Cloud workflow task execution time prediction method based on multi-dimensional feature fusion

Abstract

The invention discloses a cloud workflow task execution time prediction method based on multi-dimensional feature fusion; a fusion feature prediction model based on multiple dimensions is built through employing the strategies of feature extraction, fusion and screening, and the precise prediction of task execution time is achieved. The cloud workflow task execution time prediction requirement driven by big data can be met to a certain extent, and a new thought and way are provided for accurate prediction of the cloud workflow task execution time.

Description

technical field [0001] The invention belongs to the technical field of cloud computing, and in particular relates to a cloud workflow task execution time prediction method based on multi-dimensional feature fusion. Background technique [0002] With the maturity and wide application of cloud computing technology, the pay-per-use model and elastic expansion capability of resources in cloud data centers provide a fast, flexible and scalable deployment environment for distributed management and efficient execution of large-scale scientific applications . More and more scientists use workflows to build and manage complex scientific applications, and deploy or gradually migrate their scientific workflows to cloud data centers for execution. The key to efficient execution of scientific applications is workflow scheduling, and the design of scheduling algorithms, scheduling performance, and implementability of scheduling schemes directly depend on the estimation accuracy of task e...

AI Technical Summary

Problems solved by technology

[0004] However, most of the existing task execution time prediction methods lack targeted feature extraction strategies, and it is difficult to fully mine and utilize Categorical data and Numerical data at the same time.

A small number of task execution time prediction methods have feature extraction capabilities, but in the process of using the extracted features for prediction, no corresponding fusion strategy is designed, which is easy to introduce noise or extracted error information, making it difficult to achieve accurate task execution time predict

[0005] In addition, most of the existing task execution time prediction algorithms are based on simulation data, but the simulation environment is difficult to simulate the dynamic access and withdrawal of cloud resources and the instability of the network environment, and it is not easy to reflect the massive task request responses and resources in the real cloud environment. Big data features brought about by elastic scaling

Therefore, in the actual application process, there are the following deficiencies: First, the ability to analyze categorical data is insufficient, and it is difficult to solve the problem of too large search space caused by the high-dimensional sparse characteristics of categorical data, making it difficult for the model to find the optimal solution, thus affecting Prediction accuracy; the second is the lack of sufficient nonlinear expression ability, it is difficult to deal with the problem of reducing feature differences caused by the low-dimensional dense characteristics of Numerical data, making model learning difficult and easy to fall into local optimum, thus affecting the prediction effect; the third is the lack of An effective feature fusion strategy directly learns the extracted features, causing the model to easily learn invalid or wrong information, which affects the accuracy of the prediction results

Therefore, the existing task execution time prediction methods lack effective Categorical and Numerical feature extraction capabilities and feature fusion strategies, and it is difficult to meet the cloud workflow task execution time prediction requirements in the big data environment.

Images