Workload Allocation Manager for Persistent Memory Clusters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The deployment of data-intensive workloads to persistent memory capable environments poses significant complexity due to the need for intelligent workload placement and migration based on workload requirements and node capabilities, as existing systems lack efficient methods for ranking workloads and allocating resources effectively.
Innovation Solution
A system and method for allocating and migrating workloads across an IT environment based on persistent memory availability, utilizing a workload allocation and migration manager that ranks workloads and nodes based on priority, system calls, storage latency, and memory health, ensuring optimal placement and migration across a node cluster.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If workloads are deployed to persistent memory capable environments, then data-intensive application performance is improved, but deployment complexity increases significantly
Solution Approach 1:
The patent introduces a workload allocation and migration manager as an intermediary component that automatically handles the complex task of placing workloads on appropriate nodes. This manager ranks workloads based on persistent memory requirements and ranks nodes based on persistent memory availability and health, then performs the allocation and migration decisions automatically, thereby resolving the deployment complexity while maintaining performance benefits
Solution Approach 2:
The system dynamically changes allocation parameters by ranking workloads according to their persistent memory needs and ranking nodes according to their persistent memory availability and health status. These parameter changes enable adaptive workload placement that optimizes performance while simplifying deployment through automated decision-making based on multiple ranked factors
2Productivity
If intelligent workload placement is implemented based on multiple ranking criteria, then resource utilization is optimized, but system complexity increases
Solution Approach 1:
The patent segments the workload allocation problem into distinct ranking processes: workload ranking based on persistent memory requirements, node ranking based on persistent memory availability, and node ranking based on persistent memory health. This segmentation allows each aspect to be evaluated independently and combined to form comprehensive allocation decisions, optimizing resource utilization while managing system complexity through structured analysis
Solution Approach 2:
The workload allocation and migration manager serves multiple functions simultaneously: it ranks workloads, ranks nodes by availability, ranks nodes by health, performs allocation decisions, and handles migration. This multi-functional approach consolidates multiple complex tasks into a single unified system that optimizes resource utilization across all dimensions
3Reliability
If workloads are migrated based on persistent memory health ranking, then reliability is improved, but migration overhead increases
Solution Approach 1:
The system performs preliminary ranking of nodes based on persistent memory health status before workload allocation or migration decisions are made. By pre-establishing the health rankings of nodes, the system can quickly identify suitable target nodes for migration without performing complex health assessments during the migration process itself, thereby improving reliability while reducing migration overhead
Solution Approach 2:
The patent implements continuous monitoring and ranking of persistent memory health status across nodes, creating a feedback mechanism that informs workload allocation and migration decisions. This feedback loop ensures workloads are placed on healthy nodes and migrated from deteriorating nodes, improving reliability while the pre-computed rankings minimize the time required to act on health status changes
Data Source
AI summary
A method and system for allocating and migrating workloads across an information technology (IT) environment based on persistent memory availability. Specifically, the method and system disclosed herein entail the intelligent placement of workloads on appropriate nodes of a node cluster based on workload requirements and node capabilities and/or resources availability. Further, workloads may be ranked based on a workload priority assigned to any particular workload, if available, or based on logged system calls issued by virtual machines hosting any particular workload. Subsequently, higher ranked workloads may be granted priority access to nodes that host healthier persistent memory, if any, or host higher performance traditional storage.