A performance testing method, device, and medium for Ethernet switches

Abstract

This invention discloses a performance testing method, device, and medium for Ethernet switches, belonging to the field of Ethernet switch testing. The method includes: constructing a dynamic random data generation model that reflects the fluctuation of the testing process over time; generating test signals that reflect the fluctuation of test data volume over time; constructing a state control model for Ethernet switch services based on queuing theory, and adjusting the rate at which test data packets are injected into the Ethernet switch in real time using the state control model; selecting the output test data packet that best matches each test data packet, and calculating the performance coefficient of the Ethernet switch tested for each test data packet; calculating the average performance coefficient of the Ethernet switch, setting a performance coefficient threshold for the Ethernet switch, and evaluating the performance of the tested Ethernet switch. This method can achieve consistency of test data across multiple nodes, supports concurrent testing of tens of thousands of ports, and provides effective decision support for the configuration optimization of Ethernet switches.

Description

Technical Field

[0001] This invention relates to the field of Ethernet switch performance testing, and specifically to a performance testing method, device, and medium for Ethernet switches. Background Technology

[0002] An Ethernet switch is a switch that transmits data based on Ethernet, a local area network (LAN) that uses a shared bus transmission medium. The structure of an Ethernet switch is such that each port is directly connected to a host, and it typically operates in full-duplex mode. A switch can simultaneously connect many pairs of ports, allowing each pair of communicating hosts to transmit data without collisions, as if they had exclusive access to the communication medium.

[0003] Ethernet switches are widely used across various fields due to their low cost. In complex application scenarios, the stability of Ethernet switch performance is crucial. Performance testing of Ethernet switches is a key step in evaluating their stability under high-speed, high-concurrency conditions. Traditional Ethernet switch performance testing methods primarily rely on fixed traffic patterns and single-metric evaluation, which have the following limitations:

[0004] Static load model: Existing methods typically use fixed-rate traffic injection, which cannot simulate the dynamically changing load characteristics in a real network, resulting in test results that are out of touch with the real scenario.

[0005] The one-sidedness of a single indicator: evaluating performance solely through isolated indicators such as throughput and latency ignores the coupling effect between multi-dimensional indicators. For example, high throughput may be accompanied by high packet loss rate, and traditional methods cannot quantify this trade-off. Summary of the Invention

[0006] To address the aforementioned shortcomings in the existing technology, this invention provides a performance testing method, device, and medium for Ethernet switches, applicable to performance evaluation of Ethernet switches in high-speed, high-concurrency scenarios.

[0007] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows:

[0008] A performance testing method for an Ethernet switch is provided, comprising:

[0009] Step S1: Determine the maximum data processing throughput of the Ethernet switch based on its design parameters. Based on maximum throughput Construct a dynamic random data generation model that allows the testing process to fluctuate over time.

[0010] Step S2: During the testing of the Ethernet switch, a test signal with a test data volume fluctuating over time is generated based on the dynamic random data generation model. The test signal generates different numbers of test data packets and sends them to the Ethernet switch.

[0011] Step S3: According to t Number of test packets sent to the Ethernet switch at any time A state control model for Ethernet switch services is constructed based on queuing theory, and the rate at which test packets are injected into the Ethernet switch is adjusted in real time through the state control model.

[0012] Step S4: Collect the dataset of test data packets injected into the Ethernet switch at various times during the test period, as well as the dataset of test data packets output by the Ethernet switch, based on the theoretical delay time of the Ethernet switch's signal transmission. Filter the output test data packets that best match each test data packet, and calculate the performance coefficient of the Ethernet switch tested by each test data packet;

[0013] Step S5: Calculate the average performance coefficient of the Ethernet switch, set the performance coefficient threshold of the Ethernet switch, and evaluate the performance of the tested Ethernet switch.

[0014] Furthermore, the dynamic random data generation model is as follows:

[0015] ;

[0016] in, t For the time of the test cycle, This refers to the amount of test data that fluctuates with the time series during the testing period. To measure the fluctuation range of the test data volume, To test the amount of data in the Hearst index H Fractional Brownian motion, , K The number of periodic components in the test data volume. k This refers to the numbering of the periodic components. For the periodic components of the test data volume, For the frequency of the periodic component, The phase of the periodic component, phase Uniformly distributed within the range [0, 2π);

[0017] An adaptive algorithm is used to calculate the Hearst exponent of the test data in real time, and the Hearst exponent is dynamically adjusted.

[0018] ;

[0019] in, The Hearst exponent fluctuates with time series. The initial Hearst exponent, The adjustment range for the Hearst index. This is the adjustment time window for the Hearst exponent. As the attenuation factor, The time interval for adjusting the amount of test data. e It is a natural constant. For the switch to adjust the time interval The length of the queue below, For queue threshold, The rectifier function is the queue length.

[0020] Furthermore, the state control model for Ethernet switch services is as follows:

[0021] ;

[0022] in, i This represents the number of test packets in the queue of the Ethernet switch. n Parallel processing unit for Ethernet switches There are exactly 10 ... i The probability of a test data packet. For the service rate of Ethernet switches, K This refers to the queue capacity of the Ethernet switch. This represents the queue idle probability of an Ethernet switch.

[0023] Further, step S4 includes:

[0024] Step S41: Collect the dataset of test data packets injected into the Ethernet switch at various times during the test period. ,in, m This refers to the number of test packets injected into the Ethernet switch during the test period. For the first m The injection time of each test data packet, For the first m The number of test data packets;

[0025] Step S42: Collect the dataset of test data packets output by the Ethernet switch during the test period. ,in, For the first m At the time of output of each test data packet, For the output of the first m The number of test data packets;

[0026] Step S43: Based on the theoretical delay time of the Ethernet switch's signal transmission... From the dataset Filtering and each test data packet The most matching output test data packet is used to calculate the performance coefficient of the Ethernet switch for each test data packet.

[0027] Furthermore, the specific method for calculating the performance coefficient of the Ethernet switch for each test packet includes the following steps:

[0028] Step S431: Based on the first test data packet injection time Traverse the dataset Filter the matching dataset of the first test data packet. And satisfy The matching dataset includes data at theoretical delay times. The output test data packet that can match the first test data packet the most;

[0029] Step S432: Calculate the first test data packet and the matching dataset Similarity between each output test data packet ;

[0030] ;

[0031] Step S433: Obtain w Affinity data Filter out similarity data minimum value minimum value The corresponding output test data packet best matches the first test data packet, and the minimum value is set. The corresponding output test data packet is from the dataset. Remove from middle;

[0032] Step S434: Set the minimum value The performance coefficient of the Ethernet switch was determined by the first test data packet. ;

[0033] Step S435: Repeat steps S431-S434, sequentially selecting the output test data packet that best matches the test data packet from the remaining output test data packets, and calculating the performance coefficient of the Ethernet switch for each test data packet. .

[0034] Furthermore, step S5 specifically includes:

[0035] Step S51: Calculate the average performance coefficient of the Ethernet switch. , vThe number of the test data packet, For the first v The performance coefficient of an Ethernet switch is determined by testing a single test packet.

[0036] Step S52: Set the performance coefficient threshold for the Ethernet switch. ,like If the Ethernet switch fails to meet the performance requirements, it is deemed to be unqualified; otherwise, the Ethernet switch is deemed to be qualified.

[0037] An electronic device is provided, comprising: a memory storing program code thereon; and a processor connected to the memory, which implements the method described above when the program code is executed by the processor.

[0038] A computer-readable storage medium is provided, on which program instructions are stored, which, when executed, implement the method described above.

[0039] The beneficial effects of this invention are as follows: By constructing a dynamic traffic data generation model, this invention simulates the randomness of data volume in real-world application scenarios during Ethernet switch testing, and simulates queue congestion in actual networks, ensuring the consistency of performance test results with real-world scenarios. It can achieve consistency of test data across multiple nodes, supports concurrent testing of tens of thousands of ports, and has a high accuracy rate in matching input and output data packets, effectively improving the accuracy of performance testing and evaluation, and providing effective decision support for Ethernet switch configuration optimization. Attached Figure Description

[0040] Figure 1 This is a flowchart of a performance testing method for Ethernet switches. Detailed Implementation

[0041] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0042] like Figure 1 As shown, a performance testing method for an Ethernet switch includes:

[0043] Step S1: Determine the maximum data processing throughput of the Ethernet switch based on its design parameters. Based on maximum throughput Construct a dynamic random data generation model that reflects the fluctuations of the testing process over time.

[0044] ;

[0045] in, t For the time of the test cycle, This refers to the amount of test data that fluctuates with the time series during the testing period. To measure the fluctuation range of the test data volume, To test the amount of data in the Hearst index H Fractional Brownian motion, , K The number of periodic components in the test data volume. k This refers to the numbering of the periodic components. For the periodic components of the test data volume, For the frequency of the periodic component, The phase of the periodic component, phase Uniformly distributed within the range [0, 2π);

[0046] An adaptive algorithm is used to calculate the Hearst exponent of the test data in real time, and the Hearst exponent is dynamically adjusted.

[0047] ;

[0048] in, The Hearst exponent fluctuates with time series. The initial Hearst exponent, The adjustment range for the Hearst index. This is the adjustment time window for the Hearst exponent. As the attenuation factor, The time interval for adjusting the amount of test data. e It is a natural constant. For the switch to adjust the time interval The length of the queue below, For queue threshold, The rectified function is the queue length;

[0049] The backplane throughput of an Ethernet switch, also known as backplane bandwidth, is measured in packets per second (pps). It represents the maximum amount of data that can be processed between the switch's interface processor or interface card and the data bus. The higher the backplane bandwidth of a switch, the stronger its data processing capability.

[0050] This embodiment tests Ethernet switches by generating dynamically changing test data volumes, thereby simulating the periodic characteristics of network traffic, such as backup traffic in a data center (which typically occurs at night).

[0051] The test measures frame rate fluctuations in video conferencing (typically around 30Hz) and QoS guarantee periods for cloud services (such as AWS's 5-minute monitoring window). Furthermore, based on fractional Brownian motion, the test data exhibits drastic traffic fluctuations on small timescales (e.g., milliseconds) and smooths out fluctuations on large timescales (e.g., minutes), making the characteristics of the test signal highly consistent with the "burst-smooth" alternation of TCP traffic. This allows for the generation of traffic with strong bursts, testing the switch's queue management capabilities under extreme conditions. The periodic component gives the traffic stable temporal characteristics, enabling precise measurement of the switch's latency jitter under periodic loads.

[0052] Step S2: During the testing of the Ethernet switch, a test signal is generated based on the dynamic random data generation model, which generates test data volume that fluctuates with the time period. The test signal generates different numbers of test data packets and sends them to the Ethernet switch.

[0053] Step S3: According to t Number of test packets sent to the Ethernet switch at any time A state control model for Ethernet switch services is constructed based on queuing theory, and the rate at which test packets are injected into the Ethernet switch is adjusted in real time through the state control model.

[0054] ;

[0055] in, i This represents the number of test packets in the queue of the Ethernet switch. n For Ethernet switches, the parallel processing unit represents the number of forwarding engines or queue processing threads operating in parallel, characterizing the Ethernet switch's concurrent processing capability and directly impacting performance in multi-stream scenarios. There are exactly 10 ... i The probability of a test data packet. The service rate of an Ethernet switch represents the rate at which an Ethernet switch port processes data packets, i.e., the number of data packets that can be forwarded per unit of time. K Queue capacity is the maximum number of test packets that the Ethernet switch's port buffer can store. This represents the queue idle probability of an Ethernet switch.

[0056] The first equation of the state control model is the steady-state equation for data processing by the Ethernet switch. Based on this equation, the data processing capacity of the Ethernet switch can be adjusted in real time to ensure stable operation. This effectively simulates the congestion risk of the Ethernet switch under high load and the packet loss rate under different loads. The second equation is the queue length probability distribution formula, used to describe the probability distribution characteristics of the number of data packets in the queue of the Ethernet switch.P i With parallel processing unit n When the packet loss rate increases and then decreases rapidly, it indicates a low probability of queue congestion; if the loss rate decreases slowly, it suggests that the packet loss rate may increase under high load.

[0057] Queue idle probability The calculation method is as follows:

[0058] ;

[0059] Step S4: Collect the dataset of test data packets injected into the Ethernet switch at various times during the test period, as well as the dataset of test data packets output by the Ethernet switch, based on the theoretical delay time of the Ethernet switch's signal transmission. The process involves selecting the output test data packet that best matches each test data packet and calculating the performance coefficient of the Ethernet switch for each test data packet. Step S4 specifically includes:

[0060] Step S41: Collect the dataset of test data packets injected into the Ethernet switch at various times during the test period. ,in, m This refers to the number of test packets injected into the Ethernet switch during the test period. For the first m The injection time of each test data packet, For the first m The number of test data packets;

[0061] Step S42: Collect the dataset of test data packets output by the Ethernet switch during the test period. ,in, For the first m At the time of output of each test data packet, For the output of the first m The number of test data packets;

[0062] Step S43: Based on the theoretical delay time of the Ethernet switch's signal transmission... From the dataset Filtering and each test data packet The most matching output test data packet is used to calculate the performance coefficient of the Ethernet switch for each test data packet.

[0063] The specific method for calculating the performance coefficient of the Ethernet switch for each test packet includes the following steps:

[0064] Step S431: Based on the first test data packet injection time Traverse the dataset Filter the matching dataset of the first test data packet. And satisfy The matching dataset includes data at theoretical delay times. The output test data packet that can match the first test data packet the most;

[0065] Because Ethernet switches exhibit varying latency and packet loss rates for different test data packets during data transmission, the dataset... The first output test data packet may not correspond to the first test data packet. Therefore, this embodiment introduces a matching process. Within a reasonable delay period corresponding to the output test data collected by the Ethernet switch, the output test data packet is matched with the test data packet to ensure that the calculated performance coefficient is accurate enough and to further improve the performance test accuracy of the Ethernet switch.

[0066] Step S432: Calculate the first test data packet and the matching dataset Similarity between each output test data packet ;

[0067] ;

[0068] In the formula for calculating similarity, the first term represents the packet loss rate of the Ethernet switch during data transmission. This is calculated by subtracting the number of output test packets from the number of test packets injected into the Ethernet switch, dividing by the number of test packets, and then using the ln function to obtain the packet loss rate. The second term represents the data transmission delay as a percentage of the theoretical delay. The ratio is calculated, and then the latency data is obtained using the ln function. Finally, the latency data and packet loss rate data are used to obtain performance coefficients that characterize the performance of the Ethernet switch.

[0069] Step S433: Obtain w Affinity data Filter out similarity data minimum value minimum value The corresponding output test data packet best matches the first test data packet, and the minimum value is set. The corresponding output test data packet is from the dataset. Remove from middle;

[0070] Step S434: Set the minimum value The performance coefficient of the Ethernet switch was determined by the first test data packet. ;

[0071] Step S435: Repeat steps S431-S434, sequentially selecting the output test data packet that best matches the test data packet from the remaining output test data packets, and calculating the performance coefficient of the Ethernet switch for each test data packet. .

[0072] Step S5: Calculate the average performance coefficient of the Ethernet switches, set the performance coefficient threshold for the Ethernet switches, and evaluate the performance of the tested Ethernet switches. Step S5 specifically includes:

[0073] Step S51: Calculate the average performance coefficient of the Ethernet switch. , v The number of the test data packet, For the first v The performance coefficient of an Ethernet switch is determined by testing a single test packet.

[0074] Step S52: Set the performance coefficient threshold for the Ethernet switch. ,like If the Ethernet switch fails to meet the performance requirements, it is deemed to be unqualified; otherwise, the Ethernet switch is deemed to be qualified.

[0075] An electronic device includes: a memory storing program code thereon; and a processor connected to the memory, which, when the program code is executed by the processor, implements the above-described performance testing method for an Ethernet switch.

[0076] A computer-readable storage medium having program instructions stored thereon, which, when executed, implement the above-described performance testing method for an Ethernet switch.

[0077] This invention constructs a dynamic traffic data generation model to simulate the randomness of data volume in real-world application scenarios during Ethernet switch testing, and also simulates queue congestion in actual networks, ensuring the consistency of performance test results with real-world scenarios. It can achieve consistency of test data across multiple nodes, supports concurrent testing of tens of thousands of ports, and has high accuracy in matching input and output data packets, effectively improving the accuracy of performance testing and evaluation, and providing effective decision support for Ethernet switch configuration optimization.

Claims

1. A method of performance testing of an Ethernet switch, characterized by, The method comprises the following steps: Step S1: Determine the maximum throughput of data processing of the Ethernet switch according to the design parameters thereof , construct a dynamic random data generation model of test process fluctuation over time period based on the maximum throughput , construct a dynamic random data generation model of test process fluctuation over time period based on the maximum throughput Step S2: in the test process of the Ethernet switch, a test signal with test data amount fluctuating with time period is generated based on a dynamic random data generation model, and the test signal generates different amounts of test data packets and sends them to the Ethernet switch; Step S3: According to t The number of test data packets sent to the Ethernet switch at the moment , based on queuing theory to build a state control model of Ethernet switch service, through the state control model to adjust the rate of test data packets injected into the Ethernet switch in real time; Step S4: collecting the data set of the test data packets injected into the Ethernet switch at each time in the test period, and the data set of the test data packets output by the Ethernet switch, according to the theoretical delay time of the Ethernet switch transmitting signals , screening the output test data packet most matched with each test data packet, and calculating the performance coefficient of the Ethernet switch tested by each test data packet; Step S5: calculating the average value of the performance coefficient of the Ethernet switch, setting the performance coefficient threshold of the Ethernet switch, and evaluating the performance of the tested Ethernet switch; The dynamic random data generation model is: ； wherein, t is the time of the test cycle, is the amount of test data fluctuating over time during the test cycle, is the amplitude of the fluctuations of the amount of test data, is the fraction Brownian motion of the amount of test data under the Hurst exponent H , , K is the number of periodic components of the amount of test data, k is the number of the periodic component, is the periodic component of the amount of test data, is the frequency of the periodic component, is the phase of the periodic component; The Hurst index of the test data amount is calculated in real time by using an adaptive algorithm, and the Hurst index is dynamically adjusted; ； wherein, Hurst exponent fluctuating over time, is an initial Hurst exponent, is a modulation amplitude of the Hurst exponent, is a modulation time window of the Hurst exponent, is a decay factor, is a modulation time interval of the amount of test data, e is a natural constant, is a queue length of the switch at the modulation time interval , is a queue threshold, is a rectification function of the queue length; The state control model of the Ethernet switch service is: ； wherein, i is the number of test packets in the queue of the Ethernet switch, n is the parallel processing unit of the Ethernet switch is the probability that there are exactly i test packets in the queue, is the service rate of the Ethernet switch, K is the queue capacity of the Ethernet switch, is the queue idle probability of the Ethernet switch.

2. The method of claim 1, wherein the performance test of the Ethernet switch is performed by a test tool. The step S4 comprises: Step S41: collecting the data set of the test data packets injected into the Ethernet switch at each time in the test period wherein, m is the number of times of injecting the test data packets into the Ethernet switch in the test period, is the injection time of the i-th test data packet, m is the injection time of the i-th test data packet, is the number of the i-th test data packet; and m is the number of the i-th test data packet. Step S42: Collect the dataset of test data packets output by the Ethernet switch during the test period. ,in, For the first m At the time of output of each test data packet, For the output of the first m The number of test data packets; Step S43: According to the theoretical delay time of the Ethernet switch transmission signal , the data set is filtered from the most matching output test data packet for each test data packet , and the performance coefficient of the Ethernet switch tested by each test data packet is calculated.

3. The method of claim 2, wherein the performance test of the Ethernet switch is performed by a test packet generator. The The specific method for calculating the performance coefficient of the Ethernet switch tested by each test data packet comprises: Step S431: Based on the first test data packet injection time Traverse the dataset Filter the matching dataset of the first test data packet. And satisfy ; Step S432: Calculate the similarity between the 1st test data packet and the matching data set the similarity between each output test data packet ; ； Step S433: obtaining w the similarity data , screening out the minimum value in the similarity data The minimum value corresponding to the output test data packet is the most matched with the first test data packet, and the minimum value corresponding to the output test data packet is removed from the data set ;​ Step S434: determining the minimum value Performance coefficient of the Ethernet switch tested as the first test data packet ; Step S435: repeat steps S431-S434 to select the output test data packet that is most matched with the remaining test data packets from the remaining output test data packets, and calculate the performance coefficient of the Ethernet switch tested by each test data packet .

4. The method of claim 3, wherein the performance test of the Ethernet switch is performed by a test packet generator. The step S5 specifically comprises: Step S51: Calculate the average value of the performance coefficient of the Ethernet switch , v is the number of test data packets, is the performance coefficient of the Ethernet switch tested by the v th test data packet Step S52: set the performance coefficient threshold of the Ethernet switch If , it is determined that the performance of the Ethernet switch is unqualified, otherwise, the performance of the Ethernet switch is qualified.

5. An electronic device, comprising: The method comprises the following steps: A memory having program codes stored thereon; A processor coupled with the memory, and when the program codes are executed by the processor, the method of claim 4 is implemented.

6. A computer-readable storage medium, characterized in that, A memory having program codes stored thereon, and the program codes are executed to implement the method of claim 4.

Images