Oh great...
Instant resume from power down, straight back to that pron (or insert sensitive data) you weren't surfing! There are advantages to volatile memory.
It's all downhill from here: Avalanche spins STT-RAM
Startup Avalanche is sampling an STT-RAM chip offering DRAM/SRAM speed, persistent storage, unlimited endurance and scalability beyond 10nm. The holy unified memory chip grail is to replace DRAM, SRAM and flash with a single chip tech, offering the benefits of each without the drawbacks. Thus NAND cannot scale down psst 10nm …
-
-
Wednesday 1st July 2015 16:16 GMT Bronek Kozicki
Re: Oh great...
There is much bigger problem since all current memory architectures make lots of distinctions between filesystems and RAM. Some of these distinctions will have to go (imagine filesystem optimized for low latency RAM) but some will remain (it is not convenient to have persistent memory only).
-
Wednesday 1st July 2015 18:00 GMT Martin Gregorie
Re: Oh great...
I dunno, single level storage, which is what you're talking about, works pretty well and has been around for a while. It first surfaced around 1970 as IBM's Future Series project, which was canned for marketing reasons, later surfacing in the 1979 as the System/38 and getting a refresh in 1988 as the AS/400 series. It currently lives on as the POWER 7+ series. The hardware has changed, but the operating system, OS/400 and the single level storage system that supports it are still there.
The storage basis for OS/400, which has remained fairly much unchanged since the AS/400 first appeared, is that all data and running processes share a single, flat address space which was originally mapped onto RAID5 disk arrays. The main processor of course has a chunk of RAM, but this is best thought of as a page store: there is virtually nothing in it that isn't a copy of a disk block or page (the two are synonymous) and files/databases are best thought of as memory structures that have been written to disk.
I'm not an IBM fan, but I have spend a fair time using S/38 and AS/400 kit. It worked well and was very reliable, so I'm here to tell you that single level storage is not a problem.
PS: the other well-known system that used single level storage was the Palm - remember them?
-
Sunday 5th July 2015 19:30 GMT Michael Wojcik
Re: Oh great...
IBM's Future Series project, which was canned for marketing reasons
Future Systems, not "Future Series", was killed primarily for technical reasons - namely that it was utterly infeasible on reasonably-priced hardware (even by IBM's standard for "reasonable"). The Wheelers were prominent in getting it killed, and Lynn Wheeler has posted details numerous times in forums like alt.folklore.computers.
later surfacing in the 1979 as the System/38 and getting a refresh in 1988 as the AS/400 series
S/38 inherited a lot of design elements and technology from FS, but it's not really accurate to say FS "resurfaced" as S/38 or OS/400. Direct descendants, sure; but hardly the same thing.
But yes, OS/400 - later iSeries and System i, now just "i" - is the quintessential example of Single-Level Store, and it works very effectively. Obviously there's a performance cost but there's much to recommend in SLS and the other abstractions enforced by OS/400.
Even without SLS you can treat volatile and non-volatile storage as a unified hierarchy of pages with different attributes. AIX did this as far back as AIX 3 with its unified VMM: all access to RAM or disk went through the VMM, which treated RAM as backing store for disk.
-
-
-
-
Wednesday 1st July 2015 21:14 GMT Sgt_Oddball
Will no one think of the marketers?
There's so many different competing replacements for ram/storage where will the fun end and we get down to having an actual workable solution instead of all these different routes to the sane problem.
Besides once any of this stuff hits the market it'll be a few years before it can even start to get utilised fully.
Biting the hand that feeds IT
