Russian programmer Dmitry Zavalishin is working on a revolutionary new concept for an operating system. In Zavalishin’s Phantom OS, snapshots of the entire system will be continually saved to disk, preserving state for running applications and active data. There are no files in the traditional sense, instead a file is merely an object with persistent state. Neat!
Phantom is new, but it can currently boot, execute applications, and continually save system state to disk. A GUI is planned, as are cross-development tools. The new OS is planning to support interpreted scripting languages like Ruby and Python, and virtual machine programming languages like C# and Java. What, no perl?!
The saved-state nature of Phantom makes it a real contender for both the mobile market, as well as servers. Switch off your phone running Phantom and then turn it back on with no loss of data or apparent interruption in what you were doing. For server environments, a power loss may no longer be a big deal.
It’s great to see real innovation in the OS space. Challenging the status quo may mean that Phantom has an uphill battle before it, but it’s refreshing to see people really thinking outside the box.
Via El Reg
But… isn’t this what PalmOS has always done? PalmOS boots up.. runs.. and any data that is entered is persistently stored in memory and in backups as an image of what the persistent data in memory looks like so that it restored back to the same locations?
With the move towards “connected”/”cloud”/”internet”/”hosted” computing which is totally dependent on a “connected” protocol like – TCP/IP, I am not sure how this device “swich off/on” is going to take off as in “saving app states and resuming from there”.
Not that it is possible, but making it work will required adding a complex layer of abstraction in application implementation in both client / server side.
With technologies, design philosophies needs to be changed and it is already in a very slow pace.
I don’t think we’ll see a wholesale move to “cloud” computing or storage for a long time yet. We’re really only taking babysteps toward that end with things like GMail and online office applications.
For a long time yet there will be a need for local storage. Even in a cloud-based world, local storage can be used in many useful ways. As such, saving changes to a local storage medium has some tremendous value.
Actually it makes perfect sense for a cloud enviorment, instead of “saving” to disk you “save” to wherever your OS is hosted. It’s neat because you can upload as much as you need into the client, and if the computer goes Kaboom you can continue from any other computer.
I’m not holding my breath. For a desktop os, I see more disadvantages than anything of doing continual backups. When you have 8gigs of ram how much lag will build up from limited disk write bandwidth? Let’s assume a 20mb/s bandwidth (typical performance I get on my discs), it would take nearly 7 mins to dumb all ram to disk (incremental updates would be smarter but what about memory intensive applications such as video editing or finite analysis?).
Assumption “it is needed to dump ALL and EACH TIME” is wrong. Consequentally, calculation is wrong too. Keyword is “work set”.
Both Linux and Windows do something similar already but they just treat the writing to disk as only RAM. Swap and Virtual memory are pretty much the same thing as this and those don’t cause that much lag while continually saving to disk.
Keep in mind, the average Russian programmer has a computer that is most likely VERY outdated. Saving the system state is nothing when running on 128 MB or so. I’d be very interested to know what his development environment is. I seriously doubt he has anything close to 8gigs of RAM.
The main problem with computers today isn’t necessarily and exclusively the OS. It’s storage. Storage is extremely inefficient and RAM isn’t as efficient as the CPU either.
When the hard disk is completely done away with a fast solid state solution, and RAM can run at the same speed as the CPU so it doesn’t need a cache, then you’ll see performance shine and bottlenecks disappear.
The problem is the CPU has outpaced everything else. Storage needs to catch up to it.
Sorry, but statistically computers in Russia are LESS outdated, than in the USA. :) Especially in programmers’ domain. 2G RAM is usual amount, 4G is not rare.
In Russia, most computers are just two potatoes with wire strung between them.
In Soviet Russia,
computer outdates YOU!
2 gigs of ram costs like $30 here in the USA, really it’s no problem to get this amount of RAM, my computer has 2 gigs as well. Retail computers rarely come with less than 1 gig of ram here.
The problem with solid state is that it has only a limited number of writes. less aware people would go out and buy, say a 500GB SSD, and then cry after about a year because all their data has been lost. The time limit would shorten even further if you were continuosly saving the OS state to disk because it would be writing mass amounts of data, hundreds of times a day. HDD is sound, it should not be irradicated.
This has it’s place in truly mission critical apps. There is no reason though you could not build this layer between the hardware and Linux or BSD for example. It could be useful in critical control systems, flight controls, etc, but on a desktop who wants the HD burning out in 4 months, or their SSD based netbooks running for 17 minutes per 3 cells of battery. Not only that, but you would have to have two disks or you would be waiting forever while the app you load argues with the image of the same being rewritten to disk.
I’m with Mat, but will grant where instant on and instant recovery override power consumption and the need for 6 disk raid array superfast disks is a minor consideration.
TripleII
Sounds like IBM’s OS/400 operating system on the AS/400 computer. No “files” really, just one large memory address space. (I think it was 48 bits) Parts of memory (objects) could be persisted to disk and brought back after a reboot.
“parts” is usual – you can memory map such part in Unix as well. To really benefit from persistence, you need ALL of the system state to be saved. It changes way of software is created dramatically.
This should only change the way the OS runs not applications. Applications will still start up, do some stuff then shut down. Even in a persistent state environment all that will be lost is the shut down part (and the start up part after the first start up).
So the creation of applications should not change very much.
I heard that scientists are working on storage made of protein. The plan is to use protein molecules as forms storage. Also we will be able to have terrabytes to pentabytes of data in a small amount of protein the size of an M&M.
It would not be very hard to virtually partition the memory and hard drive into checksumed sections, the image snapshot would then just have to update any past of the complete OS THAT HAS CHANGED.
These “changesets” could be imaged in RAM, all I/O would be to and from the DUAL PORT RAM buffers between the OS and the HD.
THis would reduct the hard drive read/write cycles and extend the HD’s life.
After this comes out, how long will it take for the FOSS/Linux boys to “invent” this system for themselves, as seems to be their way.
Generally, you are right. And it is not even needed to calculate checksum – it is possible to tell what is changed much easier, using VM memory page attributes.
Finally! This is the way forward. Why would a traditional boot-procedure be an advantage? This new way of computing might require a slight hardware change, but; I am confident that this is the way to go.
It is definitely an interesting concept: The OS is one big virtual machine, and the ‘file system’ is merely object storage.
I don’t see this going anywhere very fast unless they can gather together a huge following. They need the huge following in order to code up all of the things necessary to get anywhere with a new OS; namely Drivers and programs (or a compatibility layer to allow running of foreign programs). Of course there is also the marketing side of getting the word out and trying to convince people that this new OS is superior enough to their current OS that they need to switch.
This is not a Windows or Linux killer (at least it won’t come close for another 10-30 years). At best I see this filling some niche in the server market. OO vs. non-OO at the os level makes no difference to the casual user.
Better if he contributed to existing FreeBSD or Linux kernels.
I would love to see an instant-on computer, I just don’t think we’ll realistically see it anytime soon.
I am not including cellphones, smart phones, or pdas in my definition of computer.
Didn’t come from and American college? How cum? Sports cars and football get in the way? Too many drunken orgies to think straight? Too many shipments of really good dope lately? Spending more time on elaborate schemes that lead nowhere? Too much space-travel gobbledygook going on? Old profs with tenure still lost in 1967 and wasting your time? Remember, the Russians used pencils to write with in space, because they worked! Is that what is wrong, or are you so busy trying to line up with Microsoft for a cushy sales job you didn’t have time to study? WTF!
First off. If you want this on Linux, simply use hibernate. Secondly, this would be VERY slow, you would have to right to RAM and HDD. Thirdly, very wide margin for data corruption and completely crash the system. Fourthly, energy consumption of contently righting to memory. Fifthly, RAM and HD would get extremely fragmented. SIXTHLY, if it works for phantom OS then Microsoft and Linux will just copy it and make it better. This is not the OS of the future, thats just crazy talk, its not even a new idea MANY portable OS’s use this method the only reason you don’t see it on desktops is because desktop applications are generally larger and more resources, so if you were to use this method on a desktop it will be very unefficient and laggy.
here we go again about whats going to kill Windows and Linux……..
Thanks for article. Realy great. Translate to MX lang and publish to my linux web pages
In my opinion, Linux is killing windows…slowly….. but safely.
Read the whitepaper before making conclusions.
(MS-Word, English)
http://dz.ru/en/documents/dz_phantom_whitepaper_11Feb2009_00.doc
I find this idea at least interesting enough to speculate upon easiness of developing apps for such an OS.