Service Cluster Workload Prediction Using Attention-Weighted Graphs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing workload prediction methods for services in a serverless architecture lack accuracy, particularly when relying on historical load information, which may be privacy data, and do not effectively utilize the relationships between multiple services in a service cluster.
Innovation Solution
A method involving obtaining indicator sequences for system status and workload, calculating correlation coefficients to construct similarity graphs, processing these graphs with an attention mechanism, and performing graph convolutional processing to aggregate representations, thereby improving prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If workload prediction is performed based only on historical load information of a single service, then the prediction process is simple, but the prediction accuracy is insufficient
Solution Approach 1:
The patent merges historical load information from multiple services in the service cluster to perform workload prediction. Instead of predicting based on a single service's history, the method combines data from n services, where each service contributes its historical load information. This merging approach improves prediction accuracy by capturing cluster-wide patterns while managing complexity through structured data integration and feature engineering.
2Measurement precision
If only single-service historical data is used for prediction, then data privacy is maintained, but prediction accuracy cannot satisfy requirements
Solution Approach 1:
The patent introduces an intermediary aggregation layer that combines historical load information from multiple services without exposing individual service's private data. The system uses structured feature engineering and distributed computing approaches where services contribute anonymized historical data to a cluster-level model. This intermediary mechanism enables accurate prediction while preserving data privacy through controlled information sharing and aggregation.
3Productivity
If resources are statically allocated to services, then resource allocation is simple and stable, but resource utilization and service experience deteriorate under varying workloads
Solution Approach 1:
The patent performs preliminary workload prediction using the trained model to forecast future service workloads before actual resource allocation decisions are made. This advance prediction enables the system to proactively adjust resource allocation in response to predicted workload changes, improving resource utilization and service experience. The preliminary action of prediction precedes the resource allocation action, creating a responsive dynamic allocation system.
Data Source
AI summary
Embodiments of this specification provide a workload prediction method and apparatus for a service in a service cluster. The method includes: obtaining n+1 indicator sequences respectively corresponding to services in a same historical time period, which include n status indicator sequences corresponding to n system status indicators and a load indicator sequence corresponding to a workload indicator; calculating a correlation coefficient of a same indicator between different services based on the n+1 indicator sequences corresponding to the services, and constructing a similarity graph corresponding to the same indicator, to obtain n+1 similarity graphs; processing the n+1 similarity graphs by using an attention mechanism, to obtain an aggregated similarity graph; performing, by using a first convolutional network, graph convolutional processing on an initial representation of each service in the aggregated similarity graph, to obtain an aggregated representation corresponding to each service; and performing workload prediction based on the aggregated representation.