Russian rides Phantom to OS immortality It's nine degrees Fahrenheit in Moscow right now. That's the kind of cold that makes a man concentrate on his work so he can keep from remembering that the tears the wind blows out of his eyes will freeze on the side of his face before they hit the ground. In the middle of this icebox, Dmitry Zavalishin is cooking up a new …
The big question is, is this availiable or just some vapor ware like windows, which you can only get if you pay huge sums of money?
Appart from what, what's so different about his approach to, let's say the LISP-Machine machine one? There you don't have any user visible RAM, but insteadt everything is done on disk. RAM is only a cache for disk.
9° F is a bit cold, but not all that bad. -9° is unpleasant but bearable. For tears to freeze on your face it has to be at least twenty below. *That's* cold.
@Christian Berger
This takes the opposite approach: everything is done in RAM, and the disk is only used for memory snapshots (and probably virtual memory).
Yeah, I bet that isn't a big performance hit (chuckle). This is frickin stupid, the disk responds in milliseconds. DRAM responds in nanoseconds. That is a 1,000,000 X performance hit. Why even bother anyway, with new solid state storage we might not even have disks anymore.
Unless that is the point. If this guy is anticipating RAM-like non-volatile storage, than maybe there is a point, but without even being able to anticipate what the hardware will look like, performance tuning would probably make whatever architecture he has come up with irrelevant.
Sorry, but unless there is some hugely important fact that the writer left out out of the story, then I call BS on this one.
(I would like to see a serious challenger to WIndows though! BTW, a Russian condescendingly talking about "communist" Linux is some hilarious irony!)
I'm not really saying "it's been done before" because the goals of the Cray were far different. But, from my understanding.. Cray 1 had no memory management hardware, Cray insisted because an MMU would slow the machine down. My understanding is later Crays DID have an MMU; since they had memory management anyway, they went full-tilt and treated EVERYTHING as flat storage -- as in this article, no files* , everything just appeared as system memory.. the portion that WAS in system memory just happened to be faster. I have no particular insight or suggestion from this, but he may want to look into that just so he isn't doomed to repeat any mistakes the Cray may have made.
*Well, just like this Russian OS, I think the Cray OS had some concept of files after-the-fact, to make apps that require files happy.
Os level serialisation and memory persistence instead of file? Fantastic idea? Not so sure.
First of all, The computer is the network. Try to move data around without the concepts of file, stream or message. End of the day, they will have to be added to his wonderful contraption so that the computer can communicate effectively with other systems.
This will also be the day when his "elegant" "better than Windows" system will understand why Linux and Unixes have came back from the brink of extinction and have continued to gain inroads in the market at the expense of Windows - they make network programming indistinguishable from file programming. I agree, the file paradigm sucks, however we are yet to find anything better for network communications.
Second, I have had to work with people who carried this idea (nailed onto an Agile Java Buzzword Compliance banner). Their project was wonderful (at power point level). It also failed miserably in the field. The moment the serialised data size went above a certain threshold it went belly up and stayed belly up destroying all data in the process.
... this sort of melodrama:
"It's nine degrees Fahrenheit in Moscow right now. That's the kind of cold that makes a man concentrate on his work so he can keep from remembering that the tears the wind blows out of his eyes will freeze on the side of his face before they hit the ground."
But surely only if he insists on coding outdoors? I expect he'll just turn up the heating or put on a sweater.
If the state is by default persisted then how do you get rid of a program that's crashed.
Some sort of program manager I'd guess. But since you don't do manual saves to disk and it just persists it's current state, you'll not be recovering whatever you were working on, or even any configuration changes you made.
Actually I started work on a multiprocessor distributed concept like this in 1987. It's definitely the way forward.
Linux is a copy of a 1976 OS trying to Mimic OS X and Windows.
OS X (it's self based on BSD, a free copy of a 1976 OS) and Windows can only get more bloated, more unreliable and poorer performance.
Nothing really new here as far as I can see. Many of the old PDA OSs of yore spent a lot of time ensuring that state was preserved, at the very least ensuring the system was consistent in case of a sudden loss of power.
These days the easiest way to achieve this sort of thing is just by using virtual memory, and using the swap file as your memory snapshot.
It'll be interesting to see how well it fares on power limited devices such as a mobile, as all that writing state to flash will kill battery and reduce the lifespan of the flash. I'd also be interested in seeing how well it copes with code that isn't expecting to 'run forever' and contains memory leaks, and other common bugs that don't surface for short lived exectution spans.
I'm sure he's thought of all this, of course! ;-)
>"... and it is difficult to surpass it [iPhone] technologically"
Indeed /:-|
I've developed a few simple database classes in Python for my web application development which means my applications don't have to open/read/write/close files directly in order to persist data. This gives the benefit of managed concurrent access too, in case more than one user wants to update the same record at the same time. Any web application developer who uses a back end database and a reasonably useful database API does something similar.
Interesting idea to develop a full OS around the concept possibly for mobile/embedded use, though the concept isn't a new one.
Those that can't learn from history are doomed to re-create it.
http://en.wikipedia.org/wiki/Canon_Cat
http://en.wikipedia.org/wiki/Jef_Raskin
http://en.wikipedia.org/wiki/The_Humane_Interface
And I'm not even sure Raskin didn't picked his ideas somewhere. See, those big irons of the 60's were neat beacause the core memory that was used in those days was by design impervious to power on-off cycles, being magnetism based. No need to compute ? Turn the beast off. Something to do ? Restart it, execution can resume at the point the operator left. I believe the space shuttle's computers work like this too.
So it's not really a novelty ; it's a mere routing around current amnesiac memory technology, to restore a long forgotten functionnality our predecessors could have taken for granted 5 decades ago.
So what happens if a program crashes? In a traditional system, as long as the file(s) you were working on aren't actually being written to at that exact moment, they should be safe from corruption (most of the time anyway!)... but if these snapshots are more-or-less continuous, a crash or even a rather sudden application close would presumably damage the file being worked on.
Apart from that, it sounds like an interesting idea... especially for mobile devices or hardware like routers (which mostly seem to run linux at the moment if they have an OS).
i might be wrong, with the VM porting going on, developers might like it enough to really use it, but what about porting the snapshotting to other systems?
could it be implemented in Linux or BSD/Darwin (and hence OS X) with a custom file-system?
or maybe I read too much into last night's top gear best of where they said the FCX Clarity was the most important car of the future because it is the same as the cars of today.
This post has been deleted by its author
As Kanhef says: "Pish-Tosh!"
It's a mere 23F below freezing point, roughly -12C. Here in Stockholm today it's a mere -5C and we're practically walking around in our t-shirts - you should see the girls in miniskirts and tights and bosom-baring tops. Everyone's longing for -12C so we can have a decent winter for a change with crunchy snow, nice durable ice and air that gives your lungs a good scouring.
Also a freezer worth its salt (heh) is at least -18C. Which is normal up north, where they hate the raw damp namby-pamby winter here in the south. But then, they've got saunas and distill their own poteen and keep each other warm, and build hotels made of ice. And where they're going to move a whole town a ways from where it now to avoid subsidence.
Now if it gets down to -30C *and* the wind is blowing, you start wrapping up. Around -40 you start thinking about keeping the kids home from school. In Vietnam they do that at +10C (+18F).
(-18C = -65F... -30C = -88F... -40C = -106F... )
And the Sami and Inuit snigger at this and laugh at the soft southern immigrants...
(Paris cos she knows all about keeping in heat...)
Oh, yeah, I forgot to mention Siberia... ;-)
I can't believe Ted the Sweartard is so credulous about this bullshit.
Storing data to disc ready to go again is one thing. Yay! 30 seconds saved each day! This is clearly much better than the "suspend" mode my machine already has. And it's not like I ever *want* to reset everything's state by rebooting, because all the software I use has zero defects.
But wait! Then there's this huge leap of imagination - suddenly you don't need "files" any more. w00t! No more serialization code! You don't need it!
Oh, unless you want to send stuff over a network.
Fail and you, Ted, fail and you.
Hang on a minute... Disk on an iPhone (any phone, netbook...) is Flash, which has limited write capacity, and Phantom is going to continuously snapshot memory?
I'm not convinced this is needed at the OS level, but it would be nice if more languages supported proper persistent objects.
Why is it that everyone seems to paint linux as a single unified entity that is endlessly struggling against windows, and doomed to failure? It seems to me that it is doing pretty well on everything except the "average computer user's" desktop where they have come to expect windows and won't settle for anything else. Any new OS will be in the same position... it doesn't matter that its creator think it is going to be better than windows: the market simply will not care, because it is not windows.
Now, any other system competing with the iPhone... that's a totally different matter. It combines some very nice hardware engineering with a very slick UI. You can't beat that without some serious, skilled hardware and UI engineers, but if you do have access to those sorts of people, the underlying OS is utterly irrelevant so long as it does its job. The GPhone looks like a nasty plasticy prototype running partially finished beta-like software. How on earth did anyone ever expect it to fight the iPhone?
Clive Sinclair had a laptop called the Z88 in 1988. That would remember what you were doing when you switched it back on. You could be mid-word in the middle of a letter, switch it off, switch it back on next day and the cursor would be flashing, waiting for the next letter in the word.
It also allowed you to change its batteries (4 AA cells if I remember) while in the middle of typing.
Anyway , I don't know if that was a function of the OS or of the machine.
Have a look at what used to be NonStop Services for (spit) SCO UNIX (or UNIXware).
You don't save everything every time it is updated (unlike the toy here).
You do save a sufficient subset, eg externally visible externally relevant state on a periodic or event driven basis (such as a process and its associated PCBs etc when it finishes its timeslice or whatever).
Part of the same setup also offered single system image clusters, again for SCO Unix.
Shame it was for SCO Unix.
"The moment the serialised data size went above a certain threshold it went belly up and stayed belly up destroying all data in the process." ... By Anton Ivanov Posted Tuesday 3rd February 2009 07:00 GMT
The same problem as experienced by the Junkie, overwhelmed by something over which/to which they have seeded/ceded Control. It is though, but a Natural Weakness Barrier which most efficiently sorts out the Weeds from the Flowers, relegating some to Compost and Slurry and others to Harvest and Future Resistant to Useless Junk Strains.
Has anyone here, any real Idea of what is being developed Eastwards for Westwards transmission/BroadBandCasting ...... or is IT all Guesswork and Spooky Speculation?
Windows Kernel Control would give you Stealthy Proxy Power Points/Virtual Nodes all over the Place in Space.
"Not sure I'd want an OS where any third-party program errors / memory leaks are preserved for all eternity - unless it's open source it benefits from the occasional forced reboot."
Oh yeah cause open source software NEVER needs restarting as its perfect in every way.
how are people reading this considered high level users with comments like that coming out?
How is this really any different to the hibernation mode Windows now comes with? I never shut my computer down anymore, I always select "Hibernate" which writes the state of memory (and probably some other things it needs to re-init) to disk and then shuts the computer down fully - not just suspend, but a full shutdown.
All this seems to do is offer an automated snapshot system at the expense of requiring substantial rewrites of software.
Hmm seems some people, xjf I'm looking at you here, can't convert right
0 C = 32 F
0 F = -18 C
C to F = (1.8 *C)+32
F to C = (F-32)/1.8
So (-18C = -65F... -30C = -88F... -40C = -106F... ) isn't that it
-18C = 0F ; -30C = -22F ; -40C = -40F
-18F = -27.77C ; -30F = -34.44C ; -40F = -40C
So I've no idea what calculations you were using!
Back on topic
Good idea but can you get it to work in practice and SELL it, can see uses for militry kit for sure.
Ok, this needs to hit slashdot because people here just aren't getting it. This isn't about saving '30 seconds' of boot up time, nor just taking snapshots so it's "just like hibernation". If the OS can save a consistent *incremental* memory snapshot to disk in a usable fashion, you have many benefits over current systems:
This automatically applies to all applications, with no danger of some driver or program preventing your machine hibernating or resuming properly.
Hibernating and resuming will be near instant, making it perfect for hand held devices, DVR's, games consoles. Heck, even thin clients would benefit from instant boot.
Instant boot times also have huge benefits when it comes to power saving, which is a huge topic these days. Imagine a rack of servers where their boot time is fast enough that you can afford to simply power off unneeded machines under low load conditions.
The idea has a massive amount of potential. The only question is whether he's got the ability to pull it off.
If this ever succeeds and all those Windows apps get ported, you're going to need a big "Re-Boot" button somewhere. Guess you could do that in software, or maybe run multiple VMs, but then you've sort of defeated the whole object of the OS (pun intended).
Yep, error recovery is going to be the real problem. It'll be just like having a wife next to you all the time to remind you forever of all your little mistakes.
Sounds like if you could couple this with ZFS, you'd have a complete history of everything you do on your computer. Phantom to snapshot the changes to program state, ZFS to snapshot the disk changes, all that's missing is a big timeline with a pointer on it so you can roll your machine state back to last Friday (as in "Damnit, I know this was working when I left on Friday!...")
Sorry mate, but both the core ideas in Phantom were first implemented over 35 years ago and remain in production use today:
NO FILES: first implemented around 1970 by IBM's Future Series. FS was canned, to reappear in the late '80s as System/38 and its descendants the AS/400 and System i. These machines have no filing system, only virtual memory. File-type objects are in reality persistent in-memory tables.
PROGRAM STATE PRESERVATION: first implemented in 1975 in the Tandem Non-Stop series of fault tolerant computers. HP's Integrity series are linear descendants. The internal state of all programs is saved and replicated every time an externally recognisable event occurs to a program, so the program can always be restarted from that point. This is better than Phantom since the replication is to another copy of the program in a separate CPU - if the active program's CPU fails the backup copy takes over seamlessly a millisecond or so later.
Both systems maintain separation between program and data so recovery from programming errors is never more of a problem that it would be on a conventional *NIX or Windows box.
Am I incredibly dense or is it really obvious that a) there will probably be a hard reset option for when the system is in an unrecoverable state and b) it's actually most useful for instantly bringing back servers that die because of ridiculously unlikely circumstances.
It's naive to criticise it for obvious things when you haven't seen it, because it's likely that the dev has already thought of it if it's that obvious. No? Presumably if you have snapshots of the system leading up to a crash and working fine, you could possibly even resume applications from a previous point in time if you wanted to?
I think the big win here would be for the boffins at DARPA and for the larger governments who need an OS that could reboot immediately to it's previous state after a massive EMP discharge. If you can harden the system and the flash memory you could significantly reduce the time needed to get critical systems online and the vehicle or device back in the fight faster. For warships and tanks that may not be that large of a concern but any modern fighter plane or bomber that looses all computer control is basically a flying brick until those systems can come back online, every second counts there.
We may never see it in the commercial sector but there is a niche for it, and a highly lucrative one at that.
Explicit serialization of data (ie a "file format") has huge memory advantages.
A PNG is stored zlib compressed in a fairly efficient file format, but in order to blit it onto the screen it has to be expanded to 24 bits per pixel. The same applies to just about any data structure you care to think of - hashtables, linked lists, whatever.
I think this is a nice idea and I can see the advantage for portable devices that can power off/on quickly, but doing away with files altogether is missing the point of files methinks.
He admits he can't siphon off users from the Windows world - I'll go one step further and say the Mac users will remain faithful to Apple... That leaves his target demographic as those who are using some remaining OS... and then he proceeds to insulting their choice. Brilliant.
These are the languages this guy plans to support (he's even explicitly ruling out C)... so he's just designing a browser then? Then there is the fact that any remote connexion will be killed by the shutdown... so we're left with a RIA-only machine that can't keep connections alive... what is the use of a persistent "OS", in this context? Appart from ruining SSDs? Agreed, in the odd powercut situation, you won't loose your unfinished comment on El Reg. Waitaminit, my FireFox (SeaMonkey, IceWeasel, SunSquirrel, ClayOtter, whatever) does that already. Then there is the economy on data conversions due to the "all-in-memory" approach, which can improve speed. But this is likely to be more than nixed by the VM form-factor... and the constant snapshotting.
Also, a single address space sounds like a big vuln. No pointer=no access? Well, no-one knows my IP, so I'm safe from crackers... or am I?
Also, he braggs a lot about UNIX and Linux, but after browsing his website I have the distinct impression that he mostly knows about Windows.
It's called "PalmOS" (well the *old* PalmOS at any rate, don't know about this fancy new one).
PalmOS didn't have files, didn't separate active and storage memory, and you could easily write programs that returned to the same state they were in when you left them (though this was not a requirement). The only thing it was missing was actual multitasking (in the sense of being able to run background processes).
It is an optimisation allowing the software recognise the limitations of slow but capacious disk data storage vs fast but more limited RAM data storage.
Persistence might be applicable to small devices like a phone, but I can't see it working for general apps.
Hiding too much detail results in bloat (hello windows, acrobat reader...) If you want performance, you have to know what you're dealing with, which is why C is still used now and how computers in the 1980's managed to get a spreadsheet , word-processor or database functionality into 64k - 640k of RAM.
This is the really important headache part for those snooping agencies/phorm classes which rely on collecting Third Party information for analysis and subsequent reaction ......."One can argue that VM makes system run slowly, but nowadays this problem is solved with effective JIT compilers, so we don’t expect real degradation due to the VM." ..... http://www.dz.ru/en/solutions/phantom/
For that easily evolves and grows into Spontaneous Support for Simulations of Stimulated and Stimulating ProAction .....ZerodDay Trading with HyperRadioProActive Network Elements ..... Capable of Rendering Virtually and dDelivering Pragmatically, via ITs AIR&dDs from Deep Cover Cloud Bases with NIRobotIQ Quantum Communication TelePort Thought Function, a Reality Beta for Global CoInterIndependent Society Future Builds ....... ITERative Ground Projects with Virtual Space ProgramMIng ..... CyberIntelAIgent Strings and Thread/Tangents and Trialling TrAILs 42 Follow with Third Party, Positively Reinforcing, Constructive Support ..... which in many cases/all cases can also be supplied via Zero Reaction this allowing for Unhindered Further Progress in Fields which may Require a Certain Kind of MaJIC to be XXXXercised and Shared Transparently.
It were called "Persistance Programming" when I were a lad.
It can be implemented at language level - APL has had external variables (mapped onto files) for donkey's years.
Palm OS, yes, but you did lose data without battery power.
As someone above said, virtual memory, mapped to solid state disks is the future, but you still need a "hard reboot" to stop really nasty problems and/or a better way of programming.
I can't remember the exact figure, but it was 80% of code is devoted to transforming data from a version useful for storage into a form useful for processing - and vice versa.
If the OS is handling all storage, then we can get on with the real work - which sadly is spending 90% of the remaining coding time transforming data from a form useful for processing into something which is "splatted" onto a screen...
And before Palm was the Newton. Persistant coding, all data is structured and serialised... we've seen it all...
Called 'em soups in my day, they did. Made 'em do everything. Even had one that would work underwater. And had wireless. Could read the screen in daylight, too. Not like these modern lumps of dingoes kidneys.
I don't know, kids of today, thinking they have new ideas. "Too late!" I says to them, "We've already invented everything." And then they laugh at me. Me! I'll show them! Right after me custard.
Mine's the tatty one next to the walking stick.
As others have noted, this is an interesting idea, but has been done many times before. EROS were particularly fun with it, going to trade shows and kicking out the power lead to their machine and showing how quickly it recovered - and challenging IBM et al to try the same test!
Protecting against that failure mode - which can also be done with a UPS - means you're vulnerable to application bugs destroying all of your data. If all of your data is "live" and persistant, then an application crash wipes it all out. You might be able to restore to the previous state, but that might just crash again. No files = no backups.
In any case, the programming model is sufficiently different from normal OSes that this will require application support. Which it doesn't have. And no files means no migration path for existing data, even if the apps were there. Which means it's dead in the water except, perhaps, for embedded applications. So, anyone think that the US/UK military is going to buy from the Russians? No, nor do I...
I'd pay for one when it is established, I liked my Atari OS-on-a-ROM. The 6 minutes I spend booting up each morning are the low point of my office day.
To be fair, only 33% of the delay is due to windows itself, the rest is thanks to our company OS, which overwrite many of the windows routines to help me stop using standard OS functionality like loading programs, reading USB sticks or listening to music.
At least I can while away the wait with copious opportunities to type in passwords.
-40F is *cold*.
Really seriously though...
Good on ya Ted. That's the most interesting OS possibility I've seen since Atheos mysteriously disappeared. I'll keep my eyes peeled.
Prior art and some OSes that have no market share aside... if I can download it and run it on an x86 or x64 box then I will and I can see some ways it is good.
If you seriously believe Russia will honour the American loose-patent-as-business-tactic then you are dreaming.
I don't know if it is vapourware, hope not. Just having an OS to play with that keeps persistent state will be fun and possibly extremely useful. I hope it is a real product.
All the nay-sayers and it's-been-done-before-criers should consider that none of us have easy access to such an OS outside of closed PDAs/cells/Crays/etc. Wouldn't it be nice if we did?
Of course, I have no idea from the article about the license.
' "You can not compete with Windows, repeating it," Zavalishin says. "It is impossible to compete with Windows, creating a functionally weaker system, such as Linux." '
Linux is functionally equivalent to Windows. It's less popular due to the prevalence of Microsoft that makes it easier for the public to use the Windows OS. Zavalishin is a hack.
I have written a custom operating system that did not have the concept of a file. It kept states, data, and parameters in a database which was continuously saved to non-volatile EEPROM memory. If the computer was turned off the data was left in place and when the on switch was pressed it started from where it left off. This was for an expensive, custom, specific-purpose built computer and such an OS would be very useless for general purpose use. The concept and functionality of a stored file can not be replaced.
the concept of having the file as a persistent object is pretty ingenious, but how exactly is the file stored to a storage device such as a hard disk or something ? ... like in most OS a filesystem is used , if a filesystem exists in Phantom , then what would be its nature ??
Who the hell needs a new bloody bleeding o/s when existing server h/w and o/s (running Oracle) gives me the following uptimes?
7:54am up 625 days, 10:49, 9 users, load average: 0.63, 0.54, 0.47
7:55am up 745 days, 15:54, 1 user, load average: 0.59, 0.68, 0.71
7:56am up 691 days, 21:01, 3 users, load average: 0.50, 0.61, 0.51
3 different servers - ranging from development to production OLAP and OLTP.
I've read that Phantom FAQ.. and it sounds just like yet another typical J2EE developer that thinks o-o is the be-all and end-all and think that his leet o-o skillz can revolutionise the face of computing. Utterly. Fricken. Pathetic.
Mine is the one with a lead pipe in one pocket and a Unix and Linux kernels in the other.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
