Miao Xie sent over a patch to linux-btrfs asking for comments about this patch that provides an extent buffer cache for each i-node. "This patch introduce extent buffer cache for every i-node. By this way, we needn't search the item from the root of b+ tree, and can save the search time. Besides that we can also reduce the lock contention of the root."
When looking at the small file write performance via sysbench, he found that this patch he wrote could make the file-system 5~10% faster. He plans to run more performance tests soon to validate the work. He also has further cleans to clean-up the patch before asking it be pulled into the tree.
If this per i-node extent buffer cache patch moves quickly, it's possible to become a candidate for the Linux 3.4 kernel release that will happen in a few months. This will also be good news for Fedora 18, which should be the first tier-one distribution using Btrfs by default later in the year.
Benchmarks of Btrfs and EXT4 from the in-development Linux 3.3 kernel are going to be published soon. The most recent EXT4/Btrfs benchmarks on Phoronix as of right now is from the now-stable Linux 3.2 kernel.