Read Throttling for Storage Load Balancing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In storage systems with disk drives, uneven load distribution leads to increased response times and potential timeouts due to slower disk drives, even with uniform distribution algorithms, as small biases can cause request accumulation on certain drives, resulting in delays.
Innovation Solution
A method that introduces a delay in responding to read requests if the requested data is not cached and if a predetermined number of disk drives are overloaded, ensuring the response time difference is maintained within a desired limit by tracking read request queues and thresholds, thereby balancing the load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If uniform distribution of requests across disk drives is implemented, then load balance is improved, but response time increases due to slower disk drives
Solution Approach 1:
The patent applies preliminary action by proactively introducing a delay to read requests before they are processed by overloaded disk drives. The control entity monitors queue depths and predetermined response times, and when thresholds are exceeded, it intentionally delays request processing. This prevents the accumulation of requests on slow disk drives before the problem manifests, thereby maintaining load balance across the storage system while avoiding the response time penalties that would otherwise occur.
2Productivity
If read requests are accumulated on slower disk drives, then data retrieval capacity is improved, but response time increases causing timeouts
Solution Approach 1:
The patent implements feedback by continuously monitoring the response times and queue depths of disk drives. The control entity compares actual response times against predetermined thresholds and uses this feedback to dynamically adjust request routing. When a disk drive shows signs of overload or slow performance, the system receives feedback about its degraded state and responds by delaying or redistributing requests, thereby preventing timeout conditions while maintaining optimal data retrieval capacity.
3Productivity
If delay is introduced to read requests from overloaded disk drives, then load on slow drives is reduced, but system response time increases
Solution Approach 1:
The patent applies local quality by introducing delays selectively rather than uniformly across all disk drives. The control entity identifies specific overloaded or slow-performing drives and applies delay mechanisms only to requests targeted at those particular drives. Requests to healthy, fast-performing drives are processed normally without delay. This localized application of quality control maintains overall system response time while effectively redistributing load away from problematic drives.
Data Source
AI summary
A method for managing read requests, the method may include receiving from a requesting entity a read request for reading an information unit stored in a storage system; determining by a control entity of the storage system whether the information unit is cached in a cache memory of the storage system and whether at least a predetermined number of disk drives of the storage system are currently overloaded; introducing a delay to a response to the read request thereby increasing a time difference between a time of the receiving of the read request and a time of a provision of the information unit to the requesting entity, if it is determined that the information unit is not cached in the cache memory and that the at least predetermined number of disk drives of the storage system are currently overloaded; and providing the information unit to the requesting entity.