Storage Replication Cache Flushing to Cut Journal Disk Wear
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The long distance and limited fiber transmission rate between source and destination storage devices in data replication processes lead to prolonged data replication times, degrading user experience and increasing wear on journal log disks.
Innovation Solution
A method and device that write data to a cache and flush it to a storage disk while recording the last logical block address (LBA) in the destination storage device, allowing for skipping the journal log disk and sending the LBA to the source to clear journals, thereby reducing replication time and disk wear.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data is written directly to the storage disk through the journal log disk during replication, then data integrity is maintained, but replication time is prolonged and disk wear increases
Solution Approach 1:
The patent segments the data replication process into two distinct phases: a fast replication phase where data is written to the cache without flushing to the journal log disk, and a subsequent synchronization phase where only the last flushed LBA is sent to clear journals. This segmentation allows the bulk of data replication to occur quickly while maintaining integrity through the journal clearing mechanism.
Solution Approach 2:
The patent performs preliminary action by writing data to the cache first and then flushing it to the storage disk before completing the replication process. The destination storage device records the last flushed LBA and sends this information back to the source storage device, which then clears the journals beforehand. This preliminary flushing ensures data is safely stored before the replication is considered complete.
2Reliability
If data is written to the journal log disk during replication, then data integrity is ensured, but disk wear increases
Solution Approach 1:
The patent extracts the data writing operation from the journal log disk by introducing a cache layer. Data is first written to the cache memory, then flushed to the storage disk, and finally the journal is cleared using the recorded LBA information. This extraction removes the repetitive write operations from the journal log disk, significantly reducing wear while maintaining data integrity through the journal clearing mechanism.
Solution Approach 2:
The patent discards the need for continuous journal writing during replication by using the cache and LBA recording mechanism. Instead of continuously writing to the journal log disk, the system discards the journal entries up to the recorded LBA after replication, effectively removing the wear-causing operations while recovering data integrity through the clearing mechanism.
3Speed
If the fiber channel transmission rate is used as the theoretical maximum, then transmission speed is optimized, but actual transmission rate is reduced due to overheads
Solution Approach 1:
The patent introduces a cache as an intermediary between the fiber channel transmission and the storage disk. The cache receives data from the fiber channel at high speed without the overhead penalties of direct disk writes, then flushes data to the disk in bulk operations. This intermediary mechanism allows the system to achieve higher effective transmission rates by decoupling the high-speed network interface from the slower disk interface.
Data Source
AI summary
Techniques involve: receiving, by a destination device, first input/output (I/O) requests from a source device, the first I/O requests having first logical block addresses (LBAs); writing data in the first I/O requests to a cache of the destination device; flushing the data from the cache to a storage disk of the destination device; recording a particular LBA in the destination device, the particular LBA corresponding to the last LBA to which the destination device is flushed; and sending the particular LBA from the destination device to the source device to enable clearing of journals before the particular LBA in the source device. Such techniques reduce the time spent on replicating data from the source device to the destination device to improve the user experience and reduce the wear and tear of the journal log disk in the destination device to prolong its service life.