Storage Network Bandwidth Management via Latency-Based I/O Throttling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computer storage network management technologies face challenges in managing administrative I/O requests, which can cause significant delays and impair application performance by interfering with application I/O transactions, leading to suboptimal bandwidth utilization.
Innovation Solution
The method involves sampling application I/O requests to determine a maximum latency value, comparing it with a current latency value, and throttling administrative I/O requests using a sliding window protocol when the current value exceeds the maximum, while also increasing administrative I/O requests when the latency is below a specified threshold, thereby optimizing bandwidth management across multiple units of storage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If administrative I/O requests are increased to improve storage network bandwidth utilization, then bandwidth efficiency is improved, but application I/O latency increases and application performance deteriorates
Solution Approach 1:
The system continuously monitors application I/O latency and uses this feedback to dynamically adjust the rate of administrative I/O requests. When latency exceeds a threshold, the system reduces administrative I/O rate; when latency is acceptable, the system increases administrative I/O rate to improve bandwidth utilization. This closed-loop feedback mechanism resolves the contradiction by making administrative I/O adaptive to application performance conditions.
Solution Approach 2:
The system transitions from static administrative I/O scheduling to dynamic adjustment based on real-time latency conditions. The administrative I/O rate is no longer fixed but varies continuously according to monitored application performance, allowing the system to optimize bandwidth utilization without permanently degrading application latency.
2Loss of time
If administrative I/O requests are throttled to reduce application I/O latency, then application performance is improved, but storage network bandwidth utilization decreases
Solution Approach 1:
The system implements periodic sampling of application I/O requests to determine maximum latency values at regular intervals. This periodic measurement approach allows the system to capture latency characteristics over time and adjust administrative I/O rates in corresponding periods, achieving a rhythm that balances application performance with bandwidth utilization rather than applying constant throttling.
Solution Approach 2:
The system dynamically changes the parameter of administrative I/O request rate based on measured latency parameters. By adjusting the rate parameter in response to latency measurements, the system optimizes the trade-off between reducing application latency and maintaining bandwidth utilization, rather than using fixed throttling parameters.
3Adaptability or versatility
If volume management operations are performed to provide storage virtualization and flexibility, then storage management capability is improved, but administrative I/O increases causing delays
Solution Approach 1:
The system performs preliminary sampling of application I/O requests to determine maximum latency values before executing administrative I/O operations. This preliminary action allows the system to plan and schedule volume management operations during periods when latency thresholds are acceptable, reducing the impact of administrative I/O on application performance while maintaining storage management flexibility.
Data Source
AI summary
Techniques for storage network bandwidth management are disclosed. In one particular exemplary embodiment, the techniques may be realized as a method for storage network bandwidth management comprising sampling, using at least one computer processor, application Input/Output (I/O) requests associated with the unit of storage during a specified period of time, determining a maximum latency value based on the sampling of the application Input/Output (I/O) requests, comparing the maximum latency value with a current latency value, and throttling administrative I/O requests in the event that the current latency value exceeds the maximum latency value.