Non-deterministic separation web server cluster scaling method

Abstract

The nondeterministic separation web server cluster resource scaling method comprises the following three steps: step 1, constructing a nondeterministic separation system model and a decision-making model, establishing an accurate mathematical model for a system as a system model in an ideal nondeterministic separation environment, and establishing a decision-making model for the decision-making model in the ideal nondeterministic separation environment; utilizing an Aloy modeling tool to obtain a system state transition matrix to construct a decision model; step 2, processing environmental uncertainty factors, defining compensation coefficients to describe the uncertainty factors caused by environmental changes and modeling errors, and dynamically estimating the distribution of the compensation coefficients by utilizing Kalman filtering; and step 3, solving an adaptive scaling strategy, and solving an optimal resource scaling strategy on the system decision model obtained in the step 1by integrating a rolling optimization thought and a game theory method of model prediction control. According to the method, the thought of attention separation is adopted, system model design and nondeterministic factor processing are separated, and the influence of nondeterministic factors on telescopic decision making is processed through independent steps.

Description

technical field [0001] The invention relates to a non-deterministic separation web server cluster resource stretching method, which belongs to the field of computer software development. Background technique [0002] Web server cluster scaling technology is one of the main ways to improve website performance. The server cluster scaling method needs to decide how to effectively adjust the number of servers to cope with load changes when the web service is running. Software self-adaptation is an effective method to realize the scaling of WEB server clusters. Server clusters continuously monitor system and load changes during operation, and dynamically make scaling strategies to ensure that the functional and non-functional requirements of the system are met. [0003] At present, forward-looking software adaptive methods are widely used in scenarios where adaptive decision-making is delayed. Nowadays, there are two mainstream forward-looking software adaptive methods that can...

AI Technical Summary

Problems solved by technology

The PLA method can only solve the non-deterministic factors that have been modeled, while the non-deterministic factors that have not been modeled, such as system model errors, environmental prediction errors, etc., will have a greater impact on adaptive decision-making

At the same time, the PLA method models the probabilistic changes of the environment as a Markov decision process, which consumes a lot of time during runtime.

Images