Chebyshev Node Interpolation for Service Demand Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional performance prediction models for multi-tiered web applications are inaccurate due to their inability to capture varying service demands, leading to incorrect throughput and response time predictions at higher workloads, as they assume constant service demands or average values.
Innovation Solution
The method involves identifying a range of concurrencies and computing Chebyshev nodes for interpolation to optimize load testing points, generating service demand samples, and using spline or linear interpolation to create an array of interpolated service demands, which are then incorporated into a queueing network or simulation model to predict throughput and response times.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If conventional Mean Value_analysis models use constant or average service demands, then the modeling process is simple, but the prediction accuracy deteriorates at higher workloads
Solution Approach 1:
The patent applies dynamics by transitioning from static constant service demand models to dynamic models where service demands vary with workload. The system computes service demands at multiple concurrency levels and uses interpolation to capture the dynamic behavior of service demands across different load conditions, enabling accurate predictions at higher workloads while maintaining manageable model complexity through structured computation.
Solution Approach 2:
The patent segments the performance prediction problem by dividing the concurrency range into multiple discrete levels and computing service demands separately at each level. This segmentation allows the system to capture varying service demand characteristics at different workloads rather than using a single average value, thereby improving prediction accuracy without overwhelming complexity.
2Measurement precision
If more load testing points are used to capture service demand variations, then prediction accuracy improves, but the number of testing points and computational complexity increases
Solution Approach 1:
The patent applies parameter changes by transforming the service demand computation into a function of concurrency levels. Instead of fixed testing points, the system computes service demands at variable concurrency levels and uses interpolation parameters to estimate service demands at any given concurrency, reducing the need for exhaustive testing points while maintaining accuracy.
Solution Approach 2:
The patent introduces interpolation as an intermediary mechanism between discrete service demand measurements and continuous performance prediction. The interpolation function acts as a mediator that estimates service demands at untested concurrency levels based on measured values, reducing the number of required load testing points while maintaining prediction accuracy.
Data Source
AI summary
Systems and methods for service demand based performance prediction with varying workloads are provided. Chebyshev nodes serve as optimum number of load testing points to minimize polynomial interpolation error rates. Chebyshev nodes are identified for a pre-determined range of concurrencies in the application under test. An array of interpolated service demands at the Chebyshev nodes, when integrated with a modified multi-server Mean Value Analysis (MVA) technique provides superior throughput and response time predictions.