Magnetic storage - Future



  • Came across this today when looking up something.  Kind of interesting what may be in store (pun intended) for the future.

    Magnetic storage - Future
    https://en.wikipedia.org/wiki/Magnetic_storage#Future



  • There are actually several non-volatile memory products right around the corner. XPoint from Intel and Micron is already out, but the first generation seems to be highly limited and slightly less spectacular than promised, but it should become very useful as it matures.



  • There are actually several non-volatile memory products right around the corner.

    Definitely a variety of "Real Great Things - Coming Soon" …..

    I've been watching for hints and first runs of Nantero's NRAM.
    According to the releases and news bits it's actually in some industry Fab plants now.

    As always, the proof is in the pudding, but the design specs all point to a form of memory that could possibly change the way systems are built.
    If there's little difference in the speed and power requirements of system RAM vs long term storage (what we call a hard drive today) why not use the fact and build a system that has nothing but RAM?  You could easily have a system with TB's of RAM.

    Certainly smaller devices like phones, cameras,etc. could leverage this stuff dramatically.

    Verrrry interesting if they pull it off in the real world.......



  • IBM or HP, I forget which, is already working with XPoint to create a new OS memory and storage subsystem that treats them the same, because they are. What you have is everything is memory and storage, just a difference in locality. You can do all kinds of interesting things. Nearly all forms of file-system cache become moot, there is no more notion of "loading files", they're already ready. To take full advantage of this new paradigm, your memory subsystem and storage system need to merge, even your filesystem needs to be aware in order to make sure stuff like alignment is proper. You also have the issue that the FS ACLs need to be respected.



  • There are a bunch of implementation, OS-reliability and security issues to be thought through. If the BIOS just goes looking for attached "memory disks" and maps them into the 64-bit physical address space seen by the CPU(s) then BIOS memcheck diags, BIOS functions that zero all of memory and the like have to be VERY careful to still understand what is built-in-local-run-time memory and what is some poor sods "memory disk" with a file-system that they care very much about.

    Similarly if the "memory disks" are mapped into physical address space by some OS loader code (or done already by the BIOS) - the OS has to be VERY careful with its physical address space management, not to accidentally zero some pages that are somebody's file system storage.


Log in to reply