Intel Has Been Working On A New User-Space File-System For Persistent Memory

Intel developers have been working on a new user-space file-system designed for persistent memory. This user-space file-system is designed to be high-performance and does not make use of FUSE.

While FUSE for file-systems in user-space tends to be controversial like Linus Torvalds believing it's for toys and misguided people and can often be faced with performance issues compared to file-systems backed by kernel-mode drivers, Intel's new user-space file-system for persistent memory is designed for maximum performance and thus they want to get the kernel out of the way.

The "pmemfile" project builds upon the persistent memory object library (libpmemobj) and Non-Volatile Memory Library (NVML) and implements Direct Access (DAX) capabilities to provide a speedy user-space file-system for persistent memory. Pmemfile uses the syscall intercept library to transparently redirect I/O calls to this user-space implementation without clients needing any adjustments.

The pmemfile project is ultimately aimed as a stepping stone prior to applications making direct use of persistent memory if they find the performance gains worthwhile.

Current limitations of this approach include only Linux x86_64 support and no multi-process access. There also currently is not support for extended attributes, file locks, and no I/O event notification handling.

Intel developers found pmemfile to perform up to three times faster than the EXT4 file-system with DAX support, depending upon workload.

Future work items include fork-safety, mmap() support, multi-process handling, and even better performance improvements.

Those wanting to learn more about pmemfile can see these PDF presentation slides from this week's Emebedded Linux Conference Europe with the presentation by Intel's Krzysztof Czuryło.