next up previous
Next: Extent maps Up: State of the Art: Previous: Large inode support and


Extents, delayed allocation and extent allocation

This section and the next (Section 4) will discuss features that are currently under development, and (as of this writing) have not been merged into the mainline kernel. In most cases patches exist, but they are still being polished, and discussion within the ext2/3 development community is still in progress.

Currently, the ext2/ext3 filesystem, like other traditional UNIX filesystems, uses a direct, indirect, double indirect, and triple indirect blocks to map file offsets to on-disk blocks. This scheme, sometimes simply called an indirect block mapping scheme, is not efficient for large files, especially large file deletion. In order to address this problem, many modern filesystems (including XFS and JFS on Linux) use some form of extent maps instead of the traditional indirect block mapping scheme.

Since most filesystems try to allocate blocks in a contiguous fashion, extent maps are a more efficient way to represent the mapping between logical and physical blocks for large files. An extent is a single descriptor for a range of contiguous blocks, instead of using, say, hundreds of entries to describe each block individually.

Over the years, there have been many discussions about moving ext3 from the traditional indirect block mapping scheme to an extent map based scheme. Unfortunately, due to the complications involved with making an incompatible format change, progress on an actual implemention of these ideas had been slow.

Alex Tomas, with help from Andreas Dilger, designed and implemented extents for ext3. He posted the initial version of his extents patch on August, 2003. The initial results on file creation and file deletion tests inspired a round of discussion in the Linux community to consider adding extents to ext3. However, given the concerns that the format changes were ones that all of the ext3 developers will have to support on a long-term basis, and the fact that it was very late in the 2.5 development cycle, it was not integrated into the mainline kernel sources at that time.

Later, in April of 2004, Alex Tomas posted an updated extents patch, as well as additional patches that implemented delayed allocation and multiple block allocation to the ext2-devel mailing list. These patches were reposted in February 2005, and this re-ignited interest in adding extents to ext3, especially when it was shown that the combination of these three features resulted in significant throughput improvements on some sequential write tests.

In the next three sections, we will discuss how these three features are designed, followed by a discussion of the performance evaluation of the combination of the three patches.



Subsections
next up previous
Next: Extent maps Up: State of the Art: Previous: Large inode support and
Mingming Cao 2005-07-26