December 2004 Entries
“Although Microsoft hopes to ship a test version of WinFS in late 2006, it could be several more years before the revamped storage mechanism finds its way into Windows Server.
'WinFS in not in the Longhorn client,' he said in an interview. 'It is also not in Longhorn Server.'
It is not even clear if Microsoft will include it with the Longhorn update that is scheduled to follow a couple of years later.”
from and article on C|Net's News.Com
It's really most unfortunate that we won't be getting a new Windows file system anytime soon, but there are a few alternatives worth looking into.
As sizes of modest fileservers gets larger and larger, what can we turn to when NTFS is completely unacceptable in performance and capability. One of them is to use embedded Linux, iSCSI and one of the excellent journalling filesystems that art in the GPL like AFS, XFS, ZFS or Reiser.
If you don't know already:
“Architecturally, the iSCSI driver combines with the client TCP/IP stack, network drivers, and NICs to provide the same functions as a SCSI adapter driver with an HBA.
“The daemon and the kernel driver are available under the terms of the GNU General Public License.”
Why am I even talking about this stuff?
I have a passion for creating personal library systems and that means I need a cheap yet easy to use way to access really large storage. This storage needs to conform to interop standards so I can access it from any OS and not have a huge hassle with security logins, but not just allow anyone with access to the box to get in.
iSCSI is just one way of doing this, Its really amazing that the Linux community is pulling WAYahead in this area. I can make a 4 Drive, 1 Terrabyte fileserver (750mb Raid 5 useable) for under $1000. You can't buy 4 - 250G scsi drives for that right now. For that matter, you can't buy Windows Server 2003 for that with a driveless box to run it either.
I am not sure what to do about this little system publicly but I have some ideas on how to at least help people create some plug and play mass storage for all that stuff we are collecting at home now.
What iSCSI can do that plain old NAS and SAN (without iSCSI) cannot is that it lets you just plug in a storage box to the network and treat it like a local device, I mean Windows can see it as a local drive D: but the performance and reliability is off the charts compared to plugging in a big USB drive. If you want to store all you home movies, digital pictures, etc. in a single safe place that is possible with this scenario. Need more capacity? Just plug in another box.
The main advantage over NAS and SAN is that it can be brought into the local system and not make you go through the hassles of setting up security with Samba for use in Windows.
If using Linux scares you because of the rediculous threats from SCO, then you can do the same thing with NetBSD and it's not contested in any way.
Wasabi Systems is one company providing a way to do exactly what I am talking about here, their hardware is a little restrictive, and so is their policy for becoming an OEM, but it at least a start into the arena. Hopefully there will be some Open Source projects into this area that make sense of reliable storage for everyone to use, not just the geek elite or people who can pay $10K for a file server.
The one technology that has me really excited about something other than .Net is the Cell Architecture from Sony, Toshiba and IBM.
The “big if” is what OS is going to support this new architecture and will it have the development tools needed to really take off. Microsoft is out unless they have decided to bring back the Power PC support for NT (I seriously doubt it...), Apple is in because Mac already runs on the PPC, so does BSD, Linux, and Solaris.
I find it very interesting that IBM has put its PC division up on the auction block at the same time this new architecture is about to be released. Its also interesting that Solaris has just gone Open Source.
Should we read-between-the-lines and gain understanding that this is NOT just a coincidence, but a planned effort to re-invent the PC Industry?
It makes total sense, and if anyone has then balls to throw all their weight behind it, it's these three companies. They have lost all their PC ground to HP and Dell, but a complete re-invention of the industry based on something no one else is anywhere near ready to implement is a major sign that should not be ignored.
I wonder if the fight between Sony and Toshiba over Blueray vs. HD-DVD has any merit when we know they are working together on something this big. I am sure that Blueray can be made to read HD-DVD without much effort, so we will just have the duality like we do today with DVD-R vs. DVD+R.
I have been in the PC Industry for over 20 years and this is not the first challenge to the Wintel domination but it's certainly the one with the most potential if they can truly produce a universal chip that can be used in everything from routers to game consoles to cellphones to PCs to mainframes.
Of course Apple tried it, Sun tried it, and those are the only two left alive. DEC (the alpha chip) was swallowed by Compaq, who was swallowed by HP. Transmeta never could get it together. It seems no one was really prepared to mount a serious challenge and bet the farm on it.
If they are prepared to sell the Cell Chips for something rediculously cheap (like $50 retail) then you could stick 4 to 8 of them in a PC and have parallel processing at home like we have never seen before...
A business could have a mainframe built from 256 or even 1024 of these for a fraction of the cost of a comparable Intel based system. This would set the industry on its ear if they can demonstrate that all the parallel processing is really done at the chip level and programmers don't have to do anything special to break their code into parallelisms.
Writing code for parallel processing is really hard, but Sony is promising that this isn't the case with this new architecture. If I could write code that the compiler took care of breaking up and distributing across multiple processor without me haveing to create threads for everything, split up the processing in code and then re-combine it when its all completed, it would make parallel processing something that just happens, not something we need to architect in software at the ground level.
This is going to introduce major breakthroughs in things like graphics rendering, voice/hardwriting recognition, data mining and theoretical mathematics processing.
This is probably THE most important thing to watch as products begin to arrive over the next two years, just as Longhorn is about to be introduced.
Will Longhorn end up being a BIG YAWN now that they have removed most of the really innovative and cool stuff to make its initial release deadlines? Avalon and XAML are still cool, but they aren't anything all that impressive when you look at what you can already do with SVG and XUL not to even mention the potential for things like Looking Glass. WINFS looked great, but its pulled intil what? 2007? but look at what you get with Solaris 10 today, ZFS... and GFS from Sistina which works on Linux, etc.
One thing is for sure, we are in for some major change over the next 5 years.
The article, “Microsoft commits to XML docs for long term” is some good news.
I am using XML for nearly everything that I used to use INI files, config files, csv files and several other formats for in the past. Printed Documents is something that the XML Format can assit in as well as long as you have some sort of toll (Altova, Word, OpenOffice, etc.) that can read the format.
I recently had to store some text files for a web based mail merging utility. I first went with RTF as the text format of choice so I could stuff the documents into a database and retrieve them over the web dynamically. I quickly ran into several problems that the Office XML Formats can solve quickly. HTML I great for display presentation, but it's awful for printing. Changing internal content then outputting to PDF with Active Reports turned out to be remarkably easy and the result is a portable document generated by my server that is locked for edits, but customized in content. The XML Viewer they have is not too bad for WordProcessingML, but it only works on Windows, I needed complete portability and usually choose PDF for that.
It's nice to see some detailed schemas are getting backed in an open way to share structured information.
Blogroll Me!
Copyright © 2003-2009 H. Steele Price, IV -
All opinions are my own, not necessarily those of my employer, your mother, or any government agency.