CPI index-based cloud service performance intelligent prediction method and device

Abstract

The invention discloses a CPI index-based cloud service performance intelligent prediction method and device. The method comprises the steps of obtaining various time series data about batch processing tasks and cloud services and time spans thereof; pre-processing the time sequence data based on the time span to obtain a pre-processing result; inputting the preprocessing result into a pre-trainedSnapLSTM prediction model to obtain a CPI prediction value; and carrying out weighted average on the CPI prediction value according to a pre-acquired verification error weight to obtain a CPI prediction result. According to the CPI index-based cloud service performance intelligent prediction method and device provided by the invention, modeling prediction is carried out on the performance of thecloud service under mixed deployment starting from the CPI index, the prediction effect is good, and an operation and maintenance engineer can be effectively helped to determine the trend of the service performance.

Description

technical field [0001] The present invention relates to the technical field of cloud platforms for hybrid deployment, and more specifically relates to a method and device for intelligently predicting cloud service performance based on CPI indicators. Background technique [0002] In the current Internet data center, in order to improve the overall resource utilization of the cluster, the common method is to combine batch processing tasks (such as: MapReduce, Spark tasks, etc.) services, etc.) for hybrid deployment, such as figure 2 As shown, in the data center, tens of thousands of servers are managed, and various numbers of container instances and batch processing tasks are run on each server. Latency-sensitive cloud services are often run in containers, while most batch tasks are not containerized. [0003] Compared with batch processing tasks, containerized cloud services have certain regularities in resource usage and performance fluctuations, which are better predicta...

AI Technical Summary

Problems solved by technology

However, it is still not enough to deal with the performance interference caused by various types of resource preemption, such as memory bandwidth preemption

[0005] Existing methods use the collection and analysis of upper-layer application QoS (Quality of Service) data to measure the performance of cloud services, which is inconvenient for the operation and maintenance of data centers, and QoS (Quality of Service) such as response time ) indicators, due to the influence of various factors outside the cluster, cannot accurately reflect the performance of cloud services, resulting in low prediction accuracy

Images