On Thu, Dec 10, 2020 at 10:17:41AM +0200, Andrei POPESCU wrote:
On Mi, 09 dec 20, 15:23:44, Celejar wrote:I'm curious about this because I can't imagine that FUSE performance is as good as native, so why would automounters pay the performance penalty of FUSE when native mounting would seem easy enough to do?The ntfs-3g developers claim there is no significant penalty. Testing their claims would be difficult though, considering the kernel driver has limited functionality.
For exfat, at least, the in-kernel driver is noticeably faster. From a reasonably fast drive I can get 600-700MByte/s with the kernel driver, and 400-500MByte/s with fuse. With a slower drive I can get about 260MByte/s with kernel, 220MByte/s with fuse. Whether the difference is "significant" probably depends on how much data you're transferring and how long you're willing to wait.
In the past I've seen much larger differences for actual applications. I don't know if the fuse module has gotten faster or my quick copy test above doesn't adequately reflect real world performance. YMMV.