Predictive Microservice Autoscaling to Reduce Allocation Delays
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing microservice-based applications experience inefficiencies and delays due to the time required for resource allocation and deallocation during workload fluctuations, leading to unstable system performance and potential errors.
Innovation Solution
Implement proactive resource scaling by predicting future workloads using machine learning models and adjusting resources accordingly before workload changes occur, utilizing a centralized cache to store workload data and a resource scaling component to manage resource allocation based on predicted workloads.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If reactive resource scaling is implemented (scaling up/down based on current workload thresholds), then resource costs are reduced, but processing time increases due to allocation/deallocation delays
Solution Approach 1:
The patent applies preliminary action by using machine learning models to predict future workload patterns and proactively scaling resources before actual workload changes occur. The system analyzes historical data and forecasts to anticipate peak demand periods, initiating resource allocation in advance rather than reacting after thresholds are breached, thereby eliminating the time delay between detecting high demand and allocating resources.
2Adaptability or versatility
If resource scaling is implemented for each microservice independently, then each service can be optimized, but system stability decreases due to cumulative time delays across multiple services
Solution Approach 1:
The patent merges the resource scaling decisions across multiple microservices by introducing a centralized workload predictor that considers the entire service architecture. Instead of independent reactive scaling that causes cumulative delays, the system combines predictions from multiple services into a coordinated scaling strategy, allowing resources to be allocated system-wide based on aggregate future workload patterns rather than individual service thresholds.
Data Source
AI summary
Systems and methods include reception, at a first microservice of a microservice-based application, of an indicator of a workload of an entry microservice of the microservice-based application, determination, based on the indicator of the workload, of an estimated future workload of the first microservice, and re-allocation of computing resources to the first microservice based on the estimated future workload.