Persistent memory parts have roughly the same capacity, speed and cost as current DRAM, but do not lose state when the power goes out. Some of these parts are on the market today, more will be coming out over the next few years. The Linux IO and File System stack is already challenged in handling existing SSD devices at hundreds of thousands of IO's per second and these devices will be able to sustain an order of magnitude more IOP's.
This talk will give an overview of what is being proposed in standards bodies and the Linux based solutions being proposed that will help us take full advantage of these new parts.
Persistent memory is a new class of devices that can sustain millions of IOP's per part. Some of the parts on the market today fit into DRAM slots on special motherboards.
Challenges and unique uses for this technology abound, but the Linux community needs to be able to support these parts at every level of our ecosystem, starting with the storage and file system stack and going up to modifying key system libraries and applications to take advantage of the new capabilities.
This talk will give details about the work that has been pursued in standards bodies, Linux kernel based projects to provide persistent memory aware block device drivers and new file systems. It will conclude with a discussion of how these changes will impact our applications.
Speakers: Ric Wheeler