【摘要】 随着云计算数据量的迅速增大,对资源管理策略的要求也越来越高,而负载的预测在云资源优化配置中起着举足轻重的作用。针对云计算的负载变化兼有短期动态不确定性与长期统计规律的稳定性,利用经过改进的Seq2seq模型,可通过采集一段时间内的历史负载信息,对负载时间序列数据进行建模,以实现较为准确的未来一段时间的负载预测,并通过dropout来提高模型的泛化能力。经实验分析改进后,Seq2seq模型较原Seq2seq模型在资源负载较长期预测上的准确率有很大提升。更多还原

【Abstract】 With the rapid increase of the amount of cloud computing data, the requirements for resource management strategies are becoming higher and higher, and the load forecasting plays an important role in the optimal allocation of cloud resources. Aiming at the stability of cloud computing load with both shortterm dynamic uncertainty and stability of long-term statistical laws, the improved Seq2seq model can be used to model the load time series data by collecting historical load information over a period of time, so as to achieve a more accurate load forecast for a period of time in the future. The generalization ability of the model is improved by dropout. After the improvement of experimental analysis, the accuracy of the Seq2seq model in the long-term resource load prediction has been greatly improved compared with the original Seq2seq model.更多还原