Dynamic Workload Claiming for Resilient Cloud and HPC Execution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing HPC environments face inefficiencies due to static workload assignment that fails to account for node performance differences and resource availability, leading to delayed execution and node failure challenges, particularly in high-performance computing environments.
Innovation Solution
A workload execution framework that enables dynamic assignment of execution units to nodes, allowing nodes to claim and re-claim units as needed, provides load balancing, and ensures resilience through node failure recovery and auto-scaling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If static workload assignment is used, then workload distribution is simple to implement, but execution efficiency deteriorates due to inability to account for node performance differences
Solution Approach 1:
The patent implements dynamic workload assignment where nodes can claim and re-claim execution units based on their actual performance and availability. The system transitions from static pre-assigned workloads to dynamic claiming mechanisms where nodes actively select available units, enabling adaptive response to node failures and performance variations without complex reconfiguration.
2Device complexity
If static workload assignment is used, then assignment is straightforward, but completion time increases due to nodes waiting for slower nodes
Solution Approach 1:
The system incorporates feedback mechanisms where nodes report their completion status of execution units, and the workload manager uses this information to redistribute units. When a node completes its assigned units, it can claim additional units from the pool, creating a feedback-driven dynamic allocation that optimizes completion time based on actual node performance.
3Reliability
If node failure recovery is implemented, then system reliability improves, but execution time increases due to checkpointing and restart overhead
Solution Approach 1:
The patent implements dynamic workload reassignment upon node failure rather than static checkpoint-based recovery. When a node fails, the system dynamically redistributes its execution units to other available nodes without requiring time-consuming checkpoints. The failed node's units are immediately claimed by other nodes from the shared pool, enabling faster recovery while maintaining reliability.
4Productivity
If dynamic workload claiming is implemented, then load balancing improves, but system complexity increases
Solution Approach 1:
The system implements self-service mechanisms where nodes autonomously claim execution units from the shared pool without requiring centralized micromanagement. Each node actively selects and claims available units based on its needs and capabilities, enabling distributed load balancing that reduces management complexity while improving productivity.
Data Source
AI summary
In certain embodiments, a method includes receiving, at a framework daemon, an indication to begin execution of a workload divided into execution units; accessing, by the framework daemon and in response to the indication, an execution context that includes the execution units; associating, by the framework daemon, an identifier of a node with an execution unit of the execution units to claim the execution unit for execution by the node; executing, by the node, the execution unit; accessing, by the framework daemon and after the node completes execution of the execution unit, the execution context to determine whether the execution context includes an unclaimed execution unit; associating, by the framework daemon and when the execution context includes an unclaimed execution unit, the identifier of the node with the unclaimed execution unit to claim the unclaimed execution unit for execution by the node; and executing, by the node, the execution unit.