Host-Storage Throttling for Low-Latency Reclamation Control

Solution Overview

Problem

Existing systems face challenges in efficiently managing storage resources when multiple hosts connect to a single data storage system, leading to adverse performance and latency impacts due to background operations like SCSI UNMAP commands, which consume significant resources and queue higher priority I/O operations, especially in scenarios with varying workloads across hosts.

Innovation Solution

A data storage system provides collective feedback to hosts, adjusting the rate of background operations like SCSI UNMAP commands based on monitored conditions such as latency and storage capacity, allowing resources to be allocated efficiently for higher priority tasks.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If hosts send storage reclamation commands to the data storage system, then storage resources are reclaimed and made available for reuse, but request latency increases and higher priority I/O operations are adversely affected

Engineering Contradiction:

Improvestorage capacityVSAvoidrequest latency

Core Design Contradiction:

Quantity of substanceVSLoss of time

Solution Approach 1:

The data storage system monitors its own performance metrics (such as request latency, queue depth, processor utilization) and sends feedback signals to hosts indicating whether they should throttle or continue sending storage reclamation commands. This feedback mechanism allows the system to dynamically adjust the reclamation rate based on current performance conditions, preventing latency degradation while still reclaiming storage resources when the system has excess capacity.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system implements dynamic throttling where the rate of storage reclamation commands is adjusted in real-time based on monitored performance conditions. Instead of a static reclamation rate, the system continuously adapts the reclamation intensity to match current system capacity and workload conditions, allowing optimal resource reclamation during low-utilization periods while preventing performance degradation during high-utilization periods.

Inventive Principle:

Principle #15Dynamics

2Productivity

If hosts send storage reclamation commands at high rate, then storage reclamation efficiency improves, but system performance degrades due to resource consumption

Engineering Contradiction:

Improvestorage reclamation efficiencyVSAvoidsystem performance

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The data storage system monitors its own performance metrics (such as request latency, queue depth, processor utilization) and sends feedback signals to hosts indicating whether they should throttle or continue sending storage reclamation commands. This feedback mechanism allows the system to dynamically adjust the reclamation rate based on current performance conditions, preventing latency degradation while still reclaiming storage resources when the system has excess capacity.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system changes operational parameters (throttling rate, command frequency) based on monitored system conditions. When performance metrics indicate healthy system state, the throttling rate is reduced to allow aggressive reclamation. When metrics degrade, the throttling rate is increased to protect performance. This dynamic parameter adjustment optimizes both reclamation efficiency and system performance.

Inventive Principle:

Principle #35Parameter changes

3Adaptability or versatility

If multiple hosts connect to a single data storage system, then resource sharing and data access improve, but background operations from multiple hosts accumulate and cause performance degradation

Engineering Contradiction:

Improveresource sharingVSAvoidrequest latency

Core Design Contradiction:

Adaptability or versatilityVSLoss of time

Solution Approach 1:

The data storage system aggregates throttling decisions from multiple hosts into a unified feedback mechanism. Instead of each host independently managing its reclamation commands, the system combines the impact of all hosts' reclamation activities and provides coordinated feedback to throttle the aggregate rate. This merging approach prevents cumulative performance degradation from multiple hosts while maintaining resource sharing benefits.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The data storage system monitors its own performance metrics (such as request latency, queue depth, processor utilization) and sends feedback signals to hosts indicating whether they should throttle or continue sending storage reclamation commands. This feedback mechanism allows the system to dynamically adjust the reclamation rate based on current performance conditions, preventing latency degradation while still reclaiming storage resources when the system has excess capacity.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS12517656B2Cooperative adaptive throttling between hosts and data storage systems

Publication Date: 2026.01.06 DELL PROD LP

AI summary

In at least one embodiment, processing can include: sending, from a host to a data storage system, a first storage reclamation command that identifies a first storage region of physical storage available for reclamation and reuse; in response to the data storage system receiving the first storage reclamation command, sending from the data storage system to the host first feedback information identifying at least a first detected condition on the data storage system; and in response to receiving the first feedback information at the host, the host varying a current rate or frequency of subsequent storage reclamation commands sent to the data storage system based, at least in part, on the first feedback information regarding the first detected condition on the data storage system.