Hi Jimxugle,
 
What is HUGE server? It would be helpful if you said what capacity of HD storage you require. I have one server up with 38 Terabytes of HD storage.
 
You seem a little stuck on the DOS-like drive lettering. In theory you could have a virtually unlimited number of block devices (Hard Disk is such a device, typically on an IDE interface). Then, you have the SCSI devices ... /dev/scda etc.
 
However, you have not said what the server is for. Just because you can attach a virtually unlimited number of disks is not of consequence when you don't say how the storage is used. If you will say what the server is for perhaps people will be able to assist you better.
 
HTH
Bill

Hey.. thanks for writing!
 
I'm thinking of making a system where I insert a DVD into the DVD drive, a program automatically copeis the entire Disk into a .iso file and store it on the hard drive(s) for mounting/watching later.
 
I have no expiriance with SCSI Devices.
 
I was thinking of Using one motherbord with 6 PCI channels. In each of those channels will (theoretically) be a PCI 1-to-15 hub (yes, they make them and I'll try to post a link when I find the URL). On Each of the 90 "cloned" PCI channels, I plan to have a 5 port USB card. Conenected to each port is a Hard Drive enclosure containing the highest-capacity Drive I can find (Hitachi's proposed 1TB Drive. Bringing the Total amount of space to 460800 GB. (assuming each byte is 6 bits, each kilobyte is 1024 bytes, each megabyte is 1024 KB, etc., etc., etc.)
 
PlzkTHX!!!

howdy Jimxugle
 
that's some interesting theory you're pointing out there with your "multiply cloned PCI-controllers and USB hubs" machine. Alas, as fine as teh results of your computations may sound, there are some limitations:
 
1) Total capcity vs. partition capacity
As BSchindler pointed out correctly there is virtually no limit for the number of block devices you run under an *x-operating sys. Still though, unless you "combine" the individual drives to some big virtual drive ("logical volume", keyword: RAID), you will just have a flock of drives where each partition on each drive needs its dedicated mount point in the *X-file system. Also, the total capacity might be extreme, but the maximum disk space you can access "in one block" will be the size of the largest partition of the largest drive.
 
Not that an average disk size of -let's say- 250GB is to be considered "small", the problem will be to determine WHERE (on which drive/part.) to store a copied DVD. And for that purpose a check of all mounted drives/part. will be necessary (sooner or later).
 
The solution to this would be to use a RAID-thingie that lets you combine multiple drives to one logical storage unit. I'm almost 100% sure that setting up a somewhat usable RAID set with ~450 harddisks on 90 cloned PCI channels, on 5-port USB hubs with take you around 200 years. Don't worry about that timespan, as the mere attempt to undergo such an adventure will make you immortal amongst the *X-community
 
2) PCI capacity
I'm not even mentioning the overhead of data that it would take to control 450 drives, but I could imagine that just the "start/stop disk-unit"-commands alone will exceed the bandwidth of the standard PCI as well as the PCIe/x interface. Keep in mind that harddisks are extremely "chitty chatty" with their host controllers, and 450 drives can only be compared to something really, really "beyond" like 18 wimmin at a cafe, talking about their hubbies
 
Well, and apart from the above mentioned "management data" there is still "user data" to shoot through the PCI channels. Even if you would use PCI-X 2.0 with a (theoretical) bandwidth of up to 5GB/s you would still get a super slow storage system as the drives would outmaneuver themselves permanently by requesting and occupying the PCI bus. Also not to mention: the PCI arbiter-chip will probably go nuts and start to organize with other controller chips in unions
 
3) Failure tolerance
Normally they say "the more the merrier". Unfortunately this doesn't apply to harddisks. With the proposed 450 disks chances are good that you will probably face around 5-20 defect disk per year. Even if you stick to a much lower number of disks: the chances for a complete and total loss of a disk are superlative (even 40 disks mounted as individual drives will make you go crazy).
 
Solutions
The only thing really helpful would be to use ATA/SATA-, or SCSI-RAID storage systems, either rack mounted or server-cases). These systems easily go up to a couple of Terabytes (we run a Transtec 4 TB solution here for all of the companies backup purposes), they integrate smoothly into almost any kind of environment as they come with their own operating system that offers the logical volume to Windows, Linux, Unix, Mac OSX etc.
 
The data transfer from and to the system is done via dual SCSI lanes, and it has the advantage of being "hot pluggable" (disk kapoot? get it out and stuff in a new one while the system is up).
 
Putting it all together, my advice would be to peek into those dedicated storage systems, stack a few of them and thusly obtain the TBs you aim at.
 
hope this littany helps

thx blackpage! I thought that there might be bandwidth issues...
I'm now scrapping the idea of using PCI cloing.
 
Just General FYI: I like IDE and SATA drives... they're more fammiliar.
 
Quick question. (dumb probably)
 
If I have one RAID 0 Controller with 4 drives on it, Is it possible to have a second RAID 0 controller with another 4 drives on it and have the system recognise all 8 Drives as one Disk? Does this work with more than 2 RAID Conrtollers?
 
I'm starting to work on the actual software... anyone know of an open-source DVD-to-ISO command line software? right now I'v only been able to find some *gahsp!* Closed source products. :'(

Update on my software:
 
heh. The softwere I pulled of the net (not warez, mind you) didnt act the way I thought... I now need somesort of command-line utility that will copy every bit off of a DVD into an ISO file.
 
any suggestions?

gidday Jimxugle
 
re: clustering RAID controllers
 
To give it away quite frankly: I have no idea if something like could possibly work. "Feeling"-wise I'd say "no". When it comes down to logical volumes you always need some kind of "supervising" instance. In most cases this is the RAID controller which controls the attached drives.
 
In your suggested setup several drives would be attached to independant RAID controllers. And to make the whole logical volume accessible as one large partition, the 2 RAID controllers would also need some "supervising" instance.
 
The only thing that could possibly handle your setup is a "single-card-multiple-RAID-channels"-controller. LSI e.g. produces RAID controllers with 4 independant RAID channels. So maybe you'd also like to peek into those products.
 
As it goes for the drive types: The storage system I've mentioned in my last post runs with IDE drives. Just the data-connection to the server uses the SCSI protocol. As far as I know these Transtec thingies come in all flavors, supporting almost any kind of drives (not sure about PATA).
 
hope that helps

hey... thanks! I'm gonna have to go re-think everything with this.
 
Pitch to RAID controller manufacturers: Configuration-free transparent RAID system... Connect two hard disks to a small device that will make both disks appear to be one. It will be able to Daisy-chain these into multi-terabyte virtual drives