Slashdot Mirror
Squeezing 160G on to ATA Motherboards
MadCow-ard asks: "With the introduction of the new 160 GB hard drives there comes a problem: they only appear to work with the ATA/ATAPI-6, 48 bit-standard. This means not installing them into systems that I have already built with the de facto 28 bit ATA controllers. I build video editing systems that easily reach 800 Mb, and so the Promise solution with a 2 hard drive ATA controller card doesn't really help. Is there a way squeeze these onto my systems without dropping everything above 137.4 Gb?" 160 gigs on a single HD! How soon before terabyte drives become a reality?
It's not just a hardware issue. Which OS or OS's do you want to support? A solution that works for Linux may not work on some flavor of BSD. You might even be stuck with one of the dozens of lesser OSs.
Sounds like for the systems you're building, SCSI would be a better choice. Not only will you avoid the sort of problems you're describing, but performance will be far better, especially since you're connecting multiple drives.
Have you tried 3ware? They make IDE RAID cards that have linux driver support (in the 2.4 kernel). I'm not sure if their devices support the new 48bit LBA standard. They seem to be focusing more on their larger products but their RAID cards (which are used in their larger products so they shouldn't be going away any time soon) are here.
Promise has the FastTrak100 TX4 PCI that supposedly has four independent IDE channels (no slave/master crap, everything is master like 3ware products) so you have another option there with support for 48bit LBA in Promise drivers mentioned at linux-ide.org it sounds like a promising solution (no pun intended).
You could always put a couple Promise Ultra100's in there too - it sucks to waste PCI slots but with high end motherboards having onboard LAN, sound, etc I would expect that you have plenty of open slots. I've used both Promise Ultra/FastTrack products (with the kernel drivers, not Promises) and 3ware products and both are great.
From front page of linux-ide.org:
Leading the World to Announce Native 48bit LBA Support
Supporting Maxtor BIG DRIVE TECHNOLOGY
Releasing Support of new Promise Ultra 133 TX2 48bit HOST
Future Release Support of new Silicon Image's CMD 48bit HOST
I just bought the 6800 for my FreeBSD box; I'm gonna shove about a half-dozen 100 gig drives in it so I can archive all my live concert recordings in SHN format. The cool thing about these cards is each drive gets its own dedicated controller and they're *real* RAID, none of this fake-ass Promise half-software stuff. I won't have to bother with vinum either.
The question I have for the original poster is: why bother with 160 gig drives when 100s are cheaper, and a bunch of them (or even a few striped pairs) will be a lot faster than a few, much-more-expensive 160 gig drives?
"Remember when the U.S. had a drug problem, and then we declared a War On Drugs, and now you can't buy drugs anymore?"
"with high end motherboards having onboard LAN, sound, etc..."
Maybe it's just me, but I've always thought that the integrated boards were the low-end, since the onboard components are invariably of a lesser quality than what you can purchase and install seperately.
This sig is xenon coated, and will glow red when in the presence of aliens
My problem with almost all the hardware IDE RAID right now is they work in software. They don't do the actual RAID in the hardware so CPU use takes a huge hit.
:)
Maybe that's why they are so cheap compared to SCSI RAID.
SCSI offers more devices but can it do more primary partitions per drive?
Maybe it's time to replace the whole ISA-PCI-IDE started out as an 8-bit platform and got patched and kludged time and time again mess with something that anticipates that what now seem like big drives, big RAM sticks, and fast processors and video cards will soon be classed with 8088s and 64K ram chips. And maybe there's something better than x86, or could be.
And while I'm ranting, how about we *don't* go through another episode of incompatible form factors for motherboards, cases, and power supplies (not to mention memory) that make brand name boxes un-upgradeable except by pitching them into the landfill and buying a whole new system.
I see even classic Slashdot is now pretty much unusable on dial up anymore.
Without some scheme to lie to half of your hardware you can't have more than 4 primary partitions per drive no matter what size it is (or 3 primary and only 1 extended),
;-)
...
I seem to recall an article a while back talking about how the 64-bit version of Windows XP has a new partitioning system - GPT ("GUID Partition Table") - which is meant to sort out current problems with partition tables.
And don't worry, 64-bit Linux supports it too
MS has a document explaining their 64-bit things, including GPT and the associated support stuff -- Designing for 64-bit Windows. Things appear to be changing a fair bit -- most software will break on the new hardware they describe, but it should simplify what's left
SCSI is better than IDE, but it is really aging technology as well. FireWire is altogether more convenient, hot pluggable, and probably cheaper as well.
The partitioning has nothing to do with hardware. The same 4 primary limitation exists even on SCSI. The problem stems from the choice of MS-DOS type of partitioning. Other partition schemes exist, such as BSD partitions. And Intel is developing a new one for 64-bit architectures that should still work fine on 32-bit machines. It's just a matter of coding in the support portably in the kernel.
You can make an infinite number of logical partitions in extended space, and Linux will support 59 of them (after the 4 primary). You can also stuff the primary partitions with FreeBSD style partitions for a total of 28, or OpenBSD style for a total of 60 (but only 59 devices available).
The whole IRQ system is itself part of the problem in PC design. It is actually a hack done in the design of the original IBM PC in order to delete a real I/O interrupt controller to cut costs. They used the interrupt request LEVELS for separate devices instead of the correct way of having a controller that stored the I/O address of the device generating the interrupt. Mainframes have since the 1960's worked on the latter mechanism, having a single interrupt vector for all I/O and the first thing the handler does after saving context is get the address of the interrupting device. There was never an IRQ problem on mainframes. It is the IRQ design in the PC that needs to be tossed out. Interrupts should always idenfity the precise I/O device.
Incompatibilities will continue to plague the computing industry because designers are limited by cost controls, short sighted planners, having to avoid patents held by other companies, having to push patents held by their employer, etc. Makers of whole systems don't want you to upgrade when they can sell you a whole new system every year. That's how business works. They are not there to make great technology; they are there to sell you stuff. It will include just enough technology to get you to buy it, and no more.
now we need to go OSS in diesel cars
Maybe it's time to replace the whole ISA-PCI-IDE started out as an 8-bit platform and got patched and kludged time and time again mess with something that anticipates that what now seem like big drives, big RAM sticks, and fast processors and video cards will soon be classed with 8088s and 64K ram chips.
:) You've got to pick, keep the crap, or dump it and everything related to it. But please, don't toss your old computers into a landfill. Give them to me!
Say what you like about IA-64, at least Intel does seem intested in killing off all that crap fairly soon. I don't see why you group PCI in with ISA and IDE, however. PCI is a "real" bus, and while it's not perfect, it's high-performance enough for most things, and it's reasonably well designed. I imagine that 3GIO is going to be, for the most part, an extension of PCI to higher clock speeds, rather than completly reworking it (I haven't been able to find any actual info about anywhere, so I guess I could be wrong). Also going out over the next few years are PS/2 [I hope I can find a PS/2 -> USB adapter; I love my keyboard and want to keep using it], floppy drives, serial ports [which is rather unfortunate, lots of old Unix boxen can or have to use the serial port to get an install going], and parellel IDE, replaced by Serial ATA. It's kind of kludgy internally, but I guess from the user's perspective it will be a lot nicer. Similiar to firewire drives, but >4x faster.
un-upgradeable except by pitching them into the landfill and buying a whole new system.
It's funny that you seem to dislike the bad old standards so much, yet you also don't want to throw any of it way. I'm confused.
Where do you think I get my stuff? (Hey, I've got a PC-XT and an Apple II I'm still planning to find uses for.)
But seriously, if you have a perfectly good Compaq case and power supply holding a 486 board and you find a good deal on a Gateway Socket 7 (or the other way around brand wise), it's really annoying that one's got the video jack where the other has the keyboard and mouse or vise versa and IBM, Dell, and HP also seem to play the almost but not quite compatible game as well, which means a lot of otherwise usable stuff doesn't make it to the end of it's working life before somebody gets tired of trying to pound square pegs into round holes and sets it aside to gather dust and goes out and spends more money. Even if you find what you need on eBay the addition of shipping costs run the price up to where you might as well buy something new from a local store where you can return it if it breaks.
I probably included PCI because of all the suffering I've done with Plug and Pray and the IRQ shuffle on PCI slots as well.
If they're going to bring out new stuff that's not completely compatible with legacy stuff, then why not make a clean break with something new and completely incompatible designed from the ground up with both eyes on the future instead of one on the future and one on the past? Something that eliminates all the problems we've been fighting since '81 and trys to anticipate and avoid any new aggravations.
By the way, just what kind of old computer junk can you use? (if you weren't kidding)
I see even classic Slashdot is now pretty much unusable on dial up anymore.
I really hadn't planned to send you money via Paypal, carrier pigeon, or any other method, but it is an interesting link and worth checking out by anyone having any dealings with them, including contemplating getting in on their upcoming IPO (although I do think there may be some short term money to be made there).
I see even classic Slashdot is now pretty much unusable on dial up anymore.
But seriously, if you have a perfectly good Compaq case and power supply holding a 486 board and you find a good deal on a Gateway Socket 7 (or the other way around brand wise), it's really annoying that one's got the video jack where the other has the keyboard and mouse or vise versa
Yeah, OEMs do love to play those kinds of games. Too bad, but I guess they like being the single source (and being able to charge amounts which match that status for spare parts).
By the way, just what kind of old computer junk can you use? (if you weren't kidding)
I think this sums it up: link.