Fixed Time Offset Co-allocation for Compute Resource Scheduling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In compute environments like clusters and grids, managing resources to maximize utilization and accommodate jobs with disparate but interrelated resource requirements is challenging, as existing systems fail to efficiently coordinate resource reservations across multiple jobs with varying resource types and time dependencies.
Innovation Solution
A method and system that analyze resource availability across nodes, considering fixed time relationships between jobs, shifting and combining available timeframes to determine optimal resource allocation, enabling co-allocation of resources for jobs with interrelated requirements, thereby maximizing resource usage and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If resources are reserved for multiple jobs with disparate resource requirements, then system utilization increases, but coordinating resource reservations across jobs with varying resource types and time dependencies becomes complex
Solution Approach 1:
The patent segments the resource coordination problem by introducing a hierarchical reservation structure. A parent reservation groups multiple child reservations, each corresponding to different resource types (compute, storage, network). This segmentation allows independent management of each resource type while maintaining coordinated allocation across the entire job portfolio, reducing the complexity of coordinating disparate resource requirements.
Solution Approach 2:
The patent applies preliminary action by establishing parent reservations in advance that define the overall resource allocation framework before child reservations are created. The parent reservation pre-defines time windows, resource quotas, and dependency relationships, allowing the system to efficiently allocate specific resources to individual jobs without real-time complex coordination when jobs are actually executed.
2Productivity
If resource allocation is optimized for jobs with fixed time relationships, then resource efficiency improves, but the scheduling system becomes more complex
Solution Approach 1:
The patent uses preliminary action by requiring users to specify fixed time relationships and offsets between jobs with interrelated resource requirements during reservation creation. The system pre-calculates and locks in these temporal relationships in the parent reservation structure, enabling efficient resource allocation that respects job dependencies without requiring complex real-time scheduling decisions.
Solution Approach 2:
The patent applies parameter changes by introducing time offset parameters that define fixed temporal relationships between jobs. These parameters are embedded in the reservation structure, allowing the scheduling system to automatically adjust resource allocation based on predefined time relationships rather than treating each job independently, thereby improving resource efficiency while maintaining manageable system complexity.
3Adaptability or versatility
If disparate resource types are allocated to interrelated jobs, then job processing capability increases, but resource management becomes more difficult
Solution Approach 1:
The patent segments resource management by creating distinct child reservations for different resource types (compute resources, storage resources, network resources) under a unified parent reservation. Each resource type can be managed independently with its own allocation rules and availability criteria, while the parent reservation ensures coordinated allocation across all resource types for interrelated jobs, making management easier despite the diversity of resources.
Solution Approach 2:
The parent reservation structure serves multiple functions simultaneously: it coordinates resource allocation across different resource types, enforces time relationships between jobs, manages dependencies between interrelated jobs, and provides a unified interface for resource management. This multi-functionality allows the system to handle disparate resource types without increasing operational complexity.
Data Source
AI summary
Disclosed are a system, method and computer-readable medium relating to managing resources within a compute environment having a group of nodes or computing devices. The method comprises, for each node in the compute environment: traversing a list jobs having a fixed time relationship, wherein for each job in the list, the following steps occur: obtaining a range list of available timeframes for each job, converting each availability timeframe to a start range, shifting the resulting start range in time by a job offset, for a first job, copying the resulting start range into a node range, and for all subsequent jobs, logically AND'ing the start range with the node range. Next, the method comprises logically OR'ing the node range with a global range, generating a list of acceptable resources on which to start and the timeframe at which to start and creating reservations according to the list of acceptable resources for the resources in the group of computing devices and associated job offsets.