Cloud Workload Recovery via Cost-Ranked Context Reallocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current approaches for workload recovery in cloud environments, particularly in high-performance computing and AI applications, are inadequate in determining the optimal recovery strategy and location due to high failure frequencies and complex resource requirements.
Innovation Solution
A context controller evaluates various recovery options, including workload shrinkage, allocation shifts, deadline re-queuing, partial bursting, and secondary context reassignment, to identify the lowest-cost option for workload recovery, leveraging a resource controller node to manage compute node capacities across multiple contexts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If workload recovery is performed in the current cloud context, then recovery speed is improved, but resource availability and reliability deteriorate due to high failure frequencies
Solution Approach 1:
The patent extends workload recovery from a single cloud context to multiple cloud contexts (primary and secondary). When failure is detected in the current allocation, the system evaluates recovery options across different cloud environments and selects the optimal secondary context based on cost evaluation, thereby improving reliability while maintaining recovery speed.
2Reliability
If workload is reassigned to secondary cloud contexts, then reliability is improved, but complexity of recovery management increases
Solution Approach 1:
The patent introduces a context controller as an intermediary component that manages workload recovery across multiple cloud contexts. The context controller receives failure notifications, evaluates recovery options, interacts with the resource controller to identify secondary contexts, and executes the selected recovery plan, thereby simplifying the overall recovery management process while improving reliability.
Solution Approach 2:
The system implements feedback mechanisms where the context controller continuously monitors cloud context health, evaluates recovery options based on current system state, and adjusts recovery decisions dynamically. The resource controller provides feedback on available secondary contexts and their capacities, enabling informed recovery decisions.
3Loss of energy
If multiple recovery options are evaluated, then optimization of recovery cost is improved, but processing time increases
Solution Approach 1:
The resource controller pre-identifies and maintains information about available secondary cloud contexts and their capacities before recovery is needed. This preliminary preparation allows the context controller to quickly evaluate recovery options without performing time-consuming resource discovery during actual recovery events, thus reducing evaluation processing time while still enabling cost-optimized decisions.
Data Source
AI summary
A computer hardware system includes a context controller configured to manage recovery for a workload within a cloud environment. A plurality of compute nodes are initially allocated to the workload, as a current allocation, within a context of the cloud environment. A failure within the current allocation is identified. Responsive to the failure being identified, a plurality of costs respectively associated with each of a plurality of recovery options for the workload are evaluated. A lowest-cost option, from the plurality of recovery options being evaluated, is executed. The recovery options being evaluated includes both recovery options that include performing the workload within the current context and recovery options that include performing at least a portion of the workload in at least one secondary context of the cloud environment. A resource controller node associated with the context controller is communicated with to identify the at least one secondary context.