I just pushed new code to the bcache-3.2-dev branch. I expect this version to be more or less stable and production ready, but it does need new testing. There's two changes that should improve performance: * Reworked the code for inserting cache misses into the cache. Previously, in order to work around a particularly obscure race we had to fail a lot of inserts from cache misses that should've been fine (if there was any key at a particular offset, even a stale key or a key that just invalidated that region of the cache, inserts from cache misses would fail until the btree node was rewritten). This'll improve the cache hit ratio, how much depending on how badly the given workload was affected. * Reworked the journalling code to handle cache flushes instead of treating all writes as sync. This improves write latency and reduces the amount of journal io. Fixed some bugs too, including a memory leak. Please let me know if you find any more, and if anyone's running benchmarks I'd love to know how it's doing. -- To unsubscribe from this list: send the line "unsubscribe linux-bcache" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
