The article says that log-structured apps and filesystems conflict with underlying SSDs, because those are doing similar things and the two don't play well together. But then it says that "products that already incorporate log structured I/O will have a definite advantage" in the future. How can they have an advantage if they're conflicting with the storage layer? Is this a contradiction, or am I missing something?
The paper [0] has some suggestions: an API (already existing and somewhat supported, called TRIM) to tell the disk that certain sectors are no longer in use, and atomic writes.
I guess I buy the claim that you're better off starting with a log-structured file system that will use new APIs to allocate and free chunks on the SSD, than to use a conventional file system that the SSD tries to optimize
