Storage Workload Priority Management for Latency Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing storage systems lack the ability to prioritize workloads effectively, leading to performance deterioration under heavy loads and undesirable impacts on user operations during system-initiated tasks like snapshot or snapmirror operations, as they do not allow differentiation in resource allocation between user-initiated and system-initiated requests.
Innovation Solution
A system and method for managing workloads by assigning priority levels to different classes of requests, allowing storage administrators to prioritize user-initiated workloads relative to system-initiated workloads, and controlling CPU usage, disk reads, and NVRAM consumption based on workload classes associated with target volumes, network protocols, or communication ports.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If storage server handles heavy load in non-linear fashion, then system throughput increases, but performance deteriorates rapidly and unexpected interactions occur between storage components
Solution Approach 1:
The patent segments workloads into different classes (user-initiated vs. system-initiated, or different priority levels) and applies different handling mechanisms to each class. This allows the system to manage heavy loads by processing high-priority user requests differently from system maintenance operations, preventing performance deterioration while maintaining throughput.
Solution Approach 2:
The patent implements dynamic workload management where the system can adjust processing priorities and resource allocation in real-time based on current load conditions. The system dynamically switches between different service modes (e.g., prioritizing user requests during heavy load, or allowing system operations during lower priority periods) to maintain both throughput and performance stability.
2Reliability
If system performs snapshot or snapmirror operations, then data consistency and backup are improved, but user operations experience increased latency and decreased I/O performance
Solution Approach 1:
The patent segments requests into user-initiated and system-initiated classes, assigning different priorities. During snapshot or snapmirror operations, system-initiated requests are prioritized while user-initiated requests are throttled or delayed, ensuring data consistency without completely blocking user operations. This segmentation allows both reliability and acceptable latency to coexist.
Solution Approach 2:
The patent applies preliminary anti-action by preemptively managing workload queues before snapshot operations begin. The system pre-establishes priority rules and resource allocation strategies that will be activated during consistency point operations, preventing severe latency spikes by having mitigation mechanisms ready in advance.
3Productivity
If existing systems optimize performance automatically to achieve maximum throughput, then productivity increases, but reduced latency cannot be prioritized in multi-function environments
Solution Approach 1:
The patent implements dynamic priority adjustment where the system can switch between throughput-optimized mode and latency-optimized mode based on current workload characteristics and user requirements. The system dynamically reconfigures resource allocation, queue processing orders, and I/O scheduling strategies to prioritize either throughput or latency depending on the immediate needs of different workload classes.
Solution Approach 2:
The patent changes key system parameters such as I/O scheduling thresholds, queue processing priorities, and resource allocation ratios based on detected workload conditions. By dynamically adjusting these parameters, the system can optimize for either maximum throughput or reduced latency depending on the current operational context and user priorities.
4Device complexity
If storage system does not allow prioritization of workloads, then system simplicity is maintained, but resource allocation cannot be differentiated between mission critical and tolerant workloads
Solution Approach 1:
The patent introduces segmentation of workloads into distinct classes with different priority levels and resource allocation rules. This segmentation enables the system to differentiate between mission-critical workloads (e.g., user file access) and tolerant workloads (e.g., system maintenance operations), providing adaptability while maintaining relatively simple implementation through clear classification rules.
Solution Approach 2:
The patent introduces a workload classification and priority management layer that acts as an intermediary between the storage system and different workload types. This intermediary layer handles the complexity of workload differentiation centrally, allowing the core storage system to remain simple while still providing sophisticated resource allocation through the mediation of priority rules and queue management.
Data Source
AI summary
A system and method are provided to manage different types of workload. The method comprises receiving a request, determining a class of the request, determining a priority level associated with the class of the request, and servicing the request utilizing the priority level of the class of the request. The class of the request may be associated with a target volume, as well as with a network protocol, or with a network port.