Disruption Counters for Computing Cluster Predictability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In computing clusters, simultaneous job disruptions due to competing resource demands and unplanned events lead to unpredictability and potential service level agreement breaches, as existing job schedulers struggle to manage concurrent disruptions effectively.
Innovation Solution
The implementation of disruption counters that track concurrent disruptions and adjust precedence values to resist or permit disruptions based on predefined limits, protecting system targets by incrementing or decrementing counter values accordingly, and propagating increased precedence to undisrupted targets when limits are reached.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If job schedulers allow concurrent disruptions for resource management, then resource utilization improves, but system predictability deteriorates
Solution Approach 1:
The patent implements dynamic adjustment of disruption limits based on current system state. The disruption counter tracks the number of active disruptions and dynamically adjusts whether to permit or resist new disruption requests, allowing the system to adapt between prioritizing resource utilization and maintaining predictability based on real-time conditions
Solution Approach 2:
The system changes the parameter of disruption permission by comparing the disruption counter value against a threshold. When the counter indicates few active disruptions, the system permits new disruptions to improve resource utilization. When the counter approaches the limit, the system resists new disruptions to maintain predictability, effectively changing the system's behavior parameter based on current load
2Reliability
If disruption limits are enforced to maintain predictability, then system reliability improves, but resource utilization deteriorates
Solution Approach 1:
The patent applies partial action by enforcing disruption limits only when necessary. Rather than completely blocking all disruptions, the system permits disruptions up to a configured threshold and only resists additional disruptions when the counter reaches the limit. This partial enforcement maintains predictability while still allowing reasonable resource reallocation
3Reliability
If disruption counters track and resist disruptions, then service level agreement compliance improves, but system complexity increases
Solution Approach 1:
The disruption counter acts as an intermediary mechanism between disruption requests and the job scheduler. Rather than implementing complex validation logic throughout the system, the patent introduces a simple counter that mediates disruption decisions by comparing its value against a threshold, thereby ensuring SLA compliance through a single, manageable component
Data Source
AI summary
A system is provided that includes a memory and one or more processors in communication with the memory. The one or more processors are configured to identify a set of targets and select a first value corresponding to a number of targets from the set of targets that can be concurrently disrupted. A second value is determined that is related to a number of disruptions actually occurring. A disruption request is received for a target of the set of targets. Thereafter, the first value is compared to the second value. Based on the comparison of the first and second values, it is determined whether to resist a requested disruption. If it is determined that the disruption is to be resisted, the requested disruption is resisted. If it is determined that the disruption is not to be resisted, at least one of the first value and the second value are adjusted.