Dynamic Capacity Management for Parallel Computing Resources
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current dynamic capacity management in data centers focuses primarily on fluctuations in request rate, neglecting other significant changes such as request size and server efficiency variations, leading to inefficient server utilization and high energy waste.
Innovation Solution
The AutoScale dynamic capacity management method introduces a state-change component with a delay timer and a robustness-control component that dynamically adjusts server states based on idle time and system load, ensuring efficient capacity provisioning without predicting future request rates or sizes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If servers are over-provisioned to meet peak load and maintain SLAs, then response time service level agreements are met, but energy consumption increases significantly with only 10-30% server utilization
Solution Approach 1:
The patent implements dynamic capacity management that automatically adjusts server states between active and idle based on real-time load conditions. The system monitors workload metrics and dynamically transitions servers to appropriate states, ensuring SLA compliance during peak load while minimizing energy consumption during low-utilization periods through automated state changes.
Solution Approach 2:
The system changes the operational parameter of servers by transitioning them between different states (active, idle, suspended) based on load conditions. This parameter change allows the same physical server to operate at different power levels and performance capacities, optimizing the balance between SLA compliance and energy efficiency.
2Use of energy by moving object
If servers are dynamically switched to idle state to reduce energy consumption, then energy waste is reduced, but setup costs increase due to high reactivation overhead
Solution Approach 1:
The patent implements a warm-up mechanism that performs preliminary actions to prepare idle servers for rapid reactivation. By pre-warming cache, pre-loading configuration, and maintaining core processes in a suspended state rather than fully shutting down, the system reduces the setup time penalty when servers need to be brought back online, making dynamic state transitions more cost-effective.
3Speed
If reactive approaches are used to respond to current request rate, then responsiveness to load changes is improved, but inability to predict future request rates leads to SLA violations during sudden load spikes
Solution Approach 1:
The patent implements a feedback-based control system that continuously monitors multiple workload metrics including request rate, response time, and server utilization. This feedback loop allows the system to detect load trends and adjust server states proactively, combining reactive responsiveness with predictive capacity management to maintain SLA compliance during sudden load changes.
4Reliability
If predictive approaches are used to anticipate future request rates, then capacity planning is improved, but complexity increases and cannot account for unpredictable load patterns
Solution Approach 1:
The patent implements a self-service capacity management system that automatically monitors, analyzes, and adjusts server states without external intervention. The system uses straightforward metrics collection and threshold-based decision rules rather than complex predictive algorithms, allowing it to adapt to unpredictable load patterns while maintaining manageable system complexity through automated self-adjustment.
Data Source
AI summary
A dynamic capacity management policy for multi-paralleled computing resources (e.g., application servers, virtual application servers, etc.) that includes one or more of a state-change component, a load-balancing component, and a robustness-control component. The state-change component delays the release (e.g., powering down of a physical server, removal from a virtual-server lease, etc.) of each computing resource for a set amount of time. The load-balancing component can work in conjunction with the state-change component to reduce the number of idle computing resources by distributing incoming requests in a manner that keeps the already-processing computing resources as full of requests as possible. The robustness-control component scales capacity as a function of the current number of requests within the system of computing resources to account for variations other than request rate, such as request size, reduced processor frequency, network slowdowns, etc., that affect processing capacity.