In-Place Multidimensional Database Writes to Reduce Fragmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multidimensional database systems face fragmentation issues due to data blocks being written to new locations, leading to contention for free space, index updates, and frequent defragmentation, which causes downtime and maintenance overhead.
Innovation Solution
Implement in-place data writing, where data blocks are updated in their current location if they fit within the existing space, with a configurable minimum size threshold and fudge factor for disk allocation, reducing fragmentation and index updates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If data blocks are written to new locations on the data storage medium, then data updates can be performed, but fragmentation of the data storage medium increases and index updates are required
Solution Approach 1:
Instead of writing updated data blocks to new locations (the conventional approach), the patent inverts the approach by writing updated data blocks back to their original locations on the data storage medium. This eliminates the need for index updates and prevents fragmentation, as the data block pointers in the index remain valid.
Solution Approach 2:
The patent extracts and eliminates the index update step from the data update process. By writing data blocks in-place rather than to new locations, the system removes the need to update index structures that would otherwise point to new data block locations, simplifying the overall update operation.
2Productivity
If data blocks are written to new locations, then data updates are possible, but contention for free space occurs
Solution Approach 1:
The patent inverts the conventional data update approach by writing to the original data block location rather than seeking new free space. This eliminates contention for free space among multiple threads, as each thread can independently update its assigned data blocks without competing for available storage locations.
3Productivity
If data blocks are written to new locations, then data updates can be performed, but frequent defragmentation is required causing downtime
Solution Approach 1:
By inverting the write approach to use in-place updates, the patent prevents fragmentation from occurring in the first place. This eliminates the need for periodic defragmentation operations and the associated system downtime, as the data storage medium maintains optimal organization continuously.
4Reliability
If index updates are performed for each data block write, then data block locations are tracked, but system performance decreases
Solution Approach 1:
The patent removes the index update operation from the data update process by writing data blocks to their original locations. Since the physical location of each data block remains unchanged, the index structures remain valid without requiring updates, thereby eliminating a significant performance bottleneck.
Data Source
AI summary
In accordance with an embodiment, described herein is a system and method for use with a multidimensional database computing environment. Particularly, a system and method are described for performing in-place data writes to a data storage medium associated with the multidimensional database computing environment. For example, when multiple data updates from calculation scripts are associated with a same set of blocks, the in-place data enable the system to reduce fragmentation of the data storage medium, and also to reduce contention due to index updates.