FUSE Starts Working On Scalability Improvements With Linux 4.2
The performance of FUSE has been criticized in the past with being slower than native, in-kernel file-systems. Linus Torvalds also remarked that FUSE is for toys and misguided people. Yet there are many FUSE file-systems out there from SSHFS to WebDrive to Boxfs to a ZFS implementation and dozens of other file-systems. FUSE has also been ported to Linux to many other operating systems / kernels from OS X to MINIX3 to FreeBSD.
As good news for those desiring better performance out of FUSE, the developers have been working on scalability improvements and with Linux 4.2 comes the start of the changes. FUSE maintainer Miklos Szeredi explained, "An input queue and a processing queue is split out from the monolithic fuse connection, each of those having their own spinlock. The end of the patchset adds the ability to clone a fuse connection. This means, that instead of having to read/write requests/answers on a single fuse device fd, the fuse daemon can have multiple distinct file descriptors open. Each of those can be used to receive requests and send answers, currently the only constraint is that a request must be answered on the same fd as it was read from. This can be extended further to allow binding a device clone to a specific CPU or NUMA node."
The FUSE changes done Miklos can be found via this pull request arriving just in time for nearing the end of the Linux 4.2 merge window.
2 Comments
