Computation Resource Control via Dual Queue State Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing computation resource control systems face inefficiencies in managing and allocating resources to meet varying computation requests, leading to increased latency and operational costs due to ineffective queue management and resource state transitions.
Innovation Solution
A computation resource control apparatus with an activation unit, first and second queue managing units, and an allocating unit that activates resources, allocates them based on request type, and transitions them between active, executable, and stop states to optimize resource utilization and reduce latency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If computation resources are activated immediately upon request, then response speed is improved, but operational costs increase due to excessive resource activation
Solution Approach 1:
The system performs preliminary actions by pre-activating computation resources before they are actually needed. When a computation request arrives, resources are already in an activated state or can be quickly activated from a warm state, rather than being activated from a completely stopped state. This reduces the activation time (improving response speed) while still allowing resources to be stopped during low-demand periods (controlling operational costs).
Solution Approach 2:
The system dynamically adjusts the state of computation resources based on real-time demand conditions. Resources transition between stopped, activating, activated, and stopping states according to the arrival of computation requests and system load conditions. This dynamic state management allows the system to optimize between response speed and operational costs by activating resources only when necessary while maintaining the ability to quickly respond to requests.
2Speed
If computation resources are kept in activated state, then processing speed is improved, but resource utilization efficiency deteriorates
Solution Approach 1:
The system implements dynamic state transitions for computation resources, moving between stopped, activating, activated, and stopping states based on real-time demand. This allows resources to be activated quickly when needed (maintaining processing speed) while being stopped during low-demand periods (improving resource utilization efficiency).
Solution Approach 2:
The system prepares resources by keeping them in an activating or warm state rather than completely stopped, so they can be quickly deployed when computation requests arrive. This preliminary preparation maintains processing speed while avoiding the need to keep resources fully activated during low-demand periods, thus improving overall resource utilization efficiency.
3Device complexity
If queue management is simplified, then system complexity is reduced, but latency in resource allocation increases
Solution Approach 1:
The system segments the queue management into multiple specialized queues: an activation queue for managing resources being activated, a computation request queue for pending requests, and a releasing queue for resources completing computations. This segmentation allows each queue to be managed with specific optimization strategies, reducing overall latency while keeping individual queue management relatively simple.
Solution Approach 2:
The system introduces intermediary queue structures that mediate between computation requests and resource activation. The activation queue acts as an intermediary buffer, allowing resources to be prepared in advance without blocking the main computation request processing. This intermediary mechanism reduces latency in resource allocation while maintaining manageable system complexity through clear queue separation.
Data Source
AI summary
A computation resource control apparatus includes an activation unit, a first queue managing unit, an allocating unit and a second queue managing unit. The activation unit activates a computation resource being in a stop state in accordance with a computation request. The first queue managing unit adds the computation resource which is being activated to a first queue. The allocating unit allocates the computation resource, which is output from the first queue, to the computation request to execute a computation process corresponding to the computation request. The second queue managing unit adds the computation resource which has completed the computation process to a second queue and places the computation resource, which is output from the second queue, in the stop state.