Issues Fragmentable Structure
When data are not available or produced in sequence, it is useful to support fragmentable structure which allows to store data in out-of-order manner.
Fragmentable Structure
Since the format itself does not support fragment-ability according to the block structure of the file system, it is appropriate to use a transformation block for these purposes. This technique works so that the data block it interpreted as a sequence of blocks of constant size and allows their classification in any order, thus allowing noncontinuous sequence of regular blocks representation.
Description of the Solution
This is a transformational type of data-block tree. Individual clusters has constant length and are either structural or data type. The first block is the root and structural. The proposed solutions is not likely final.
UBClusterIndex - Next Block
UBENatural - SubBlockPartSize
UBClusterIndex - First SubBlock
...
UBClusterIndex - Last SubBlock
UBNatural - DataPartSize
UBNatural - MainBlockPartSize
UBClusterIndex - First MainBlock
...
UBClusterIndex - Last MainBlock
While the indexes to SubBlock refer to the structural blocks, an index to MainBlock is referring to the sequence of data block.
SubBlockPartSize value equals to infinity means that the file is equivalent to data block, otherwise it is a node / leaf block.
Advantages of the Solution
Mentioned technology brings some advantages, especially it's more flexible way how to edit the file, to add to the other blocks on any depth level and change their size.
- The potential increase or reduce of the block does not lead to the need to rebuild the entire file
- More efficient organization reduces the number of necessary accesses to disk
Possible Optimalizations
It should be possible to improve low-efficiency of the solution.
For example, with the introduction of bitmaps using blocks, use of the relative values of the clusters indexes, or assembling a large number of small blocks into one.
It could be also possible to establish totals sums for increased robustness, or caching of certain values for greater efficiency.
Problems
The implementation of this solution is also facing some problems.
- How to align the entire block to disk access block - There should be an advantage to use SLRUB type.
- How to address the problem where the value of Next Block is longer than the cluster size
- How to determine data vs. leaf block
Page Source