AI-Based Dynamic Architecture Sizing for Virtual Machines
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computer system architecture sizing is complex and dependent on subject matter expert (SME) insights, lacking empirical data, and varies by vertical industry and workload periods, making it challenging to accurately size system resources such as CPU and memory, especially during peak and slack times, and failing to correlate with service-level agreement (SLA) requirements for high availability.
Innovation Solution
A method using AI-based statistical analysis to dynamically size computer system architecture by capturing real-time data from virtual machines and bare-metal servers, calculating mean, maximum, and standard deviation (SD) for CPU and memory usage, and creating a reference architecture for each vertical industry, optimizing system architecture based on empirical data and time zone considerations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional SME-dependent architecture sizing is used, then expert knowledge can be applied, but the process becomes complex and inconsistent across different vertical industries
Solution Approach 1:
The patent replaces the manual, expert-dependent architecture sizing process with an automated AI-based system that uses statistical modeling and machine learning algorithms. The AI model analyzes historical performance data, workload characteristics, and system configurations to automatically determine optimal architecture specifications, eliminating the need for manual SME intervention while improving consistency and accuracy across different vertical industries.
Solution Approach 2:
The system transforms the architecture sizing problem from a qualitative expert judgment process into a quantitative parameter-based analysis. By collecting and analyzing multiple parameters including CPU utilization, memory usage, storage I/O patterns, network throughput, and workload characteristics, the AI model can precisely calculate optimal architecture configurations. This parameter-driven approach enables consistent sizing decisions across different industries and eliminates variability introduced by different SMEs.
2Adaptability or versatility
If static architecture sizing is used, then implementation is simpler, but the system cannot adapt to varying workloads during peak and slack periods
Solution Approach 1:
The patent implements dynamic architecture sizing that adapts to varying workload conditions. The AI model continuously monitors system performance metrics and workload characteristics, then adjusts architecture recommendations in real-time or near-real-time. This enables the system to optimize resource allocation during peak periods while reducing capacity during slack periods, providing workload adaptability without requiring complex manual reconfiguration. The dynamic nature is achieved through automated feedback loops and continuous learning from operational data.
Solution Approach 2:
The system incorporates feedback mechanisms where actual system performance and workload patterns are continuously fed back into the AI model. This feedback loop allows the model to learn from real-world operations and refine its architecture sizing recommendations. The feedback-driven approach enables the system to adapt to changing conditions automatically, adjusting architecture configurations based on observed performance patterns without requiring complex manual intervention or pre-defined static rules.
3Measurement precision
If comprehensive data collection is performed, then more accurate statistical analysis can be done, but data processing time and computational resources increase
Solution Approach 1:
The patent implements preliminary data collection and preprocessing steps that prepare data for future analysis. Historical performance data, workload characteristics, and system configuration information are collected and stored in structured formats in advance. This preliminary action enables the AI model to perform rapid analysis when architecture sizing is needed, as the raw data processing work has already been completed. The system maintains pre-processed datasets that can be quickly queried and analyzed without requiring extensive real-time computation.
Solution Approach 2:
The system creates simplified copies or representations of complex data sets for analysis purposes. Instead of processing all raw operational data in real-time, the AI model works with aggregated statistics, summary metrics, and representative samples that capture the essential characteristics of the workload. These copied representations maintain the statistical accuracy needed for precise analysis while dramatically reducing the computational burden and processing time required.
Data Source
AI summary
In an approach to improve the field of architecture generation by dynamically sizing computer system architecture requirements for virtual machines. Embodiments define static parameters and dynamic parameters for customer consuming resources of the computer system architecture and store data of the static parameters and dynamic parameters in data warehouse database (DWDB) tables. Further, embodiments compute, using the data of the DWDB tables, calculate the minimum, maximum, mean and standard deviation (SD) for the user count and the CPU and memory usage, and update the DWDB tables based on the minimum, maximum, and SD values per customer. Additionally, embodiments classify an architecture size associated with each of the customers, create a reference architecture for each of the one or more vertical industries and the architecture size, and optimize, by an analytical database, the computer system architecture provided to one or more customer based on the computed data of the DWDB for the one or more vertical industries.