gain factor includes: Determine the current link buffer size and the second bandwidth; the second bandwidth represents the initial bandwidth allocated to the current data stream. The first gain factor is determined based on the current link buffer size and the second bandwidth.

6. The method according to claim 1, characterized in that, Determining the second value based on the adjustment factor includes: Determine the fourth value; the fourth value represents the source congestion window size; The fourth value is adjusted using the adjustment factor to obtain the second value.

7. The method according to claim 6, characterized in that, The determination of the fourth value includes: Determine a second gain factor and a fifth value; the second gain factor represents the congestion window gain factor, and the fifth value represents the bandwidth-delay product; The fourth value is determined based on the second gain factor and the fifth value.

8. A network optimization device, characterized in that, include: A processor and a memory for storing computer programs capable of running on the processor; When the processor is used to run the computer program, it performs the steps of the method according to any one of claims 1 to 7.

9. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images