Conditional Resource Throttles for Database Workload Concurrency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Modern priority scheduling technologies are ineffective in managing resource distribution across multiple subject areas and categories simultaneously, particularly when the number of subject areas is large and disparate, and struggle to manage resource allocation at granularities beyond small time intervals, such as 60 seconds, leading to inefficiencies and resource wastage in database management systems.
Innovation Solution
A system and method for workload management that dynamically adjusts system resource consumption against a target level by controlling workload concurrency levels, using conditional resource throttles to manage resource distribution across subject areas and categories, allowing unclaimed resources to be utilized by active workloads, and monitoring consumption to ensure optimal system utilization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If priority scheduling technologies are used to manage resource distribution, then resource allocation within small time intervals is improved, but the ability to manage resource distribution across multiple subject areas and categories simultaneously deteriorates
Solution Approach 1:
The patent segments resource management into multiple independent dimensions: subject-area dimensions (e.g., division, application, country) and category-of-work dimensions (e.g., priority, workload type). Each dimension is managed separately through independent throttle mechanisms, allowing the system to handle complex multi-dimensional resource distribution without the limitations of traditional single-dimension priority scheduling.
2Speed
If priority schedulers manage resources within small time intervals, then short-term resource distribution is improved, but the ability to manage resource distribution at larger time slices deteriorates
Solution Approach 1:
The patent adds a time-dimension to resource management by introducing averaging intervals that can span from seconds to hours or days. This allows the system to simultaneously manage short-term resource allocation (for immediate responsiveness) and long-term resource distribution (for overall fairness and utilization), resolving the contradiction between fast response and sustained resource management.
3Productivity
If contemporary priority schedulers are used, then resource distribution within current time interval is improved, but fair management of large number of divisions with varying weights deteriorates
Solution Approach 1:
The patent implements feedback mechanisms that continuously monitor resource consumption across all divisions and categories. The system calculates actual resource usage, compares it against target consumption levels, and dynamically adjusts throttle settings to ensure fair distribution. This feedback loop enables precise control over resource allocation fairness even when managing hundreds of divisions with widely varying weight preferences.
Data Source
AI summary
A system, method, and computer-readable medium that facilitate workload management in a computer system are provided. A workload's system resource consumption is adjusted against a target consumption level thereby facilitating maintenance of the consumption to the target consumption within an averaging interval by dynamically controlling workload concurrency levels. System resource consumption is compensated during periods of over or under-consumption by adjusting workload consumption to a larger averaging interval. Further, mechanisms for limiting, or banding, dynamic concurrency adjustments to disallow workload starvation or unconstrained usage at any time are provided. Disclosed mechanisms provide for category of work prioritization goals and subject-area resource division management goals, allow for unclaimed resources due to a lack of demand from one workload to be used by active workloads to yield full system utilization at all times, and provide for monitoring success in light of the potential relative effects of workload under-demand, and under/over-consumption management.