Adaptive Resynchronization Speed for Storage Nodes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In storage platforms with mirrored volumes, resynchronization processes can burden both owner and backup storage nodes, leading to performance degradation for clients, especially when nodes become unavailable or are newly configured, as they need to transfer accumulated data changes efficiently without disrupting ongoing service requests.
Innovation Solution
Storage processors track activity metrics to dynamically adjust the speed and chunk size of resynchronization data transfer based on current workload, allowing for efficient data synchronization without impacting client service performance by processing data in manageable chunks during idle times and slowing down during busy periods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If resynchronization data is transferred at high speed to quickly synchronize backup volume, then synchronization time is reduced, but client service performance deteriorates due to burden on owner and backup storage nodes
Solution Approach 1:
The patent implements dynamic adjustment of resynchronization data transfer speed by monitoring workload metrics from storage processors. When workload is high, the transfer speed is reduced to prevent performance degradation; when workload is low, transfer speed is increased to complete resynchronization quickly. This dynamic control resolves the contradiction between fast synchronization and maintaining client service performance.
Solution Approach 2:
The system changes the parameter of data transfer rate based on monitored workload conditions. By adjusting this parameter in response to system state, the patent achieves both fast resynchronization when possible and maintains client service performance when workload is high, resolving the technical contradiction.
2Productivity
If resynchronization data is transferred in large chunks to reduce number of transfers, then resynchronization efficiency is improved, but storage node burden increases causing service degradation
Solution Approach 1:
The patent dynamically adjusts chunk size based on monitored workload metrics. When storage nodes have low workload, larger chunks are transferred to improve efficiency; when workload is high, smaller chunks are used to reduce burden. This dynamic adaptation resolves the contradiction between resynchronization efficiency and storage node burden.
Solution Approach 2:
The system changes the parameter of chunk size in response to workload conditions, allowing efficient large transfers when possible while reducing to smaller chunks when node burden would cause service degradation.
3Reliability
If resynchronization operation is performed immediately when backup node returns or is newly configured, then data consistency is restored quickly, but client service performance deteriorates due to processing burden
Solution Approach 1:
The patent implements dynamic control of resynchronization initiation and speed based on real-time workload monitoring. When the backup node returns or is newly configured, the system monitors storage processor workload and adjusts the resynchronization process accordingly - slowing down or pausing when workload is high to maintain client service performance, and accelerating when workload is low to restore data consistency quickly.
Solution Approach 2:
The system uses feedback from workload metrics to control the resynchronization process. By continuously monitoring storage processor activity and adjusting resynchronization behavior based on this feedback, the patent resolves the contradiction between quickly restoring data consistency and maintaining client service performance.
Data Source
AI summary
A resynchronization operation (300) adapts according to activity histories (148) within a storage platform (100). An owner node (A) and a backup storage node (B) may track activity such as the IO operations begun and use recent activity to determine respective amounts of data the owner (A) and backup (B) can expect to process without unacceptably degrading storage services. The owner (A) transfer resynchronization data in chunks, each having a size limited by the current amounts the owner (A) and backup (B) determined from current activity. When activity is low or idle, large chunks may be sent to quickly complete resynchronization, while a busy system uses smaller chunks such that the system performance is not adversely affected.