Cloud Workload Deployment with Function Accelerator Cards
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional cloud systems face inefficiencies in resource utilization, where load is often concentrated on specific servers with function accelerator cards, limiting the ability to distribute workload effectively based on changing resource situations after deployment.
Innovation Solution
A method for task deployment in a cloud system that involves determining resource status, calculating performance estimation values for host servers and function accelerator cards, and selecting optimal deployment locations based on these values to prevent load concentration and enhance resource utilization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If a program using a specific function is fixedly deployed to a server capable of accelerating the specific function, then the function execution performance is improved, but the load is concentrated on a specific server while resources of other servers are in an idle state
Solution Approach 1:
The patent implements dynamic workload deployment by continuously monitoring resource status and performance estimation values, and adjusting task allocation in real-time. The system transitions from static fixed deployment to dynamic adaptive deployment, where tasks are automatically redistributed based on current system state, resolving the contradiction between performance optimization and resource utilization.
Solution Approach 2:
The system changes deployment parameters dynamically by calculating performance estimation values based on current resource status and using these values to determine optimal deployment targets. This parameter-based adaptive approach allows the system to maintain high function execution performance while distributing load across multiple servers based on real-time conditions.
2Power
If the load is concentrated on a specific server to ensure function execution performance, then the function acceleration capability is improved, but the ability to distribute workload effectively based on changing resource situations is reduced
Solution Approach 1:
The patent implements a feedback mechanism where the system continuously monitors resource status, calculates performance estimation values, and uses this feedback to adjust workload deployment decisions. This closed-loop control enables the system to maintain function acceleration capability while adapting workload distribution to changing resource situations, resolving the contradiction between performance and adaptability.
Solution Approach 2:
The system transforms the static workload deployment into a dynamic process that continuously adapts to changing resource conditions. By implementing real-time monitoring and recalibration of performance estimation values, the system maintains function acceleration while gaining the flexibility to redistribute load according to current system state.
3Speed
If programs are fixedly deployed to specific servers for function acceleration, then the execution speed for specific functions is improved, but the system cannot efficiently redistribute load when resource situations change
Solution Approach 1:
The patent implements dynamic workload deployment that maintains high execution speed for functions while enabling efficient load redistribution. The system continuously calculates performance estimation values based on current resource status and automatically redirects tasks to optimal servers, ensuring both fast execution and adaptive load distribution.
Solution Approach 2:
The system performs preliminary calculation of performance estimation values and resource status assessment before making deployment decisions. This preliminary action enables the system to maintain execution speed by pre-evaluating potential deployment targets while preserving the ability to redistribute load efficiently when resource conditions change.
Data Source
AI summary
The present disclosure relates to a method, apparatus, system, and computer program for task deployment in a cloud system including a function accelerator card, and more particularly to a method, apparatus, system, and computer program for task deployment in a cloud system including a function accelerator card, which can efficiently perform task deployment in the cloud system including the function accelerator card to increase the efficiency of the cloud system.In the present disclosure, disclosed is a task deployment method in a cloud system including one or more host servers and one or more function accelerator cards, which is performed by one or more processors in a task deployment apparatus, the task deployment method including: determining the status of each of available resources for the one or more host servers and the one or more function accelerator cards; calculating each of performance estimation values when the task to be deployed is executed in the one or more host servers or the one or more function accelerator cards under the condition of each of the available resources; and selecting the host servers or function accelerator cards in which the task is to be deployed in consideration of each of the performance estimation values.