Throughput Estimation via Virtualized Load Vector Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Estimating the throughput of information processing functionalities executed concurrently on general purpose computers is challenging due to performance degradation from shared resources, and requires separate performance evaluations for each combination of software and hardware, which becomes impractical with increasing types of hardware and software.
Innovation Solution
A throughput estimation device using a feed-forward network of calculation units that models software processes and hardware resources, allowing for the estimation of throughput by inputting a load vector and outputting a result value, while reducing the number of required performance evaluations through optimization and objective function minimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If separate performance evaluations are performed for each combination of software and hardware, then measurement precision is improved, but device complexity and time consumption increase significantly
Solution Approach 1:
The patent creates a virtualized performance evaluation system that copies the essential characteristics of physical hardware and software combinations into a virtual environment. By constructing virtual machines and virtual networks that replicate real system behaviors, the system can perform multiple performance evaluations simultaneously without requiring separate physical testbeds for each combination, thus reducing the number of evaluations needed while maintaining measurement precision.
Solution Approach 2:
The patent implements a universal performance evaluation platform that can assess multiple software applications across different hardware configurations through a single integrated system. The virtualized infrastructure allows the same evaluation framework to handle diverse scenarios (different VMs, containers, and physical devices) without requiring separate specialized measurement systems for each combination type.
2Measurement precision
If separate performance evaluations are performed for each combination of software and hardware, then throughput estimation accuracy is improved, but loss of time increases significantly
Solution Approach 1:
The patent performs preliminary actions by pre-configuring virtual machine images, virtual network topologies, and performance evaluation templates before actual measurements are needed. These pre-prepared virtual environments can be rapidly deployed and configured for different software-hardware combinations, eliminating the need to build entire testbeds from scratch for each evaluation scenario and significantly reducing measurement time.
Solution Approach 2:
The patent enables continuous performance evaluation through parallelized virtual testing. Multiple performance evaluations can run simultaneously in the virtual environment without interfering with each other, as virtual machines and networks can be isolated and executed concurrently. This continuous parallel processing allows the system to complete many more evaluations in the same time frame compared to sequential physical testing.
3Adaptability or versatility
If general purpose computing hardware of different specifications is considered, then adaptability is improved, but device complexity increases
Solution Approach 1:
The patent utilizes parameter changes by allowing virtual machine configurations to be dynamically adjusted to represent different hardware specifications. Instead of managing complex physical testbeds for each hardware type, the system changes virtual parameters (CPU cores, memory size, storage capacity, network bandwidth) to simulate various hardware configurations, thereby achieving broad adaptability while keeping the evaluation system itself simple and manageable.
4Adaptability or versatility
If concurrent execution of multiple information processing software is considered, then adaptability is improved, but device complexity increases
Solution Approach 1:
The patent introduces virtualized infrastructure as an intermediary layer between the software applications and physical hardware. This virtualization layer (including virtual CPUs, memory managers, and network virtualizers) acts as a mediator that abstracts away the complexities of resource sharing and hardware interactions. By evaluating software performance through this intermediary virtual layer rather than directly on physical hardware, the system can handle multiple software combinations without proportionally increasing evaluation system complexity.
Data Source
AI summary
The throughput estimation device 10 comprises a storage 11 which stores tuples, each including an evaluation formula and parameter data, and a calculation device constructing unit 12 which constructs a calculation device according to a configuration including data associated with software processes, data associated with hardware resources, data associated with interconnections between a software process and a hardware resource, and the tuples, wherein the calculation device inputs a load vector describing input information load to the software processes, and outputs a result value.