Software system performance optimization method based on generative countermeasure network

Abstract

The invention discloses a software system performance modeling and optimization method based on a generating countermeasure network, which mainly solves the problems of high time cost, too much difficulty in training, easy collapse of training network and limited optimization space in the prior art. The implementation scheme is as follows: 1) acquiring the feature sample set of the software system; 2) fixing that hardware environment, configuring the software system, carry out performance testing according to the characteristic sample set, obtaining throughput or time delay, carry out pretreatment on the throughput or time delay, and sequentially carrying out single thermal coding and normalization to obtain structured data; 3) at that start of each iteration, randomly selecting half of the structured data from the step 1 as a training sample, iteratively train the generated antagonism network to obtain an optimized sample; 4) comparing the performance of the training sample and the optimization sample to verify the optimization effect. The invention reduces the time cost, improves the robustness and stability of the network, has obvious optimization effect, and can be used for theprocessing of the Internet and the big data.

Description

technical field [0001] The invention belongs to the technical field of computers, and in particular relates to a method for optimizing the performance of a software system, which can be used for Internet and big data processing. Background technique [0002] At present, in the era of rapid development of the Internet and big data, with the increasing amount of data, how to optimize the performance of the software system under certain conditions of the hardware system has been widely discussed. Some of the currently popular distributed software systems include Spark, Hive, HBase, and Kafka. Therefore, how to conduct refined performance modeling and optimization of various software systems is still a hot issue in the industry and academia. [0003] A data-aware Spark configuration is disclosed in the patent document "A data-aware Spark configuration parameter automatic optimization method" (application number: 201611182310.5 application date: 2016.12.20 publication number: CN...

AI Technical Summary

Problems solved by technology

The disadvantage of this method is that it is necessary to evaluate the impact of each configuration on the performance of the distributed memory computing framework Spark cluster in the actual environment, as the training set of the random forest model, which wastes a lot of time and cost

The disadvantage of this method is that with the improvement of the complex network fitting effect, the complexity of the network is also increasing, and the difficulty of training is increasing, which leads to problems such as network crashes.

[0005] Therefore, for the performance optimization of the software system, there are still some problems, including the curing and aging of ideas, and the traditional method of solving the weight of each feature to the final performance by various methods cannot be avoided, and the innovation level has not made much progress.

At the same time, due to the consistency of the basic ideas of the method, it is impossible to make great progress in performance improvement and encounter bottlenecks

When using traditional machine learning methods, the time cost of the algorithm is too large due to the need for too many samples

Or when using some new machine learning algorithms, such as complex networks, although they can achieve better results, they are difficult to train, easy to crash, and not suitable for some practical scenarios

Images