Storage Pool Management via Queue Depth Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data storage systems with virtual storage pools (VSPs) face challenges in maintaining consistent latency and minimizing head-of-line blocking due to differences in command execution latencies among data storage devices (DSDs) with varying firmware and hardware architectures.
Innovation Solution
A data storage system that includes a storage pool with multiple DSDs, each with single or multiple command queues, and a storage management device that receives queue information from each DSD to select a target DSD that minimizes head-of-line blocking and maximizes parallel queue usage for incoming commands.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If DSDs with different firmware and hardware architectures are pooled together to form a VSP, then the overall storage capacity and utilization are improved, but the command execution latency becomes inconsistent due to head-of-line blocking
Solution Approach 1:
The storage management device dynamically selects target DSDs based on real-time queue information, adapting to changing system conditions. The system continuously monitors queue depths and adjusts command routing decisions to minimize latency while maximizing storage utilization across heterogeneous DSDs.
Solution Approach 2:
The system changes the routing parameter (target DSD selection) based on queue information. By selecting DSDs with lower queue depths for new commands, the system optimizes execution latency while maintaining high storage pool utilization across diverse hardware architectures.
2Reliability
If commands are stored in queues before processing, then the DSD can service commands in order, but head-of-line blocking increases latency for later arriving commands
Solution Approach 1:
The storage management device performs preliminary selection of the target DSD based on queue information before issuing new commands. By anticipating which DSDs will have lower latency based on current queue depths, the system reduces future head-of-line blocking effects while maintaining proper command ordering.
Solution Approach 2:
The system uses queue information as feedback to continuously optimize command routing. The storage management device monitors queue depths and uses this feedback to select target DSDs that minimize latency, balancing command ordering requirements with performance optimization.
3Productivity
If a VSP receives multiple read and write commands from a host, then the storage capacity is effectively utilized, but the varying latencies among different DSDs cause inconsistent performance
Solution Approach 1:
The storage management device dynamically adapts command routing based on real-time queue information from each DSD. This dynamic selection process ensures that commands are directed to the most appropriate target DSD, maintaining performance consistency across the heterogeneous storage pool while maximizing capacity utilization.
Solution Approach 2:
The system changes routing decisions based on queue depth parameters. By selecting target DSDs with favorable queue conditions, the system maintains consistent performance characteristics across the VSP while effectively utilizing the total storage capacity of all pooled devices.
Data Source
AI summary
Aspects of the disclosure relate to improved techniques for managing a data storage device (DSD) pool, and in particular the selecting of DSDs based on commands distribution and latencies among the command queues in the DSDs. In some aspects, a DSD may have multiple queues that enable the DSD to perform certain commands (e.g., write commands and/or read commands) in parallel. Thus, such DSD has various head-of-line blocking latencies based on which queue the command is directed to. A storage management device can reduce the command latency of the DSDs, for example, based on queue information (e.g., head-of-line blocking, parallel queue utilization, etc.) learned from the DSDs.