[0022] Combine the following figure 1 and figure 2 Describe one embodiment of the present invention:
[0023] figure 1 This is a flowchart of a load sharing adjustment method in an embodiment of the present invention. The method includes:
[0024] 102. Measure the traffic of each link.
[0025] For example, the link here can be a Trunk port, a Tunnel tunnel, an equal-cost route, or other links that require load balancing. The traffic of each link can be measured regularly, and the measurement interval can be fixed or configured by the command line. For example, each link traffic can be measured every 5 minutes. You can also measure the traffic of each link according to the traffic change. For example, set a traffic change threshold. When the traffic change between two measurements is greater than the threshold, reduce the measurement time interval. When the traffic change between the two measurements is less than the threshold increase the measurement time interval. It can also combine timing measurement and measurement based on traffic changes, such as setting a fixed measurement time interval and a traffic change threshold, measuring the traffic of each link at a fixed measurement time point, and when the traffic change between two measurements is greater than the traffic When changing the threshold, one or more measurements are added between two fixed measurement time points.
[0026] 104. Estimate the predicted traffic corresponding to each hash value according to the binding relationship between the link and the hash value and the traffic of each link, where the total number of hash values is greater than the number of links. The hash value is obtained according to information such as the IP address and/or the MAC address of the packet in the flow-by-flow forwarding, and the total number thereof is determined by the adopted flow-by-flow forwarding algorithm. The binding relationship between the link and the hash value is used to select the forwarding path. For example, the ratio of the total number of hash values to the number of links may be set to be 2:1 or higher, for example, if there are 4 links, the total number of hash values may be set to be 8 or more.
[0027] For example, the predicted traffic corresponding to each hash value is estimated by an arithmetic mean method according to the total number of hash values bound by each link and the traffic of each link. For example, there are 4 links, 16 hash values, and the traffic of each link is T 1 , T 2 , T 3 , T 4 , the hash values bound by each link are 3, 3, 4, and 6 respectively, then the predicted traffic of the 3 hash values is estimated to be T 1 /3, the predicted flow of 3 hash values is T 2 /3, the predicted flow of 4 hashes is T 3 /4, the predicted flow of 6 hashes is T 4 /6.
[0028] 106. Adjust the binding relationship between the link and the hash value according to the predicted traffic corresponding to each hash value, so as to improve the uniformity of load sharing.
[0029] In order to improve the uniformity of load sharing in flow-by-stream routing and forwarding, it is necessary to minimize the dispersion degree of the predicted traffic corresponding to each link after adjustment. Therefore, it is necessary to provide a quantitative standard for the uniformity of load sharing. The predicted traffic corresponding to each link refers to the sum of the predicted traffic corresponding to all hash values bound to each link.
[0030] For example, the range of the predicted traffic corresponding to each link is used as a standard. The smaller the range is, the more uniform the load is shared. The range refers to the difference between the maximum value and the minimum value of the predicted traffic corresponding to each link. Similarly, the variance or standard deviation of the predicted traffic corresponding to each link can also be used as a standard. The smaller the variance or standard deviation, the more uniform the load sharing is. Var ( T ) = Σ i = 1 n ( T i - T ‾ ) 2 n , where n is the total number of links, T i is the predicted traffic corresponding to each link, and T is the average value of the predicted traffic corresponding to each link. The standard deviation is the arithmetic square root of the variance.
[0031] Those of ordinary skill in the art can know that other quantities that can represent the degree of data dispersion or data fluctuation can be used as a quantitative standard for the effect of adjusting the binding relationship between the link and the hash value, that is, the uniformity of load sharing.
[0032] Different algorithms can be used to minimize the dispersion degree of the predicted traffic corresponding to each link after adjustment.
[0033] For example, exhaust all possible binding relationships between the links and the hash value, and calculate the discrete degree of the predicted traffic corresponding to each link under the possible binding relationship between each of the links and the hash value. The binding relationship between the link and the hash value is adjusted with the least discrete degree.
[0034] For another example, the greedy algorithm is used to adjust the binding relationship between the link and the hash value, so as to minimize the dispersion degree of the predicted traffic corresponding to each link after adjustment. For example, calculate the average of all link traffic, starting from the link with the most traffic exceeding the average and the link with the most below the average, and by adjusting the hash value bound by the two links above, to the average. value adjusts the predicted traffic for both links to minimize the sum of the absolute values of the deviations of the two links from the mean. Then start the adjustment from the two links with the largest deviation between the traffic flow and the average value among the adjusted links, and repeat the above steps until the quantification standard of the degree of dispersion cannot be further optimized or reaches the pre-specified expected threshold.
[0035] Those skilled in the art may know that other methods capable of dealing with optimization problems, such as simulated annealing, etc., can be used to minimize the dispersion of the predicted traffic corresponding to each link after adjustment. Taking simulated annealing as an example, simulated annealing is a general probability algorithm used to find the optimal solution of a proposition in a large search space. The simulated annealing algorithm can be decomposed into three parts: solution space, objective function and initial solution. In the embodiment of the present invention, the solution space is the set space of the binding relationship between all links and hash values, the objective function is the discrete degree index, and the initial solution is the binding between the link and the hash value before adjustment established relationship. Each iteration of simulated annealing can be decomposed into four steps. The first step is to generate a new solution located in the solution space from the current solution by a generating function. In the embodiment of the present invention, the new solution is to adjust the binding relationship of only one hash value. produced. The second step is to calculate the difference of the objective function corresponding to the new solution, that is, the variation of the discrete degree of the new solution relative to the previous iteration. The third step is to judge whether the new solution is accepted. The most commonly used acceptance criterion is the Metropolis criterion: if Δt′<0, accept S′ as the new current solution S, otherwise accept S′ with probability exp(-Δt′/T) As the new current solution S, where Δt is the objective function difference and T is the objective function. The fourth step is to replace the current solution with the new solution when it is confirmed to be accepted. At this point, the current solution achieves one iteration. On this basis, the next round of trials can be started. When the new solution is judged to be discarded, the next round of testing is continued on the basis of the original current solution.
[0036] figure 2 This is a block diagram of a load sharing adjustment apparatus in an embodiment of the present invention. The device includes:
[0037] The measurement module 202 is used to measure the traffic of each link.
[0038]The calculation module 204 is configured to estimate the predicted traffic corresponding to each hash value according to the total number of hash values bound by each link and the traffic of each link, where the total number of hash values is greater than the number of links.
[0039] For example, the predicted traffic corresponding to each hash value is estimated according to the method that each hash value corresponding to the same link equally shares the link traffic. For example, there are 4 links, 16 hash values, and the traffic of each link is T 1 , T 2 , T 3 , T 4 , the hash values bound by each link are 3, 3, 4, and 6 respectively, then the predicted traffic of the 3 hash values is estimated to be T 1 /3, the predicted flow of 3 hash values is T 2 /3, the predicted flow of 4 hashes is T 3 /4, the predicted flow of 6 hashes is T 4 /6.
[0040] The adjustment module 206 is configured to adjust the binding relationship between the link and the hash value according to the predicted traffic corresponding to each hash value, so as to improve the uniformity of load sharing.
[0041] The method for adjusting the binding relationship between the link and the hash value is the same as figure 1 The method of 106 in the described embodiment is the same.
[0042] Combine the following image 3 and Figure 4 Describe one embodiment of the present invention:
[0043] image 3 This is a flowchart of a load sharing adjustment method in another embodiment of the present invention. The method includes:
[0044] 302. Measure the traffic of each link.
[0045] The method for measuring the traffic of each link in this embodiment is the same as figure 1 The method of measuring the traffic of each link in the illustrated embodiment is the same.
[0046] 304. Estimate the predicted traffic corresponding to each hash value according to the binding relationship between the link and the hash value, the traffic of each link, and the historical predicted traffic corresponding to each hash value.
[0047] For example, there are 4 links and 16 hash values, and the current traffic of each link is T. 01 , T 02 , T 03 , T 04. The binding relationship between each link and the hash value is T 01 :H 1 -H 3 , T 02 :H 4 -H 6 , T 03 :H 7 -H 10 , T 04 :H 11 -H 16. The predicted traffic corresponding to each hash value at the last measurement is HL 1 -HL 16. Then the estimated flow corresponding to H1 is T 01 ·[HL 1 /(HL 1 +HL 2 +HL 3 )], and so on, to calculate the predicted traffic corresponding to each hash value during this measurement.
[0048] Those of ordinary skill in the art may know that other estimation methods for citing historical records can also be used to estimate the predicted traffic corresponding to each hash value. For example, refer to the predicted traffic corresponding to each hash value estimated in the past three measurements, weight the predicted traffic estimated in the past three measurements corresponding to each hash value according to a certain weight, and estimate each hash value bound by each link. The weighted proportion of the predicted traffic in the traffic of each link. The weight may be a preset fixed value, or may be a variable adjusted according to the traffic of each link, the traffic stability of each link, the measurement time interval or other parameters. Still taking the aforementioned 4 links and 16 hash values as an example, referring to the predicted traffic corresponding to each hash value estimated in the past two measurements, the weights of the past two measurements are Q respectively. 1 and Q 2. The predicted traffic corresponding to each hash value in the past two measurements is H 1 L 1 -H 1 L 16 and H 2 L 1 -H 2 L 16. Then the estimated flow corresponding to H1 is T 01 ·[(Q 1 ·H 1 L 1 +Q 2 ·H 2 L 1 )/(Q 1 ·H 1 L 1 +Q 1 ·H 1 L 2 +Q 1 ·H 1 L 3 +Q 2 ·H 2 L 1 +Q 2 ·H 2 L 2 +Q 2 ·H 2 L 3 )], and so on, to calculate the predicted traffic corresponding to each hash value during this measurement.
[0049] 306. Adjust the binding relationship between the link and the hash value according to the predicted traffic corresponding to each hash value, so as to improve the uniformity of load sharing.
[0050] The method for adjusting the binding relationship between the link and the hash value according to the predicted traffic corresponding to each hash value in this embodiment is the same as figure 1 In the illustrated embodiment, the method for adjusting the binding relationship between the link and the hash value according to the predicted traffic corresponding to each hash value is basically the same.
[0051] Figure 4 It is a block diagram of a load sharing adjustment apparatus in another embodiment of the present invention. The device includes:
[0052] The measurement module 402 is used to measure the traffic of each link.
[0053] The recording module 404 is configured to record the historical predicted traffic corresponding to each hash value.
[0054] The calculation module 406 is configured to estimate the predicted traffic corresponding to each hash value according to the binding relationship between the link and the hash value, the traffic of each link, and the historical predicted traffic corresponding to each hash value. The total number of hash values is greater than the number of links.
[0055] The method for estimating the predicted traffic corresponding to each hash value is the same as image 3 The method 304 in the described embodiment is the same.
[0056] The adjustment module 408 is configured to adjust the binding relationship between the link and the hash value according to the predicted traffic corresponding to each hash value, so as to improve the uniformity of load sharing.
[0057] The method for adjusting the binding relationship between the link and the hash value is the same as image 3 The method 306 in the described embodiment is the same.
[0058] By adopting the technical solution of the embodiment of the present invention, because the total number of hash values is greater than the number of links, the binding relationship between links and hash values can be adjusted according to the traffic of each link, so as to improve the evenness of load sharing of the flow-by-flow routing forwarding technology sex.
[0059] Those of ordinary skill in the art can understand that all or part of the steps in the methods of the above embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium, and the storage medium can be a ROM /RAM, disk or disc, etc.
[0060] The above description is only a preferred embodiment of the present invention, but the protection scope of the present invention is not limited to this. Substitutions should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.
