It's not often having anything to write about on the Oracle Cluster File-System v2 (OCFS2), but with Linux 6.10 it's seeing a rather significant performance optimization.

Andrew Morton on Sunday sent out his set of non-MM updates for the Linux 6.10 merge window. Among those many random patches are two sets of OCFS2 patch series: one providing better write I/O performance and the other providing random bug fixes as a result of the file-system testing (fstests).

Heming Zhao of SUSE spearheaded the work to improve the write I/O performance particularly when fragmentation is high. The optimization is to avoid unnecessary searches and especially when there are a lot of small files.

The impact of this write I/O performance optimization for OCFS2 is rather significant. As summed up by Heming Zhao:

"[before the patch] the loop times: 7*242*49*(32256/64) = 41835024 (~42 million times)

In the worst case, user space writes 1MB data will trigger 42M scanning times.

under this patch, the timing is '7*242*49 = 83006', reduce by three orders of magnitude."

A nice improvement for anyone using this shared disk cluster file-system. These OCFS2 improvements are part of [1]the non-MM updates heading into Linux 6.10.



[1] https://lore.kernel.org/lkml/20240519115735.acabb2269a1a91ee0ebdbca5@linux-foundation.org/