> Last time I looked at IDE in any technical
> depth, I only saw four addresses "reserved" for
> IDE controller use. I guess you can have any
> address, but the BIOS couldn't boot off any
> address, it has to know where to look for the
> controller. Predetermined list of 4 seems to
> ring a bell.
There are 4 addresses, but you can only boot off the first 2 in most operating systems. There are ways to get more than 4 up and running to expand to lots of drives, but not sure what OSs it works with.
> Secondly, IDE seems to REALLY hit the breaks
> when you do two independant operations on two
> drives on the same channel (say, a read on
> drive 1 and writer on drive 2).
The issue is that most ATA implementation don't support command queueing, therefore there is no bus release. Each command finishes to completion until the bus is released, while the other drive sits idle. Upcoming drives will be implementing queueing and won't have this performance limitation.
> If my 4 controller addresses educated guess is
> right, and performance does crawl, you'd
> probably want to have 4 drives on 4
> controllers, one each.
The secondary port isn't inherently slower than the primary port. However, each port uses a controller address. (0x178 or something for the first, can't remember offhand)
Best performance is achieved with one drive per cable.
> If all the above is correct, this guy is plain
> wrong. He's published, I'm not, I'm willing to
> admit defeat - where am I wrong? Do the raid
> controllers emulate being scsi hosts, run off
> OS drivers (=likely windows ones), etc?
Everything except ATA hard drives are emulated as SCSI hosts. ATAPI (the CDROM protocol) is simply a packet scsi over an ATA cable. The raid controllers also just use the built-in scsi layer in the OS.
eric
http://www.t13.org for the real ATA specs if you're curious
> They really really need to design a IDE-II
> specification that gives the SCSI performance
> traits to IDE.
They already have it -- tag command queueing has been in the ATA spec for years, since ATA-5 I think. Most vendors either have command queueing IDE drives, or are coming out with them soon.
http://www.t13.org for more info on the various ATA specifications
> He's trying to use software raid, but he has 4
> Promise FastTrak 100TX2 raid controllers. WTF?
> First off, each of those cards supports 4
> drives on 2 channels... Why does he need 4
> cards when he only has 8 drives? He only needs
> 2 cards.
I'm a firmware engineer for Maxtor... if you're going for performance, you want 1 drive on each bus, and you don't want to use the motherboard connectors. With 2 drives on each bus, you are limiting the average transfer rate out of cache to 50% of the max transfer rate. On a modern drive with their 60-65MB/sec channel rates, you cannot stream sequentially off of 2 drives without saturating an ATA-100 cable. Even running ATA-133 won't help starting a year from now.
Additionally, every bios I have looked at sucks in terms of performance. In most cases they have small DMA FIFOs which stutter the pipe during high speed transfers -- they literally hang the DMA lines while they empty their fifo into memory, then come back and grab another 8 words or something sad. They also tend to be very poor managers of the IRQ line. This causes delays at times when your hard drive could be giving you more data, but the host hasn't gotten around to asking for it yet.
All the 3rd party cards have like 2Kbyte FIFOs which prevents any overrun from occurring, which alone is quite helpful in high bandwidth applications.
The cards we include with our drives are in the lower end of Promise's spectrum... you can spend more and get more performance if you want to, which is what I suspect the author of the original article did.
When I was at MIT in the 90s the campus minimum wage was about $7.85 or so. This number is dictated by the need that a 15hour/week need-based work study program cover some fixed percentage of tuition, room, and board.
The AI lab and Media lab were usually paying in the $8.50-9 range, which back in 92 before the boom we thought was a decent salary.
Then of course we took one look at off-campus jobs, as the boom started, and realized we could consult for $50/hr as freshmen to some mom-and-pop business that wanted to connect to that internet thing.
Adrian Carmack: That's it? If you had a million dollars, you'd do two chicks at the same time?
John Romero: Damn straight. I always wanted to do that, man. And I think with my rippin' Ferarri and my wad I could hook that up, cause chicks dig a dude with money.
Carmack: Well, not all chicks.
John Romero: Well the kind of chicks that'd double up on a dude like me do.
Pierce Brosnan - 16 May 1953 (46)
Clint Eastwood - 31 May 1930 (71)
Tom Hanks - 9 July 1956 (45)
Denzel Washington - 28 December 1954 (47)
Bruce Willis - 19 March 1955 (46)
Arnold Schwarzenegger - 30 July 1947 (54)
Robert DeNiro - 17 Aug 1943 (58)
Sean Connery - 25 Aug 1930 (71)
I guess with a name like Metrollica, I guess having only 6 of your 8 facts wrong is pretty good.
non-operational shock characteristics for most drives allow them to absorb a flat-on shock in the 200-400G range. This is when they're dropped flat, because the HSA (head stack assembly) and media flex with nearly the same resonant frequencies, preventing severe damage in most cases.
When the drive is dropped on its end, or worse, on a corner, they can take significantly less of a beating, because the shock can result in misalignment of the platter on the spindle which can be a huge problem with today's data densities. They should still function, but performance will go WAY down because every seek will take an extra rev (11.11ms at 5400) due to misaligned skew.
Operational shock is only a few G's, but if you're dropping in a plugged in and operating computer, you're a moron in the first place.
One note about utilization... in our performance tests here, chipset-based IDE implementations that are built into most motherboards are not as good performance as PCI add-on cards, usually due to smaller FIFO sizes on the DMA bus. Older VIA chipsets were notoriously bad at stalling the bus every 8 or 16 words transferred, while some add-in boards implement several K worth of DMA FIFO.
The ATA/ATAPI-6 specification has support for command queueing, which is the asynchronous component of SCSI. SCSI also implements out-of-order data transfers within a command, which is not part of the ATA spec, however this doesn't help quite as much as command reordering in the queued world.
In the queued-ATA design, the command phase consists of writing all the same task-file registers as before. However, instead of a data transfer phase, an ATA-6 drive has the option to disconnect from the bus and report a 0x40 status instead of 0x50, indicating it is working on a queued command. At this point in time, up to 31 other commands may be issued while the drive is working on the first command.
Once the drive has the data for any of these commands, it then enables the service request bit, at which point the host is expected to issue the service command. The drive, upon receiving a service command, puts the tag that the drive is servicing into the task file and begins data transfer for that command.
To my knowledge, this is pretty similar to how SCSI drives implement this, the difference being that in ATA land the drive must complete the data transfer for a single command while in SCSI land, the drives can disconnect in the middle of a transfer and resume that transfer later after servicing other commands.
Media rates on most drives are in the 50-70MB/s range, so the other poster saying that it only affects performance out of cache is mostly correct. The only difference here at Maxtor for the 133 vs 100 is basically a few timing changes in our ASIC.
I guess what I want in a calculator is a lot like what I want in a computer...
For that reason, I love my 42S that I've been using for 12 years. This is actually my second, first one was stolen back in 1991. Nice simple display, batteries last a long time, good stack manipulation, and moderately programmable. It has even survived two coke spills.
I am a firmware engineer for a large hard drive company, and though I guarantee I know how to make the disk unreadable by these tools, it is impossible to do with any "user" program.
The way I imagine most of these recovery tools work is by reading sideband data off of the drive... When the write head is hauling ass around the platter and you want it to write to a given LBA, it never writes in exactly the same place twice. It might be in slightly different phase with the start of the LBA (5-20ns is common), and since it is a mechanical system, an LBA isn't a perfect arc... it can tend to wobble.
Using in-house diagnostic tools we can "force" the servo code that is supposed to keep the read/write heads centered to a prescribed amount off to the side... If you had an event where the sensitive data was written.1 tracks towards the outer diameter from center, and on a subsequent pass (the 7x overwrite) you wrote your data smack down the center of the track, then it would be possible to position your read head around.3 to.4 tracks towards the OD, crank up the gain in the read channel, and recover that "sideband" data. It would be an absolute pain in the ass, but it is possible. Of course, this setup would probably take roughly 30 mins-2 hours per LBA to calibrate, read, and decode, and on a 100 gig disk that'd take a LONG time...
The price of RAM is still 30x more than hard disk space, but if Moore's law continues, it will eventually be cheaper. I am eagerly looking forward to this day since I believe it result in some incredible changes in OS and Application design. Those of that have been around a while will remember the day when hard disks finally became affordable. Improved access speeds and larger storage size allowed all sorts of new applications to appear. Consider how pervasive SQL databases are, yet we would not have all the applications associated with them without affordable hard disk technology. I think the transition from hard disk based systems to memory based systems will create many opportunities in the high tech world.
Do you think that RAM is the only part of your computer that follows Moore's law? I work for a major HD manufacturer, and the stuff that is for sale now is only a fraction of what we have working in the labs which will go on sale in 4-6 months...
Granted, in 10 years we might not be using hard drives anymore, but for the forseeable future (e.g. within what people forcast in business) the hard drive is here to stay.
RAM is going up on hard drives too as prices drop...
Our business model however has nothing to do with the power user --- in IDE land it is all about reducing cost so e-machines can sell a $399 PC with one of our drives in it. Modern IDE drives are around $0.0025 / MB, which is about 1/10 the cost of the cheapest SDRAM.
Actually, MIT has one of the most comprehensive athletic programs in the entire world, with more inter-school varsity sports than most other colleges. Something like 45 Division III or D3-Club athletic programs.
MIT has a hockey rink that is free for all students with 90 minutes of open-ice hockey and ~4 hours of open-ice skating every day.
A rifle range, pistol range, tons of basketball courts, weight rooms, pool, and tons of fields for baseball, soccer, lacrosse and football round out the facilities.
What really separates MIT in my opinion is that all these facilities are designed with the students and their health in mind, as opposed to a big D1 school that just wants to pack alumni into the stadium each saturday afternoon.
If MIT ever was on ESPN (Our men's hockey team was invited to the D3-Club national championships a few years ago), I would definitely be there cheering for my alma matter.
hrm... wonder if I could get MIT athletics on PPV like I get NHL Center Ice now...
The problem with FDB motors is they generate about 15-18C more heat in normal operation. You better have monster airflow around the thing if you want to keep the sucker cool.
Most drives run around 30-35C in normal operation, and will guarantee they work up to mid-50s. (55C is pretty standard)
Oh, and if your drive *does* overheat, then your FDB motor will start outgassing which will eventually contaminate the media, producing defects and lost data.
They say they use "quiet" seek algorithms... A quiet seek = slow in most cases, since you get quieter by just not accelerating the heads quite so hard.
Like the others, I'll believe it when I don't hear it...
The order is not quite being interpreted the way everyone says it is. According to the legal commentary I heard on NPR this morning on the way to work, part of the reason for throwing out this remedy is that they didn't feel it would prevent Microsoft from exerting its monopolistic strengths in OS and Browser markets even after the company was split in two. They have asked for a new court to come up with a remedy that should actually weaken Microsoft's position with regards to its control of these markets.
Due to industry demands, virtually all drives that we ship are tuned more for decibel level than for performance, because that is what Dell etc are demanding.
High performance = more actuator movement = noisy
Retail drives will be faster than the ones included in pre-built systems, because of this distinction of priorities. I'm sure all drive manufacturers do this too, not just us.
The "real" support for high performance gamers is OpenGL. Stuff like smoke / dynamic lighting are a hundred times faster in OpenGL than DirectX on my GeForce2.
If you read the scores more closely, all the MS-SQL are clustered solutions, while the best single-machine performance is Oracle 8i.
Since throwing more machines at a problem will eventually get you more performance in many cases, it would suggest that MSSQL can't compete per 1 piece of hardware, and that the Oracle folks never bothered to put their software to the test on a "real" cluster.
Though that reason helps, the real reason is to avoid headstack skew which is the #1 technical difficulty in 4+ platter drives. It is also why almost no drive companies make the 4+ platter products anymore.
Slashdot Mirror
> Last time I looked at IDE in any technical
> depth, I only saw four addresses "reserved" for
> IDE controller use. I guess you can have any
> address, but the BIOS couldn't boot off any
> address, it has to know where to look for the
> controller. Predetermined list of 4 seems to
> ring a bell.
There are 4 addresses, but you can only boot off the first 2 in most operating systems. There are ways to get more than 4 up and running to expand to lots of drives, but not sure what OSs it works with.
> Secondly, IDE seems to REALLY hit the breaks
> when you do two independant operations on two
> drives on the same channel (say, a read on
> drive 1 and writer on drive 2).
The issue is that most ATA implementation don't support command queueing, therefore there is no bus release. Each command finishes to completion until the bus is released, while the other drive sits idle. Upcoming drives will be implementing queueing and won't have this performance limitation.
> If my 4 controller addresses educated guess is
> right, and performance does crawl, you'd
> probably want to have 4 drives on 4
> controllers, one each.
The secondary port isn't inherently slower than the primary port. However, each port uses a controller address. (0x178 or something for the first, can't remember offhand)
Best performance is achieved with one drive per cable.
> If all the above is correct, this guy is plain
> wrong. He's published, I'm not, I'm willing to
> admit defeat - where am I wrong? Do the raid
> controllers emulate being scsi hosts, run off
> OS drivers (=likely windows ones), etc?
Everything except ATA hard drives are emulated as SCSI hosts. ATAPI (the CDROM protocol) is simply a packet scsi over an ATA cable. The raid controllers also just use the built-in scsi layer in the OS.
eric
http://www.t13.org for the real ATA specs if you're curious
> They really really need to design a IDE-II
> specification that gives the SCSI performance
> traits to IDE.
They already have it -- tag command queueing has been in the ATA spec for years, since ATA-5 I think. Most vendors either have command queueing IDE drives, or are coming out with them soon.
http://www.t13.org for more info on the various ATA specifications
--eric
> He's trying to use software raid, but he has 4
> Promise FastTrak 100TX2 raid controllers. WTF?
> First off, each of those cards supports 4
> drives on 2 channels... Why does he need 4
> cards when he only has 8 drives? He only needs
> 2 cards.
I'm a firmware engineer for Maxtor... if you're going for performance, you want 1 drive on each bus, and you don't want to use the motherboard connectors. With 2 drives on each bus, you are limiting the average transfer rate out of cache to 50% of the max transfer rate. On a modern drive with their 60-65MB/sec channel rates, you cannot stream sequentially off of 2 drives without saturating an ATA-100 cable. Even running ATA-133 won't help starting a year from now.
Additionally, every bios I have looked at sucks in terms of performance. In most cases they have small DMA FIFOs which stutter the pipe during high speed transfers -- they literally hang the DMA lines while they empty their fifo into memory, then come back and grab another 8 words or something sad. They also tend to be very poor managers of the IRQ line. This causes delays at times when your hard drive could be giving you more data, but the host hasn't gotten around to asking for it yet.
All the 3rd party cards have like 2Kbyte FIFOs which prevents any overrun from occurring, which alone is quite helpful in high bandwidth applications.
The cards we include with our drives are in the lower end of Promise's spectrum... you can spend more and get more performance if you want to, which is what I suspect the author of the original article did.
--eric
When I was at MIT in the 90s the campus minimum wage was about $7.85 or so. This number is dictated by the need that a 15hour/week need-based work study program cover some fixed percentage of tuition, room, and board.
The AI lab and Media lab were usually paying in the $8.50-9 range, which back in 92 before the boom we thought was a decent salary.
Then of course we took one look at off-campus jobs, as the boom started, and realized we could consult for $50/hr as freshmen to some mom-and-pop business that wanted to connect to that internet thing.
eric
Adrian Carmack: That's it? If you had a million dollars, you'd do two chicks at the same time?
John Romero: Damn straight. I always wanted to do that, man. And I think with my rippin' Ferarri and my wad I could hook that up, cause chicks dig a dude with money.
Carmack: Well, not all chicks.
John Romero: Well the kind of chicks that'd double up on a dude like me do.
Courtesy of IMDB...
Actor - Birthdate (age)
Pierce Brosnan - 16 May 1953 (46)
Clint Eastwood - 31 May 1930 (71)
Tom Hanks - 9 July 1956 (45)
Denzel Washington - 28 December 1954 (47)
Bruce Willis - 19 March 1955 (46)
Arnold Schwarzenegger - 30 July 1947 (54)
Robert DeNiro - 17 Aug 1943 (58)
Sean Connery - 25 Aug 1930 (71)
I guess with a name like Metrollica, I guess having only 6 of your 8 facts wrong is pretty good.
My system is quad-ROT13, a $100 upgrade to your double-ROT13
actually no, it is cheaper to have 2 suppliers incase one f's up you can still ship your product
nobody smart goes into volume production with a 1x1 supplier matrix
meant to say MORE, not less of a beating when dropped on the end or corner
non-operational shock characteristics for most drives allow them to absorb a flat-on shock in the 200-400G range. This is when they're dropped flat, because the HSA (head stack assembly) and media flex with nearly the same resonant frequencies, preventing severe damage in most cases.
When the drive is dropped on its end, or worse, on a corner, they can take significantly less of a beating, because the shock can result in misalignment of the platter on the spindle which can be a huge problem with today's data densities. They should still function, but performance will go WAY down because every seek will take an extra rev (11.11ms at 5400) due to misaligned skew.
Operational shock is only a few G's, but if you're dropping in a plugged in and operating computer, you're a moron in the first place.
--e
One note about utilization... in our performance tests here, chipset-based IDE implementations that are built into most motherboards are not as good performance as PCI add-on cards, usually due to smaller FIFO sizes on the DMA bus. Older VIA chipsets were notoriously bad at stalling the bus every 8 or 16 words transferred, while some add-in boards implement several K worth of DMA FIFO.
The ATA/ATAPI-6 specification has support for command queueing, which is the asynchronous component of SCSI. SCSI also implements out-of-order data transfers within a command, which is not part of the ATA spec, however this doesn't help quite as much as command reordering in the queued world.
In the queued-ATA design, the command phase consists of writing all the same task-file registers as before. However, instead of a data transfer phase, an ATA-6 drive has the option to disconnect from the bus and report a 0x40 status instead of 0x50, indicating it is working on a queued command. At this point in time, up to 31 other commands may be issued while the drive is working on the first command.
Once the drive has the data for any of these commands, it then enables the service request bit, at which point the host is expected to issue the service command. The drive, upon receiving a service command, puts the tag that the drive is servicing into the task file and begins data transfer for that command.
To my knowledge, this is pretty similar to how SCSI drives implement this, the difference being that in ATA land the drive must complete the data transfer for a single command while in SCSI land, the drives can disconnect in the middle of a transfer and resume that transfer later after servicing other commands.
Media rates on most drives are in the 50-70MB/s range, so the other poster saying that it only affects performance out of cache is mostly correct. The only difference here at Maxtor for the 133 vs 100 is basically a few timing changes in our ASIC.
I guess what I want in a calculator is a lot like what I want in a computer...
For that reason, I love my 42S that I've been using for 12 years. This is actually my second, first one was stolen back in 1991. Nice simple display, batteries last a long time, good stack manipulation, and moderately programmable. It has even survived two coke spills.
I am a firmware engineer for a large hard drive company, and though I guarantee I know how to make the disk unreadable by these tools, it is impossible to do with any "user" program.
.1 tracks towards the outer diameter from center, and on a subsequent pass (the 7x overwrite) you wrote your data smack down the center of the track, then it would be possible to position your read head around .3 to .4 tracks towards the OD, crank up the gain in the read channel, and recover that "sideband" data. It would be an absolute pain in the ass, but it is possible. Of course, this setup would probably take roughly 30 mins-2 hours per LBA to calibrate, read, and decode, and on a 100 gig disk that'd take a LONG time...
The way I imagine most of these recovery tools work is by reading sideband data off of the drive... When the write head is hauling ass around the platter and you want it to write to a given LBA, it never writes in exactly the same place twice. It might be in slightly different phase with the start of the LBA (5-20ns is common), and since it is a mechanical system, an LBA isn't a perfect arc... it can tend to wobble.
Using in-house diagnostic tools we can "force" the servo code that is supposed to keep the read/write heads centered to a prescribed amount off to the side... If you had an event where the sensitive data was written
--eric
The price of RAM is still 30x more than hard disk space, but if Moore's law continues, it will eventually be cheaper. I am eagerly looking forward to this day since I believe it result in some incredible changes in OS and Application design. Those of that have been around a while will remember the day when hard disks finally became affordable. Improved access speeds and larger storage size allowed all sorts of new applications to appear. Consider how pervasive SQL databases are, yet we would not have all the applications associated with them without affordable hard disk technology. I think the transition from hard disk based systems to memory based systems will create many opportunities in the high tech world.
Do you think that RAM is the only part of your computer that follows Moore's law? I work for a major HD manufacturer, and the stuff that is for sale now is only a fraction of what we have working in the labs which will go on sale in 4-6 months...
Granted, in 10 years we might not be using hard drives anymore, but for the forseeable future (e.g. within what people forcast in business) the hard drive is here to stay.
RAM is going up on hard drives too as prices drop...
Our business model however has nothing to do with the power user --- in IDE land it is all about reducing cost so e-machines can sell a $399 PC with one of our drives in it. Modern IDE drives are around $0.0025 / MB, which is about 1/10 the cost of the cheapest SDRAM.
eric
Actually, MIT has one of the most comprehensive athletic programs in the entire world, with more inter-school varsity sports than most other colleges. Something like 45 Division III or D3-Club athletic programs.
MIT has a hockey rink that is free for all students with 90 minutes of open-ice hockey and ~4 hours of open-ice skating every day.
A rifle range, pistol range, tons of basketball courts, weight rooms, pool, and tons of fields for baseball, soccer, lacrosse and football round out the facilities.
What really separates MIT in my opinion is that all these facilities are designed with the students and their health in mind, as opposed to a big D1 school that just wants to pack alumni into the stadium each saturday afternoon.
If MIT ever was on ESPN (Our men's hockey team was invited to the D3-Club national championships a few years ago), I would definitely be there cheering for my alma matter.hrm... wonder if I could get MIT athletics on PPV like I get NHL Center Ice now...
--eric, MIT alum in debtwhoever modded the Lose Lips post up is a moron...
Any d00d can run nmap and get a pretty decent idea of what a specific machine is running if they're interested in a "serious" attack.
Personally, I would be more inclined to break into the government agencies that didn't reply, knowing they were still using NT.
--eric
The problem with FDB motors is they generate about 15-18C more heat in normal operation. You better have monster airflow around the thing if you want to keep the sucker cool.
Most drives run around 30-35C in normal operation, and will guarantee they work up to mid-50s. (55C is pretty standard)
Oh, and if your drive *does* overheat, then your FDB motor will start outgassing which will eventually contaminate the media, producing defects and lost data.
They say they use "quiet" seek algorithms... A quiet seek = slow in most cases, since you get quieter by just not accelerating the heads quite so hard.
Like the others, I'll believe it when I don't hear it...
The order is not quite being interpreted the way everyone says it is. According to the legal commentary I heard on NPR this morning on the way to work, part of the reason for throwing out this remedy is that they didn't feel it would prevent Microsoft from exerting its monopolistic strengths in OS and Browser markets even after the company was split in two. They have asked for a new court to come up with a remedy that should actually weaken Microsoft's position with regards to its control of these markets.
One note about drives that we ship...
Due to industry demands, virtually all drives that we ship are tuned more for decibel level than for performance, because that is what Dell etc are demanding.
High performance = more actuator movement = noisy
Retail drives will be faster than the ones included in pre-built systems, because of this distinction of priorities. I'm sure all drive manufacturers do this too, not just us.
--eric
The "real" support for high performance gamers is OpenGL. Stuff like smoke / dynamic lighting are a hundred times faster in OpenGL than DirectX on my GeForce2.
If you read the scores more closely, all the MS-SQL are clustered solutions, while the best single-machine performance is Oracle 8i.
Since throwing more machines at a problem will eventually get you more performance in many cases, it would suggest that MSSQL can't compete per 1 piece of hardware, and that the Oracle folks never bothered to put their software to the test on a "real" cluster.
--eric
HAHAHAHAHHAHAHAHAhA
1. wrong -- more sites serve with apache
2. MS innovates the most because they have no regard for standards groups
3. wrong again, MS didn't exist when the internet was being invented
Though that reason helps, the real reason is to avoid headstack skew which is the #1 technical difficulty in 4+ platter drives. It is also why almost no drive companies make the 4+ platter products anymore.
meant to be under 5%, but slashdot ate my plain-text <