View previous topic :: View next topic |
Author |
Message |
lurid Guru
Joined: 12 Mar 2003 Posts: 595 Location: Florida
|
Posted: Tue Sep 09, 2003 12:21 pm Post subject: 'Storage' to Replace Traditional Filesystems |
|
|
Quote: | OSNews is reporting on Storage, an innovative project which aims to replace the traditional hierarchical filesystems with a new document store which is database-based (PostgreSQL). |
Hm. This looks like Longhorn to me. Microsofts idea of a searchable filesystem, rather than a hierarchal one. Screenshots of Storage show queries being typed in like "messages from bob" and all mail from Bob being listed in (what I assume is) Nautilus. Is this a good idea? Will this actually work? Personally, I'm divided.
On one hand, when Longhorn does come out, it will inevitably have problems with an untried and untested filesystem. But once the kinks are worked out, users will get used to running their computers this way. Having the same capability on Linux will make anyone switching over to Linux feel more at home.
On the other hand, this is a really stupid idea. If I know that a file is located in /usr/share/foo, why should I have to then 'search' for foo in order to get access to it? To me, this 'next generation technology' is all flash with no real benefit to the user. Having to search for files and then on top of that, wading through the results to actually find what you're looking for, seems like a big waste of time.
I think Microsofts idea is to further integrate the Web with the desktop. Users will basically being doing Google searches for information on their hard drives. The rational here is that clicking /usr/share/foo (or C:\My Documents\Work\foo, as the case may be) is more work in terms of clicks than simply typing 'foo' into a search bar. Its seems to me, though, that if I know where I'm going, extra clicks would be faster than searching, looking through results, then clicking.
Again though, having this avalible under Linux will make future converts feel right at home. And of course, being Open Source, most the bugs that Microsoft will be faced with after launch will probably be worked out in Storage by the time Longhorn is released.
What do you think? _________________ Go find a cheerleader and saw her legs off. - Nny |
|
Back to top |
|
|
Nermal Apprentice
Joined: 18 Apr 2002 Posts: 259 Location: UK
|
Posted: Tue Sep 09, 2003 12:24 pm Post subject: |
|
|
I think it will be impressive when it goes wrong |
|
Back to top |
|
|
Xiol Apprentice
Joined: 24 Jul 2003 Posts: 209 Location: UK
|
Posted: Tue Sep 09, 2003 12:31 pm Post subject: |
|
|
I thought that Storage, and Longhorn's FS, were just laid on top of the traditional filesystem.
The database just stores the metadata...
Wait, I have no idea what I'm talking about.
Think it's time to RTFA. |
|
Back to top |
|
|
robdavies Tux's lil' helper
Joined: 06 Sep 2003 Posts: 90
|
Posted: Tue Sep 09, 2003 12:32 pm Post subject: |
|
|
Storage looks like an interesting project, wonder how useful it will turn out in practice. Many users don't get on to well with locate, so they may like it.
Way back, UK mainframe firm had a contents addressable file system, might be same sort of idea.
As for M$ and Longhorn, I really do not care what they are planning. |
|
Back to top |
|
|
hook Veteran
Joined: 23 Oct 2002 Posts: 1398 Location: Ljubljana, Slovenia
|
Posted: Tue Sep 09, 2003 12:49 pm Post subject: |
|
|
yup, seems identical to M$ longhorn :]
but seriously, i think this "fs" could work only for complete noobies (on any platform!!) and people who constantly keep forgeting where they have stuff.
i don't like it, but i'm sure some guys like it ...because if there's absolutely noone who would use it, it wouldn't be there IMO.
also IMO you have to think about this "fs" not just from the "how i use it" side, but also from the "how i make it work" side ...this "fs" will require the user to rename every file he gets into a short description ...this could be VERY awkward ...imagine having millions of mp3 that are listed as numbers and have emtly id2/3 tags ...that's fun!!! ...imagine then the same situation, only not all files are the same filetype ...and now for the final stroke: imagine how you'd describe portage and the /etc/ and /var/ and /dev/ dirs?!? ...imagine litterary trying to "emerge my favourite office suite, you know the one with the little star ...i think it's abbrevated OOo ...no idea why" ...well, i'm not a person who'd enjoy such a "fs" ...but i'm not conviencing anyone of not trying ...i'm just saying i wont _________________ tea+free software+law=hook
(deep inside i'm still a tux's little helper) |
|
Back to top |
|
|
lurid Guru
Joined: 12 Mar 2003 Posts: 595 Location: Florida
|
Posted: Tue Sep 09, 2003 2:56 pm Post subject: |
|
|
Xiol wrote: | I thought that Storage, and Longhorn's FS, were just laid on top of the traditional filesystem |
It is. I've read the NTFS is still there underlying everything, but that the user won't actually have access to it. The new 'filesystem' if you can call it that, sits on top of NTFS and just organises everything via an SQL database. But as far as I know, you won't be able to open Explorer and click through directories.
robdavies wrote: | Way back, UK mainframe firm had a contents addressable file system, might be same sort of idea. |
Yeah, its actually a very old concept that was designed originally for webservers. Once again, a Microsoft 'innovation' was basically ripped off from someone elses idea. Its not even a recent one.
hook wrote: | i don't like it, but i'm sure some guys like it ...because if there's absolutely noone who would use it, it wouldn't be there IMO. |
Its there because Microsoft put it there. Its going to be the new way to do things in Longhorn regardless of whether people like it or not. It exists as an Open Source project right now because someone probably wanted to get the jump on MS and develop and release this tech first.
Like I said, this is good only for future Windows converts. Once they get used to this type of file system after using Longhorn, they'll expect that ALL operating systems will do the same thing, just as right now they expect all operating systems to play every single game avalible and to have a flashy GUI. As for existing Linux users, I think most of us agree this system has major flaws and should generally prove to be quite annoying to use. _________________ Go find a cheerleader and saw her legs off. - Nny |
|
Back to top |
|
|
krusty_ar Guru
Joined: 03 Oct 2002 Posts: 560 Location: Rosario, Argentina
|
Posted: Tue Sep 09, 2003 4:06 pm Post subject: |
|
|
I've just ovelooked the idea behing Storage, It could be usefull, but if it were my decision, I would put the effort in this that it's truly new and revolutionary, besides being one of the first mayor OSS proyects that tryes to create something new and better instead of just following M$'s bad example. _________________ I am Beta, don't expect correct behaviour from me.
Take part of the adopt an unaswered post initiative |
|
Back to top |
|
|
Unne l33t
Joined: 21 Jul 2003 Posts: 616
|
Posted: Tue Sep 09, 2003 4:08 pm Post subject: |
|
|
I really don't like the idea at all. It might be beneficial for users like my parents who have a hard time grasping the concept of directories and subdirectories and files in directories, etc. But for anyone who's used a computer for any given length of time I think it'd be a huge hassle.
"Searching" for files is one of the biggest pains in the world, to me. Probably why we have so many tools for it (find, locate, etc.), and it's still often a hassle. The fastest way to locate a file is to put it somewhere and remember where it is, or find where it is and remember that location. I don't often remember filenames exactly, but I do remember the locations of files in my directory tree. I don't need the computer to manage the locations for me, and I'm pretty sure the computer would do a poorer job of it than I can do myself. If I was forced to "search" for every file I used, I'd go insane. |
|
Back to top |
|
|
hook Veteran
Joined: 23 Oct 2002 Posts: 1398 Location: Ljubljana, Slovenia
|
Posted: Wed Sep 10, 2003 9:28 am Post subject: |
|
|
actually Unne gave me an idea how this would be useful!!!
imagine having a completely normal linux system, which hosts a few complete idiots. you could make a /home2 (or /home_idiots whatever you like) and link a storage pratition to it ...this way the admin and all the normal users would have the normal fs, and the idiot users would use the storage on /home2 (or /home_idiots) _________________ tea+free software+law=hook
(deep inside i'm still a tux's little helper) |
|
Back to top |
|
|
() l33t
Joined: 25 Nov 2002 Posts: 610
|
Posted: Wed Sep 10, 2003 1:31 pm Post subject: |
|
|
[irony]While you're at it maybe you could provide a separate boot with automated installation of packages, for those morons who can't compile programs themselves.[/irony]
There's noone saying that Storage won't provide a way of structuring files similar to hierarchical filesystems, at least if one is to believe the HCI paper: 'Planned augmentation with a category refinement system may produce a hybrid that addresses both object reference and exploration'. |
|
Back to top |
|
|
Cossins Veteran
Joined: 21 Mar 2003 Posts: 1136 Location: Copenhagen, Denmark
|
Posted: Wed Sep 10, 2003 4:42 pm Post subject: |
|
|
The idea of Storage (or the Longhorn FS, for that matter) is not to have a whole new kind of filesystem - it is to put layer in between the user and the filesystem which takes care of file destinations, names, permissions, and stuff like that so the user won't have to worry.
I imagine having a button called "Music", clicking on which would bring up a window listing all music files on the harddrive, independant of file types and such. It could then be sorted in subfolders by title, author, release date, etc.
Maybe we will need a new term for folder, maybe a whole new interface for navigating them (making it easier, mind you).
I see a lot of potential in this idea, I hope it develops into something useful...
- Simon |
|
Back to top |
|
|
RedBeard0531 Guru
Joined: 21 Sep 2002 Posts: 415 Location: maryland
|
Posted: Wed Sep 10, 2003 9:48 pm Post subject: |
|
|
doesnt reiser4 do something like this. there ia a layer of abstarction between the users and the HARDCORE filesystem structure. It uses a plugin system to allow you to view your files how ever you want.
[sarcasm]
I think that people should access files by using the CHS/LBA values for them. I want coplete control over where my files go. anything else is for noobs
[/sarcasm]
No offence, but i think that several of you are judging something before trying it, or even becomeing well informed. just b/c MS is doing something doesnt make it bad. Gates above all is a business man, I doubt he would waste millions on the r&d to develope a new fs if there was no advantages.
the following is purly conjucture, please feel free to skip it and go to the next post
My gess is that storage wouldnt effect most dirs, such as /dev or /bin. It might provide a frontend (while leaving the backend available) to dirs like /home and /etc. Imagine typing /config/sys/fluxbox/menu to access the systemwide flux menu. Mabey to would start with a differant char than / to signify the diffeances. You could even do something like /media/music/linkin\ park/meteora/Faint.
mabey a /media/seach dir that you could "mkdir faint" to create a dir containing all the media with the word faint. This sorta exists now. take a look at lufs ant the gnetfs. it does this from the gnutella network. it is proof of concept now, tho.
I appologize for my atrcios speling [/code] _________________ OH MY GOD! Kenny just killed Kenny!
That Basterd! |
|
Back to top |
|
|
ebrostig Bodhisattva
Joined: 20 Jul 2002 Posts: 3152 Location: Orlando, Fl
|
Posted: Wed Sep 10, 2003 10:14 pm Post subject: |
|
|
Nice to see that others are following in Oracle's footsteps
Oracle has had it's iFS or internet File System for quite a few years now. To the regular user it looks like a normal disk or directory under *NIX. All of the contents are stored physically in an Oracle database and the contents can be indexed and searched using any of Oracle Content managment like Spatial, Video, Image or Text.
See Oracle Content Managment
The only culprit is that this is a huge task for a database and requires a lot of resources and requires a database that is not yet available as open source in order to function properly.
Erik _________________ 'Yes, Firefox is indeed greater than women. Can women block pops up for you? No. Can Firefox show you naked women? Yes.' |
|
Back to top |
|
|
telex4 l33t
Joined: 21 Sep 2002 Posts: 704 Location: Reading, UK
|
Posted: Wed Sep 10, 2003 10:41 pm Post subject: |
|
|
One thing I don't get; maybe some can answer this...
Storage uses a PostgreSQL database for the filesystem, but also keeps the traditional filesystem heirachy. Does this mean that you do away with reiser, ext etc. and have storage, and then have two "frontends" to it... one that is emulating the traditiona filesystem, and the other that provides a meta-information based system? |
|
Back to top |
|
|
ebrostig Bodhisattva
Joined: 20 Jul 2002 Posts: 3152 Location: Orlando, Fl
|
Posted: Wed Sep 10, 2003 11:20 pm Post subject: |
|
|
telex4 wrote: | One thing I don't get; maybe some can answer this...
Storage uses a PostgreSQL database for the filesystem, but also keeps the traditional filesystem heirachy. Does this mean that you do away with reiser, ext etc. and have storage, and then have two "frontends" to it... one that is emulating the traditiona filesystem, and the other that provides a meta-information based system? |
Well, I can only comment from the one I know, Oracle iFS.
Basically it is a filesystem driver that is a frontend for the database. The files are stored in the database and along with the actual files, we also store metadata, i.e directory and hierarchy information. Depending on the type of file stored, the information is stored in different tables in order to ease the index process. Check the link I gave above for more information on how it is done.
The local filesystem driver, which is actually a database client, is then responsible for presenting the data to the user in a normal directory fashion.
Erik _________________ 'Yes, Firefox is indeed greater than women. Can women block pops up for you? No. Can Firefox show you naked women? Yes.' |
|
Back to top |
|
|
tdb Apprentice
Joined: 19 Sep 2002 Posts: 293 Location: New Orleans, Louisiana, U.S.A. (what's left of it anyway...)
|
Posted: Fri Sep 12, 2003 5:22 am Post subject: |
|
|
This whole thing reminds me of Evolution's "Virtual Folders" feature. |
|
Back to top |
|
|
corrs_fan Tux's lil' helper
Joined: 16 Sep 2002 Posts: 78 Location: Giffnock, East Renfrewshire
|
Posted: Fri Sep 12, 2003 2:43 pm Post subject: |
|
|
i just had another thought about this too. What the heck kind of CPU will you need for this
think about it, it has to (re)reference everything all the time, when you create, ammend and delete the "content", the backend must have to monitor this and update itself, which surley means more means more memory usage too. that athlon-64 is looking nicer by the day. _________________ Some say "The glass is half empty",
I usually say "Eh, There was a Glass.." |
|
Back to top |
|
|
paperwings Tux's lil' helper
Joined: 14 Jan 2003 Posts: 137 Location: Boston, MA
|
Posted: Fri Sep 12, 2003 5:29 pm Post subject: |
|
|
I don't really understand this push for a new type of filesystem anyway. What's wrong with the hierarchical type? I've heard people going one about time-rather-than-filename based systems, etc. They say it is more natural to think of the document you worked on yesterday rather than the document named "x.xxx". Whatever. I say moose muffins. Seems to me that the hierarchical system is more natural. After all, before computers, we filed things in cabinets by name.
What's that sound? It's the sound of these new filesystems flopping when no one uses them... :wq
P.S. it is not necessary to type :wq when finished with post. I will never learn... |
|
Back to top |
|
|
ebrostig Bodhisattva
Joined: 20 Jul 2002 Posts: 3152 Location: Orlando, Fl
|
Posted: Fri Sep 12, 2003 10:45 pm Post subject: |
|
|
paperwings wrote: | I don't really understand this push for a new type of filesystem anyway. What's wrong with the hierarchical type? I've heard people going one about time-rather-than-filename based systems, etc. They say it is more natural to think of the document you worked on yesterday rather than the document named "x.xxx". Whatever. I say moose muffins. Seems to me that the hierarchical system is more natural. After all, before computers, we filed things in cabinets by name.
What's that sound? It's the sound of these new filesystems flopping when no one uses them... :wq
P.S. it is not necessary to type :wq when finished with post. I will never learn... |
I actually tend to agree with you, mostly
There are situations where a database based filesystem is good. Mainly in situations where you then easily can index all of your documents and perform free-text based searches using complex queries. Of course, most single user PC's will not have any need for this, but in a large collaboration between maybe hundreds of people, it is nice to be able to locate all the documents which contains X, Y and Z, but Not A, B and (C OR D) minus E. You can't do that easily in a hierarchical system.
I guess it all boils down to what your bottlenecks are and what your needs really are.
Erik _________________ 'Yes, Firefox is indeed greater than women. Can women block pops up for you? No. Can Firefox show you naked women? Yes.' |
|
Back to top |
|
|
Tasslehoff Tux's lil' helper
Joined: 09 Jul 2003 Posts: 100 Location: BC Canada
|
Posted: Fri Sep 12, 2003 11:42 pm Post subject: |
|
|
What I wonder about is added overhead. How much will there be? I assume there should be some, unless of course the search for the files will be more efficiant thereby decreasing the overhead. |
|
Back to top |
|
|
ebrostig Bodhisattva
Joined: 20 Jul 2002 Posts: 3152 Location: Orlando, Fl
|
Posted: Sat Sep 13, 2003 1:01 am Post subject: |
|
|
Tasslehoff wrote: | What I wonder about is added overhead. How much will there be? I assume there should be some, unless of course the search for the files will be more efficiant thereby decreasing the overhead. |
I hoenstly don't know how much overhead it would be on a standalone, but if you try to run Oracle iFS on a server with less than 1GB of RAM, forget it. The larger the filesystem is, i.e number of files to be stored and indexed the more resources is needed. You also have to have a pretty fast CPU(s) in order to keep the indexes up-to-date.
I seriously doubt this will have much use on standalone PC's unless there are some serious advancement in indexing and storage technology. Maybe when we have 400Ghz CPU's and terrabyte ram sticks it will be efficient on single PC's
Erik _________________ 'Yes, Firefox is indeed greater than women. Can women block pops up for you? No. Can Firefox show you naked women? Yes.' |
|
Back to top |
|
|
watersb Apprentice
Joined: 04 Sep 2002 Posts: 297 Location: take a left turn in Tesuque
|
Posted: Wed Sep 24, 2003 2:54 pm Post subject: |
|
|
I am participating in a project to support large data archives of astronomical data.
So far, a simple hierarchy (mapped directly onto a traditional filesystem) serves as a structured storage repository for the raw visibility data, which is several terabytes. We use a simple meta-data index to locate the correct data files and serve them up to users. So far.
In the near future, we will need the ability to model the visibility data as if it were a monolithic data source, from which we may serve sub-images: a virtual sky, with image cut-out service. Jim Gray and Alex Szalay have done this already for optical surveys, and Jim has done this for Earth-observing imagry.
It is a hard problem. Simple and hard. Not-so-simple when you consider the manipulations needed to normalize the data... anyway...
I have experimented with PostgreSQL and Oracle and SQLServer, adding spatial indexes to them; this works OK.
But most science data archives look a lot like filesystems, the users expect this, the enormous amount of legacy data-reduction software expects this.
There are efforts to model the thing exactly like iFS or WebDAV. But I suspect for a specialized data repository like this, we might be able to use something like Reiser4.
My point is that for some applications, data-storage of this nature might be useful. It still isn't clear. Not clear that it isn't a bad idea. Might be a way to simplify.
IBM mainframes have been doing this for a long time... |
|
Back to top |
|
|
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
|