Slashdot Mirror
Serial ATA Coming
John Doe writes "Heatseekerz.net Has a new article dedicated to Serial ATA @ Cebit 2002. This technology will be here sooner then you think!" The article is a little thin, but I haven't heard
a lot about what looks to be a very common standard in the not so distant future.
there's already a high speed serial that can be used for ide drives. its called usb2 and also firewire.
I am using an external drive bay that takes FW in and converts (with a very small pcb) to 40pin ide (ata100). cost isn't much ($70) and the controller isn't either ($30).
I was able to copy an 80gig drive from native ide to a remote ide via firewire on the latest linux 2.4.18 kernel in about 3 hrs or less.
serial ide would probably JUST be ide. but serial usb2 and FW are more general purpose (video, etc).
I think serial ide is just too late in the market.
--
"It is now safe to switch off your computer."
That's a matter of opinion. Remember, this requires new hard drives - something that doesn't exactly happen every day in big business. You're talking new hard drive duplicators, external hard drive enclosures, etc. This is like saying fibre channel hard drives are available today - well, sure they are, but they aren't getting big play in your typical home or business.
What's your damage, Heather?
There are a number of issues that it seems that SerialATA doesn't address that it should:
1) Power to the device is still separated from the data connection.
2) Because it is backwards compatible with regular ATA it appears it will have the same limitations on the number of devices you can connect, i.e. 2 per channel.
3) It is unusable for external devices.
Why upgrade to a standard whose only advantage is a speed increase we don't need and smaller cables that can be done with parallel ATA ala "round" IDE cables? Seems like a huge investment that would be better made in FireWire 2.0 or something similar so that you can use the same interface internally and externally, with power provided, and have many devices on the same bus.
"If I can see farther it is because I am surrounded by dwarves." -- Murray Gell-Mann
I thought that Serial ATA 1 would not support daisy chain. As far as I know you need as many Serial ATA ports on the motherboard/controller as drives you want to use.
For more information go to here
As far as the name "Serial ATA," it's a smart move. It will create the impression in people's mind that it's an "extention" or "enhancement" of standard ATA, without necessarily being backwards compatible at all. But, hey, once it gains market share, and the SATA drives start filling the shelves at Best Buy, it won't really matter.
dinner: it's what's for beer
Third, Serial ATA--unlike SCSI--doesn't require you to load device drivers out of the wazoo to support devices on the bus. The only driver you'll probably need is the driver for the motherboard chipset that incorporates Serial ATA support. this is an OS design issue. you don't have to do this with Linux. there is a single SCSI driver, based on the identity of your SCSI controller. All other SCSI devices attached to the bus are accessed using this driver. this has never really been true under Windows or MacOS, but it has nothing to do with SCSI itself, just the rather silly way developers of and for those platforms have gone about creating the driver architecture.
The reason that there is a movement from parallel standards towards serial standards is not because serial transfer is faster, in most cases it is not, a parallel solution will in most cases always be faster (unless the time it takes to de-paralize the data is to high), the problem with paralell transfer at hig speed is isolation.
When you transfer data at high speed with cheep cables like the normal IDE once, the signals in the different cables tend to polute each other.
Three different solution exists to this problem:
1. We could build better cables (like the ones you usualy have with Ultra2/Ultra160/320 scsi interface).
2. We could send the data with cheap cables but with a better error correction. (like is done with ATA66/100/133)
3. We could develop a serial interface.(Like Serial ATA)
Wasn't IBM developing serial standard decade ago? Whatever happened to that? (I think it was called SSA or Fibre-Channel)
Questions not answered by the FAQ:
"The invisible and the non-existent look very much alike." -- Delos B. McKown
Parallel ATA cannot scale to support several more speed doublings, and is nearing its
performance capacity. By contrast, Serial ATA's roadmap starts at 1.5 gigabits per second
(equivalent to a data rate of 150 MB/s) and migrates to 3.0 gigabits per second (300 MB/s), then
to 6.0 gigabits per second (600 MB/s). This roadmap supports up to 10 years of storage
evolution, based on historical trends.
No the serial port on the C64 were bit bangers. Just load up the buffer and toggle the bits thru. The drives had their own 6502 and memory, they set waiting for their ID to appear on the bus and then answered back to start the read/write. They worked pretty well for what they were. The SCSI drive has an adapter with a 6502 and memroy just like the standard 1571 drive, but with a custom driver that translates the serial bits to the SCSI bus. Works very well and is about 10 times faster than the 1571 drive.
Why SerialATA and not USB 2.0 or Firewire? Two big reasons: cost and design.
Firewire and USB are not designed for what SCSI and SerialATA are. If you pull a failed drive off a firewire chain and plug a new one in it's place the new drive will have a different ID than the old one did. Firewire and USB are loose protocols that are designed for plugging in optional components not for critical drives and drive arrays. With Firewire and USB the ID goes with the device not with the connector. With SCA SCSI and SerialATA it is the other way around.
As far as cost, those bridge chips between USB/Firewire and IDE are very expensive. IDE is the cheapest drive technology by far and from what the industry is saying SerialATA will be as cheap or cheaper. The chips on the motherboards and drives will have less pins. The connectors will take less real estate on the motherboard, the cables will have fewer wires. Because it's new technology it will probably carry a price premium for a little while but in the end, this stuff will be really cheap.
I am a big fan of IDE RAID. It's price/performance is absolutely amazing compared to every other RAID option. The problem is you can never get more than 8 drives attached to a controller. 8 Drives is nice but a standard 3 channel LVD/SE scsi controller can handle 45. Even a 1 channel can take 15 drives. SerialATA will open new doors for larger IDE RAID arrays, with smaller connectors that allow more drives per controller and longer cable lengths to allow more drives per system. A standard hot plug connector will also come in handy when building compact hot swap enclosures. I expect to see SerialATA RAID controllers supporting up to 16 drives and the ability to put 4 in one machine for a massive storage server (10+TB). This will be a great help to IDE RAID.
In short, I think SerialATA is the only hard drive interface technology that has a bright future. Fiber channel is just too expensive and SCSI, while nice, is losing it's advantages over IDE. DMA mode is now standard on all modern IDE chipsets nearly eliminating the performance differences between SCSI and IDE. Now with SerialATA the cabling on IDE is much more graceful than before. While SCSI is nice for it's ability to daisy chain, that's really a bad idea for production systems. Point to point is much more reliable and much easier to troubleshoot.
Fast, cheap, flexible, huge industry support. Sounds like winning technology to me.
set softtabstop=4 shiftwidth=4 expandtab nocp worlddomination