Microsoft is breaking new ground with Longhorn, successor to XP. The upcoming WinFS file system will be the first to be context-dependent, and promises to make long search times and wasted memory a thing of the past. THG compares it to FAT and NTFS.
The WinFS File System For Windows Longhorn: Faster & Smarter
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
It talks very little about the actual WinFS technology it’s supposed to cover. I’m not suprised, since Microsoft has kept it pretty well under wraps. However, it makes you wonder if this article’s sole reason for existence is to get hits.
Jared
While not the biggest Windows fan, I think they’re onto a good thing with WinFS. In fact, it’s what I looking forward the most on Longhorn. The new paradigm is really interesting.
“However, this luxury comes at a price for XP users: even with FAT 32, XP can only format partitions of up to 32 GB. Larger volumes can only be edited if they were created by other systems. Of course, partitions of this size only make sense in isolated cases anyway.”
The above quote from the article was surprising to me. I’ve formatted partitions up to 160GB on Mac OS X with HFS+ and don’t even know if there is a limit on size. And if I have more than 100GB of related files, I want them on the same partition.
As to the future file system, it seems like vaporware at this point.
Also, I would not trust a file system to SQL Server in its current incarnation.
Aside from the last section most of this article was about current window FS’s.
Yes. I guess they want to promote NTFS over FAT32. NTFS is more safe and secure anyway, although it’s less compatible with other OSes.
“…and promises to make long search times and wasted memory a thing of the past”
‘locate’ + reiserfs > winfs
I’m not really surprised, THG have been piss poor for a long time now. They seem to have zero quality control, and articles full of errors.
Is M$ claiming to be the first to do something Be did over 8 years ago? Actually, this has been predicted on Slashdot for the past 2 years, so I guess it’s not a surprise.
BeOS already employs a DB based FS. Tom better do his research before printing such articles.
BEOS beat you to it!
You meant to say “Microsoft copied it”. That is what that company does. Hey, it is a good buisness model.
Man you guys are almost as bad the Linux zealots here. In the end WinFS will rock and people will find uses for it. Stop bashing MS for no good reason. Stick with what you lick and keep your nasty comments to yourselves.
I would have liked a section on other database based FS’s currently in deployment/development – like BFS (OpenBFS would be more likely?) and Reiser4.
Also in the first screenshot, am I the only one to find the GUI on that search function a tad complicated for the regular user – it seems so far that Longhorn will be a step backwards in usability rather than forward. And yes I have tried it, I ran a Milestone build a while back, and it looks promising technologywise, but I can’t say I’m impressed by it’s look and feel.
But time will tell.
> In the end WinFS will rock and people will find uses for it.
Interesting…Does this mean that it won’t be inherently useful? A solution waiting for a problem? You mean that MS will create a technology that no-one actually needs, and that requires the user to “find” (no pun intended) a use for it. Maybe the WinFS can help? A file system that will recursively find a use for itself. Now that’s innovation.
The BFS that was released into production, in R3 and later, was did not have a database engine at its core. It simply indexed certain attributes to make it faster to search. Semantically, it was still a hierarchical filesystem, and files were still presented to the user in a hierarchical manner. WinFS uses an actual SQL database at its core, and largely eliminates the concept of hierarchical files. Its not a new idea, just a logical extension of the entire attribute mechanism. However, it is the first implementation of a database filesystem in a production OS.
“WinFS uses an actual SQL database at its core, and largely eliminates the concept of hierarchical files. Its not a new idea, just a logical extension of the entire attribute mechanism. However, it is the first implementation of a database filesystem in a production OS.”
Actually, you’re wrong there.
IBM was first with its SFS and BFS (Shared File System and Byte File Sytem for the ignoscenti) on VM/ESA. Which incidentally had a Reusable Server Kernel – a network-aware module – that was a database management system server.
The only problem is that they never really seem to have taken it beyond the network file system stage.
Stop it with the zealot type attacks and assumptions. The guy never implied that it would not be useful. Just that people will and shall find uses for it.
ANYTHING’s better than plain FAT lol…
P.S. Don’t get mad because the market picked a superior product over BeOS. Get used to the fact that BeOS is dead and will more then likely ( am about 99.9% sure on this ) never come-back in any meanful numbers. I doubt that it would even compete with FreeBSD or Linux or even OS-X in terms of number of users even with Zeta at the helm let only with Windows.
“…The guy never implied…”
By “the guy” you meant YOU right, jeez… at least try to be less obvious. Either that or you have some sever multiple personality disorder brewing…
The original B[e]FS was a database. Eventually it was decided to make it merely database-like, because the maintenance of an actual database incurred too many performance penalties for too little gain. Let’s face it, a file system is not NEARLY using all of the power of relational calculus, SQL etc…in fact, even a simple Prolog-type database could do the exact same thing (though I’m not suggesting it, for obvious reasons). While this sounds nifty, I don’t think it’s too much worth it (although I suppose it makes journaling easier).
Also, databases are complex things. Filesystems are much simpler–what happens if SQL Server breaks or is exploited?
to get rid of the crud on the system than to hav it sorted? Although a DB filesystem sounds good I wonder is there not a better solution.
Awww man… you crack me up… 😀
Not “filesystem”, so it’s more like an indexing layer (db) on top of NTFS. Very suitable for DRM.
The things almost always looks better in paper than in use.
Every time Microsoft release a new product it present also a loot of hype, it’s normal, but I’m still using from them Win98SE.
Let’s see how this beast will behave prior to make assumptions.
I’ve said it before, I’ll say it again… Using a database as a file system is a dumb idea. Abbreviated list of reasons:
– Namespace collisions (if you avoid then somehow, it’s still tricky then to select objects.
– Performance limitations of having a DB server in memory *all the time*. I run MySQL on my Linux box, because I need the functionality for actual work, not just to sustain the system.
– Horrible memory requirements for very large numbers of files & large data sizes.
– Overheads for having to index data in the first place.
– Overheads for all the meta-data drifting around.
– Overheads to users having to specify meta-data at length every time they save a file.
– Processing pain trying to simply search mountains of meta-data for an entire file system full of thousands of files, each with multiple items of meta-data.
– Relational DB, meet heirarchical legacy file systems and user understanding. Enjoy.
Most of all, searching is not a viable alternative to structure. Complex tasks can’t be performed this way for regular users, since directories and filenames allow for users to produce strong file groupings and heirarchies of their own. Most of all, users *do not want* to Google to find their own files. They most especially don’t need to lose files because they forget the specific keywords they used to save it with, or when they did it.
It’s not that there isn’t a place for meta-data searching. It just isn’t a file system and shouldn’t be a file system. Making a more advanced search facility compared to what exists in current file selectors, allowing for meta-data searching is probably a good thing, but performance issues abound.
It’ll be interesting to see what Reiser 4 does, though…
‘locate’ + reiserfs > winfs
‘locate’ + files added since last updatedb == inaccurate searches.
The search that fails to find the file you want because it relies on a database that isn’t updated in realtime is infinitely slower than the one that finds the fike you want.
Despite these advantages, pre-Windows 2000 NTFS still fails to meet all the requirements of current systems. The NTFS system distributed with Windows NT caps the length of automatically assigned partition names at 26 letters (drives A to Z). Plus, partition changes always require a reboot. There’s also the fact that NTFS volume information is stored in the registry, making things really complicated if you want to use a disk with another system.
The improvements to NTFS in Windows XP are slight compared to Windows 2000. Data throughput has been raised, and instead of a fixed 512-byte cluster size, it now lets you define variable cluster sizes.
Bollocks.
Partition changes do _not_ always require a reboot. Only if the disk could not be unmounted (always true with the system disk, possibly true with other disks if they have open files).
Disks with NTFS partitions can be quite happily moved between machines – running Disk Administrator will insert all the necessary information into the registry. It can be a bit trickier if there are software RAID arrays, but even that is doable.
Custom cluster sizes have been possible since at least NT4 (albeit only from the command line format).
Welcome to the world of commercial press! (In conventional print and TV press, they want to get readership/audience too, for the ad money)
To BeOS zealot:
1) BeOS didn’t used a real database in its file system. It found a nice, totally non-new way of storing stuff like atrributes. But if BFS is a database… then a lot of things would be a database. It is database like, but not technically comparable with WinFS (from what we hear, unless Yukon is a BFS clone, as oppose a new version of MS SQL, which is unlikely).
2) BeOS wasn’t the first with database-like file systems to begin with! There are many OS’s – experimental and (huge) commercial (failures) that is very similar to BFS.
From the article: “All readable and writeable disks need a file system – usually a FAT (File Allocation Table).”
huh?
you have become quite windows friendly these times… i’ve read osnews a lot 4 years ago, this was a much more windows-sceptical site back then… which was a lot better… kind of strange alternate os mag, praising windows… 2 thumbs dowwn!
btw… XP stands for eXPired
WinFS does not replace FAT and NTFS; its an add-on feature. NTSF will be the only file system for MS products in the futute. NTFS already has very advanced features, like ReiserFS.
It has been said before in this thread and the post preceding me has said it again… yet it seemed to be ignored. I wonder if the above would be ignored too.
Personally I think this site is hugely valuable *because* of the fact it’s Windows (and everything else) friendly.
What I need is a reliable source of information, which also manages not through random biases into their reporting, but instead will look at the facts to decide if something is good or bad.
Windows (Linux/BSD/BeOS/QNX/Whatever) has it’s place in fulfilling a function. And it should never do more than fulfil a function – if you want a political statement… go run for office. If you want something to help you use your computer, go for an OS.
Finally. WinFS stands for Windows Future Storage and is an addon, not Windows File System or blah… This was the last thing we heard about WinFS from WINHEC articles.
blah
This has to be the worst article ever. I can’t believe I actually read it and wasted 4 min on it. It has absolutely NO useful information and is full of errors and stupid/huge ads. Person who wrote it has a marginal knowledge of file systems.
MS said it at WinHEC that WinFS will lie on TOP of NTFS!
Eugenia, how about reading articles before posting them?!
> 1) BeOS didn’t used a real database in its file system.
It did, way back. once again at that time it wasn’t suitable on current hardware, so it was dropped (RIP zookeeper).
> but not technically comparable with WinFS
Don’t compare something that is proven to work, in use, to something that doesn’t exist yet, please.
> 2) BeOS wasn’t the first with database-like file systems to begin with!
True. Many people tell bells and wistles about reiserfs… but XFS had it all even before BFS !
Just BFS was the first for a general public OS.
> huh?
I agree, I don’t think LPs did have a filesystem back in the days =)
To everyone that pointed BFS/OpenBFS is not a database file-system, you’re completelly right. They are actually better thn that as it give you the fast searches and metadata searches without the overhead of a full blown database running.
I is worth nothing that up to DR8 (Developer Release 8 – Although I am not sure this is the cirrect version), the BeOS file system *WAS A REAL DATABASE* and that was dropped due to the overhead introduced.
-Bruno
Then you should read more carafuly, troll.
http://www.theregister.co.uk/content/4/30670.html
This article feels more like advertising than real analysis. Firstly, it claims that WinFS is a real filesystem. The last time I checked, WinFS was intended to be an add-on to NTFS. It is not a filesystem at all, just a VFS layer.
It also claims that Linux needs FAT16 to share a filesystem with DOS/Windows. Linux supports FAT32 very well, and it has done so for many years. It can also read and write (although writing is still experimental) NTFS.
The article summary states that “The upcoming WinFS file system will be the first to be context-dependent,” which isn’t true either. The Be Filesystem is context-dependent, although contrary to popular belief it is based on metadata, not a database.
A filesystem that looks truly interesting is Reiser4 [ http://www.namesys.com/ ]. According to its Web site, it is due at the end of this month (although its last deadline was September last year), placing it well ahead of WinFS. Unlike WinFS, Reiser4 is a database filesystem, with a modular plug-in architecture that allows it to be extended to include other features. The Reiser4 Whitepaper [http://www.namesys.com/v4/v4.html] is an interesting read.
eepp, before you name others a troll, try to find the correct way of writing “carafuly’…
Himm..
You, a careful troll!
Windows (Linux/BSD/BeOS/QNX/Whatever) has it’s place in fulfilling a function. And it should never do more than fulfil a function – if you want a political statement… go run for office.
If MicroSoft was a responsible corporate citizen that didn’t try to squash competition by misusing its position as a monopolist, people wouldn’t need to be political about it.
Your attitude reminds me of the main character in the play “Mephisto” by Thomas Mann. He keeps repeating that he doesn’t need to get involved in politics since he’s an actor and nothing else, that he should only be concerned with his art. Unfortunately, he happens to be in Germany during the 30s…eventually, politics catches up with him.
To veer further off-topic, I think politics should definitely not be left to politicians. Getting involved in matters political is part of being human and living in a society (aren’t we called the “political animal”). As citizens of democracies I think it is our duty to get involved. Because you may not want to play politics with computing, but you can bet that MS does (and has – notice how the Bush administration has all but completely nullified the verdict of the MS anti-trust trial).
In other words, you can try to avoid politics, but that doesn’t mean that politics will avoid you…
Oracle had their own db-based file system as well called ifs. I dont know if its still available or not.
‘locate’ + files added since last updatedb == inaccurate searches.
This will be solved by Medusa actually. Additionally, Medusa does fulltext indexing and attributes. Updates will be made incremental and once feasible, it should be possible to make updates whenever they occure (like FAM).
The usage of Medusa seems to be actually the same as WinFS: Locating files fast and beeing able to locate them by more than their name. Medusa is also no vaporware but works. In fact it was included in GNOME 1.4 once. It is still under development. Together with the interesting Reiser4, this really opens up a lot of possibilities for the near future. Microsoft might not be much ahead this time, but we’ll see.
This way of working with files is definitely the future and it absolutely doesn’t matter who had this first… What matters is, that this will come and improve computing for everyone. Some will be among the first to implement it, others will follow.
“However, this luxury comes at a price for XP users: even with FAT 32, XP can only format partitions of up to 32 GB. Larger volumes can only be edited if they were created by other systems. Of course, partitions of this size only make sense in isolated cases anyway.”
The above quote from the article was surprising to me. I’ve formatted partitions up to 160GB on Mac OS X with HFS+ and don’t even know if there is a limit on size. And if I have more than 100GB of related files, I want them on the same partition.
I think it is important to keep in mind that we are talking about Fat32 which is NOT the default file system for XP. XP uses NTFS. Under NTFS there is not this 32 GB limitation problem.
As far as Linux and NTFS there is the ability to read from an NTFS partition. Further, there is the ability to utilize Linux partitions in Windows (see: http://uranus.it.swin.edu.au/~jn/linux/explore2fs.htm).
For those of you saying that THG has gone downhill…I could not agree more. I stopped reading it a while back when I found about half of what he was saying about RDRAM:
A. Didn’t make sense
B. Contradicted himself
C. Went against pretty much everything else I was reading
D. Later went against my own experiences.
Further there were a number of just off-the-wall reviews that I read regarding Athlon XP boards that I just found totally innacurate.
Those of you who swear by Tom’s and only use it to guide you, please start looking around. The info on THG seems very biased and approaches fiction.
you have become quite windows friendly these times… i’ve read osnews a lot 4 years ago, this was a much more windows-sceptical site back then… which was a lot better… kind of strange alternate os mag, praising windows… 2 thumbs dowwn!
Last I checked OS News was a web site devouted to new about operating systems and the last I checked, Windows is considered an OS (whether you like Windows or not is irrelevant).
OS News is necessarily being Windows friendly by posting info about Windows. I think it is better to have articles about Windows rather than just “alternative OSes”
If you are really so opposed to reading about Windows…don’t read the Windows articles.
If you are really so opposed to reading about Windows…don’t read the Windows articles.
I think the criticism that was made is not so much about the quantity of Windows articles, but rather the overall positive tone they have, while articles about Linux generally seem more critical. However, personally I can’t say that this is because the editors are more “Windows-friendly”, or rather because they simply are more critical of Linux because they want it to succeed and don’t want they bias to be apparent. Just like the military leader who’s harder on his son serving under him than with his other soldiers, in order to counter any appearance of favoritism…
Calling me a troll for agreeing with you. Hrm… Is this some sort of new attitude?
Again, if you are opposed to the Windows articles don’t read them.
If you think that the Windows articles are too postive fine…don’t read them.
Further:
Keep in mind, a lot of the articles that come in to OS News are linked in from other sources. It is not OS News that is at fault…it is just a reality that there are far more Windows users out there than there are Linux users. Consequently when someone writes an article about Windows getting some new widget, the spin is going to positive because the person writing the article already knows Windows. Many of the articles that are about Linux are also written by people who use Mac OS or Windows as a primary OS and “play” with Linux. Thus they don’t know what is going on REALLY with Linux. Consequently they end up writing things that are much more critical of Linux.
At last MS found again nice feature which may force people to upgrade hardware.
Recession in hardware sales is very bad, as in all other sales.
So those arguments about slowness of such solution don’t count.
Even if on 10 GHz machine with >2 GB RAM we will have same experience as with FAT on 200 MHz 64 MB machine, which is absolutely acceptable from consumer POV.
Don’t try to stop progress!
For those of you saying that THG has gone downhill…I could not agree more. I stopped reading it…
Same here. THG used to be a good site, but then it started to show more bias in the last few years. Frankly, I don’t remember the last time I went there…
sorry, misunderstood.
…than to call hard drive storage space “memory.”
“NTFS organizes its centerpiece – the Master File Table (NFT) – with hidden files. The MFT manages…”
Really… first the “memory” thing, informational issues and now really obvious typos. Come on… someone proof-read these articles.
…than to call hard drive storage space “memory.”
LOL…I thought I was the only one that went looney tunes when I heard that.