... does anybody remember Geoworks? That was a cool piece of software...
GeoWorks Ensemble! AKA PC/GEOS! Now NewDeal. That program has had more names then Prince, but it was still cool. I still miss GeoWrite. Still the best WYSIWYG word processor I've used, to this day.
For those who don't know, PC/GEOS was a multi-tasking, multi-threading, object-oriented GUI operating environment that ran on a PC/XT with 640KB of RAM and 10 MB of hard disk. And it was pretty fast, too. A company is trying to resurrect it as NewDeal Office, but I don't put high hopes on it.
Ansible was talking about an 8-port 100mbit switch, not a hub.
I'm being truly pedantic here, but technically, a hub, a switch, and a repeater are all different things.
Hub is simply a generic term for anything that forms the center of a star topology cabling system. Key phone systems have hubs. Electrical equipment often has hubs. TP Ethernet has hubs.
Repeaters do just that: Repeat an Ethernet transmission (frame) on other ports. Repeaters operate at the electrical level; they don't have any intelligence. Originally used simply to work around cable length limits, multi-port repeaters are required for TP.
Switch is a term for a multi-port bridge. A bridge is a small computer with at least two network interfaces. It receives Ethernet frames, figures out which interface the destination is on, and sends it appropriately (or not at all, if the destination and source interface are the same). If the bridge doesn't know which interface the destination is on, it sends the frame out on all of them (except the source interface).
Now, the breakup of Microsoft will certainly cost the company an immense amount of money as it buys new lodgings for the spinoff company, duplicates support staff, reorganizes, etc. But none of that money will be going to the companies that were harmed by Microsoft's illegal actions. I'd have liked to see them get a few hundred million each.
Any company that feels it was harmed by Microsoft's business practices can sue them for damages. Given the evidence, they may well win.
Microsoft will likely be beset by several hundred rather large house flies in the next few years.
Trust me, there's enough of it already. Start clicking on random things on freshmeat and you'll see what I mean. Yeah, there's lots of good stuff, but OTOH there's a lot of bad stuff too.
Of course. Sturgeon's Law: Ninety percent of everything is crap.
Linux has been above the curve for some time now; I don't think we can expect it to stay there forever. Fortunately, the really good Open Source/Free Software that is out there now will still be out there when the signal-to-noise ratio drops.
First of all, as stated by some others, Windows has already had ATA 100 support.
Which I wasn't disputing.
IBM's latest IDE release, the 75GXP, maxes out at 37MB/s sustained...
Ok, I stand corrected; ATA/66 does have some use then. ATA/100 is still completely worthless to anyone but marketing types, and is likely to remain so for its service lifetime.
SCSI 160 is bogus at this point, ignoring the fact you will need 4 of either of the above mentioned 2 drives to max the throughput.
And I can fly, ignoring the fact that gravity pulls me down.
Point being: That is a pretty big fact to ignore there! Wide SCSI supports up to sixteen devices on the bus at a time, all of which can have operations pending at once (unlike ATA). So, unlike ATA, you can very much use the full potential of SCSI Ultra160. In fact, we could use more bus bandwidth!
The problem is that it requires a 64bit PCI slot which no mainstream consumer level board carries.
That sounds an awful lot like "640K should be enough for anybody!" to me.
First, about five years ago, no mainstream consumer-level board had PCI slots at all. I am sure we will see 64-bit PCI in consumer systems soon enough.
Second, anyone buying four super-fast hard drives is probably not going to be putting them in their 80386 with 4MB of RAM. They will likely have either 66 MHz PCI slots, 64-bit PCI slots, or both.
(For example, one system we have in our integration room right now has four PCI buses with two slots each. Two of those buses are 64-bit. That same system has eight high-speed drives on two Ultra160 SCSI buses using only one IRQ. Do that with ATA!)
32bit PCI which we all have has a theoretical limit of 133MB/s for the whole bus.
First, don't use "all". Speak for yourself.
Second, if 32-bit PCI is such a limit, isn't ATA/100 kind of pushing things? Won't serial ATA be completely pointless?
Hey, let's all go back to Tandy TRS-80s and Apple ][e's, this new stuff is too fast!
The release of USB2 will eliminate almost all of SCSI's external peripherals, except for highend hard drives.
USB2 is still vaporware right now. SCSI Ultra160 you can buy today.
Personally, I think USB is appropriate for devices like scanners, cameras, and the like. I think SCSI is more appropriate for hard drives, tape drives, and other mass storage systems.
Most current gen IDE drives do support command queuing. Whether or not it actually works is anyone's guess...
Another thing to love about ATA: Nobody knows if you're actually getting the performance you should or not!
The reason ATA CDROM drives used to kill system performance was due to a lack of DMA, not because IDE could only access one device per bus at a time. The advent of DMA has elminated this bottleneck.
Then why does ATA CD-ROM multi-tasking performance still blow chunks, when SCSI systems barely even notice the load? Could it be because the whole ATA subsystem has to STOP and wait for the CD-ROM to complete its transaction before that bus can become free again?
ATA sucks, pure and simple. Rather then putting money into trying to extend a dead technology, the industry should just switch to SCSI, which is (1) here today (2) widely used (3) a standard and (4) actually works.
Win2k does indeed support USB. And win2k is essentially a marketing name for NT 5.0, so Windows NT has supported USB for about 6 months now.
Actually, NT 4.0 has had USB support since 1998, I believe. However, Microsoft decided not to release the updates which enabled that support, preferring to use it as a level to force you to upgrade to Windows 2000.
I'm sure the fact that Linux also lacked good USB support had something to do with it.
Got that? Unless they have competition, Microsoft cares nothing about you or your desire to use the technologies you want.
How does ATA/100 compare with SCSI in speed? Reliability? Price?
Reliability
These days, the HDA (Hard Disk Assembly) is usually the same for both SCSI and ATA (IDE) disk drives. They simply use a different PCB (Printed Circuit Board) for different bus interfaces. Thus, all other things being equal, reliability should be about the same.
Of course, all other things are not equal. Generally speaking, very-low-end systems will never see SCSI, and very-high-end systems will never touch ATA. Thus, some really cheap (i.e., unreliable) drives are only available in ATA versions, and some really expensive (i.e., more reliable) drives are only available in SCSI versions.
Price
Despite the fact that there is no practical difference in cost of materials, SCSI drives are almost always much more expensive then their ATA counter-parts. There are two reasons for this.
The first is volume. A lot more ATA drives get sold then SCSI. But the second reason is the more significant one: SCSI is still considered a "high end" technology. Such technologies command a premium price. Thus, manufactures charge more for SCSI drives, and people are willing to pay it.
Performance and Features
ATA does not support device disconnection. This means only one device at a time can be using the bus. Since your average hard drive is going to sustain maybe 15 MByte/sec throughput, if you're lucky, even ATA/33 is overkill. ATA/66 and ATA/100 are completely useless to everyone but marketing types.
SCSI is the clear winner here:
160 MByte/sec maximum transfer rate
Up to 16 devices per bus
External devices
Cable lengths measured in feet, not inches
More kinds of devices (scanners, etc.)
Device disconnection
Multiple buses per controller IRQ
Multiple initiators (controllers) per bus
Tagged command queueing
I've heard people claim that "most users don't need" the features and performance of SCSI. I disagree at two levels: One, I think they do. Even if it is just wondering why all their applications slow down when the CD-ROM drive is busy, regular users encounter the problems of ATA. And second, if regular users don't need that kind of performance, why are they bothering to upgrade ATA at all?
There's a lot of discussion over whether digital or analog is the better medium to use. I want to point out a few things most people are missing.
It is true that, in theory and all other things being equal, analog gives you better reproduction. Digital (by definition) requires you to sample a signal periodically. Changes occur in discrete steps. Analog gives you smooth transitions, as it isn't limited to a particular rate of sampling.
However, digital has other advantages that, IMNSHO, outweigh the advantages of analog in practical use.
First, digital can be reproduced, stored and retransmitted, without limit, without experiencing any signal degradation. Digital signals can also be encoded with redundant data for error correction. Digital media is also generally more resistant to physical degradation from repeated use then analog media (although that is more by accident then through conscious design).
The end result is that while, in theory, an analog signal offers better reproduction, a digital signal will often have better quality, because analog media tends to get worn out quicker and more easily then digital. This is why I like CD over vinyl records; CDs don't pop and hiss like my records used to. This is why I like DVD over VHS; DVDs do not degrade with multiple viewings.
With proper care, you can generally prevent analog systems from degrading in this manner, but neither I, nor your average movie theater teenage projector jocky, treat media that well.
One other thing: The analog purists argue that digital is inferior because digital is sampled. It is interesting to note that motion video of any type is already sampled: What we perceive as motion is really a series of still frames. If a sampled signal is automatically disqualified, then all motion video is disqualified.
what is the technical/financial/legal difference between different long-distance phone companies both providing service to the same area and two cable companies doing the same thing?
You can switch long distance companies because they are providing a connection to distance sites. You still have the same LEC (Local Exchange Carrier), which actually brings the wire to your house. Now, there are (in places) CLECs (Competitive LECs) available, but what they do is essentially buy local lines in bulk from the primary LEC.
The limitation is basically the fact that physical wires need to be strung on poles. You can talk of competition among the phone, power, and cable companies, but there is still only one set of power lines, one set of phone lines, and one TV cable running down your street. That is what the technical limitation is.
The problem with a company like AT&T is that they have a vertical monopoly in particular locales. They don't own more then 30% of the service nationwide, but in the areas they do service, AT&T is the:
- Phone Local Exchange Carrier (runs wires to your house, connects you to your neighbor) - Long-distance phone carrier (connects you to Aunt Marge) - Cable TV wire carrier (runs cable to your house) - Television programming producer (makes TV shows ("content")) - Internet Service Provider (connects to the 'net backbone) - Internet Content Provider
Aunt Marge may not have AT&T's monopoly, but she may have Time-Warner's instead (or whatever).
Reiser FS looks seriously cool. [However] it isn't well integrated.
That's mainly because ReiserFS is still considered to be very experimental (well, by everyone but Hans Reiser), and will not be part of the 2.4 production kernel. The earliest it can hope for adoption is the 2.5/2.6 series. This isn't because of flaws in ReiserFS; it is because of timing. ReiserFS started reaching usable stability too late for it to be integrated into the 2.4 kernel.
I have to wonder, though, what the hold-up is. I'm not trying to troll here, but what's taking them so damn long? They've been working on it since.99 and its still not done?
First, they haven't been working on it since 0.99. That's just plain false.
Second, you have to understand something about the history of Linux. It is only relatively recently that the need for a journaling filesystem has arisen in the Linux world. Up until a year or two ago, Linux was still mostly the domain of hackers and geeks. There were some businesses using it, but largely "on the quiet" and for smaller tasks. Given that user community, the desired features in the filesystem were low complexity and fast general performance. ext2 fits the bill. In the event of a system crash, fsck can put the filesystem back together quickly enough.
Now that Linux is getting attention and being deployed on larger systems, though, fsck times have become an issue. For a really large array (say, 500 GB), without special tuning, fsck times in the eight to ten hour range are not unheard of. We can no longer wait for fsck, and so Linux needs a better solution.
Which leads to the next problem: Filesystems are important. Real important. Filesystem corruption is probably one of the worst possible software failures you can have. Even kernel crashes can be dealt with (just look at NT), but if your filesystem is trashed, then your system is dead. For good. You have to recreate the system (i.e., restore from backup) to continue.
Thus, when it comes to filesystems for general production use, the Linux Kernel Team(TM) makes damn sure they do it right. A filesystem has to be extraordinarily stable before it can be considered ready for the production kernel.
(Oh, and as an aside: "I'm not trying to troll here" should go right up there with "The check is in the mail" and "It's only a cold sore".)
... someone might kinda sorta get ready to start considering how to start thinking about looking at ext3 by 2.7 or so...
When Ted Ts'o came and spoke at the local Linux User Group a couple weeks ago, he seemed much more optimistic about the ext3 filesystem making it into Linux for 2.5/2.6. He did indicate that there is little to no chance of it making it into 2.4. He also expressed concerns about ReiserFS and XFS being pushed in too quickly, though. (The argument basically being: Filesystems are damn important, so you have to make real sure you do 'em right.)
If the slashcode just forced the post to be previewed before submitting then it wouldn't be a problem.
Couldn't the bad-guy auto-post script just fake the HTML state that indicates you've previewed the comment at least once? You'd have to keep that state in a server-side database, which would add overhead to the existing system. How much overhead, I don't know.
Are you trying to say that there are not enough businesses in the entire world outside of the U.S. to keep a 'nation' about the size of a sports stadium in business?
Something like 90% of the world's Internet traffic goes though USA backbone sites. To get from Poland to France (on the net), you go though a USA site. The percentage of SSL traffic (the "important" stuff) is even higher. Even offshore business accounts often host on USA soil, because that's where the network is.
So, while the USA certainly doesn't have any magic control over the world economy, you cannot dispute the fact that most of the network traffic this thread is interested in goes through sites owned and operated by USA companies.
Now, consider that Sealand is a data haven. Without network connectivity, they are about as useful as a jet fighter without fuel. If the entire USA collectively decides to cut off Sealand, then they've lost most of the market they care about.
Is that going to happen any time soon? No. Too many laws against it, and too much popular opinion behind the laws. It would take a radical change in our culture and political climate. By the time the needed changes could be put into effect, the backbone situation will likely have changed.
... the Thinking Machines CM-5... used Sun servers. I'm sure there are others that used less-powerful system to run mathematical behemoths.
Yup. The Sun Enterprise 10000 (AKA "Starfire") uses a dedicated Ultra 5 as the console/management station. It connects via dedicated ethernet to the Starfire.
If I was moderating that day, I would have just left it at its original +1. What really gets me is that two moderators thought it was worthy of their points.
Just FYI, but there was only one +1 moderation done to that comment. My karama is high enough that my posts get the "+1 Bonus" that many talk about.
Was that post worthy of the "+1 Bonus"? No. As I said, I expected to get moderated down, and posted accordingly. I can't be good all the time.;-)
There are ways to reduce the impact of the clustering, but it will never be better than a parallel computer.
That's complete bunk. Whether a centralized multiprocessor machine or a massively-parallel distributed cluster would be faster depends completely on the task at hand. Specifically: How parallel is the task?
If the task can be broken up into many completely self-contained pieces, then a cluster will generally win. You can buy lots of low-end hardware cheaper then you can buy even very good high-end hardware.
If the task contains contention points or data access is very random, then you're better off with a single multiprocessor machine. An example of a contention point would be the locks in a database. An example of random data access would be logins to Slashdot.
Finally, it is worth pointing out that, after a certain point, most large machines have to move to a NUMA design, at which point you start to resemble a massively parallel cluster anyway.
Anyone who has submited a story and then had it posted w/o giving you credit -- or someone else posts something that you feel is much less important than your story will know exactly what you mean. People like CmdrTaco have an inside track and it seems like they can post whatever they want while sometimes it seems like it takes an act of congress for anyone else's story to get posted.
Okay, folks. Some people have a really warped idea of what this site is.
Please note that the button on the submit page is labeled "Submit Story". It is not labeled "Publish Story". It most definitely not labled "Splash my handle on the Slashdot home page and give me an ego boost".
There are a lot of people who seem to think getting a submission published on Slashdot is like loosing your virginity or getting your drivers license: Something you have to do to "be cool" or whatever.
Folks, that isn't the way it works.
This is Rob's site. The Great Taco can do whatever the hell he wants to. (Likewise, you're under no obligation to read this site. But that's another story...) The "Submit" page is so you can tell Rob & Co about something you think is cool. There is absolutely no obligation, of any kind, involved.
Anyone who thinks otherwise is having delusions of grandeur.
Come on, why the hell has this comment been moderated up to +3 insightful?... There's nothing insightful about it, it's just pure flame bait, nothing more, nothing less.
Ya know, I was wondering that myself, and I posted the comment. I really expected to be moderated down for that. It isn't quite "pure" flame-bait, as that is my honest opinion, but it surely is a flame-worthy message.
Okay, I admit I'm not a Kerberos expert, but I've looking into this issue some, and it appears to me that everyone is up in arms for all the wrong reasons.
As I understand it: Microsoft took a field in Kerberos marked for "vendor-specific data" and used it for -- get ready for this -- vendor-specific data. (If that is wrong, please feel free to correct me.)
So there is nothing wrong with Microsoft's Kerberos implementation. Getting mad at them for that is incorrect.
However, Microsoft has done some things worth getting mad about: First, the vendor-specific data is in a closed, proprietary format, designed to lock-out non-Microsoft implementations; and second, they've threatened Slashdot for what are (IMO) silly reasons (the exact merits of their case have been debated to death elsewhere; let's not repeat all that here).
We should be after MSFT to open up their protocols and compete fairly, and not after them for using a field in Kerberos for what it is designed for.
Right. Then you ask him how much it cost, and you can laugh at him. Normal people can actually afford to make Beowulf or MOSIX clusters.
That ain't anything close to VMS clustering.
Beowulf and friends are distributing processing tools. They take an easily paralizable job and handle the mechanics of distributing it for you. Beowulf is mostly application-level software; the machines still function as seperate hosts.
A VMS cluster essentially turns a group of machines into a single machine. All resources are multiplexed into a single logical unit. If one of them fails, the others pick up all the work it left behind. Beowulf is nothing like it. Doesn't even come close.
(I say this as someone who would rather bash his head into the wall then use VMS. I prefer Unix, but I know where we still haven't beat the competition -- yet.)
It shows that VMS can take much more flaky hardware errors than can UNIX.
Well, I suppose VMS could implement some kind of ECC code in software, sort of a RAID-for-RAM, if you will, and that might help -- unless the bad RAM contains the ECC engine, of course. But really, this sort of thing is a hardware issue.
It's true. Even Win95 is less susceptible to flaky hardware than Linux is. Some severly overclocked systems will run on Win95, but crash on Linux.
Well, that's kind of a mis-truth. You often hear stories of Windows running where Linux does not. This is usually because of one of two reasons:
(1) Cheap OEMs designing hardware that works with Windows, rather then designing hardware that meets the specifications.
(2) Windows doesn't use as much of the hardware as it should. For example, in any SMP system, the extra processors could be defective and Win95 wouldn't notice. Linux would. Does that make Win95 fault tolerant?
(As an aside: I suspect BeOS is the same as Linux here. BeOS also is a much more sophisticated OS then '95.)
Take a PC running Win95 for example. If one... kills an I/O processor, more likely than not, a UNIX will crash. Another bullet-proof OS, for example, could handle an I/O processor going out in the middle of operation and still keep running.
Er, not really. Most OSes I've seen will recover nicely enough if a non-critical system fails. Even Win9X will, assuming failing applications don't take out the kernel in their death. Now, if, say, the system drive fails, then you can bet that will kill most any system. The solution is redundant hardware -- it works better and faster then software solutions anyway.
(I don't know what an "I/O processor" is supposed to be. Severely failing hardware on a PC generally generates an NMI, which cannot be trapped, by any OS.)
I have no experiance with VMS...
But hey, lack of knowledge has never stopped you from posting before, right?;-)
Sure, I like UNIX, but there are many advantages to VMS about which a UNIX only geek of today might never learn. VMS is a solid, highly secure OS -- to toss this technology away is plain folly.
I honestly think we'd be better off just devoting time and effort to fixing the (few) areas where the free Unixes are not as good as VMS then we would be trying to salvage anything useful from it.
I really, really don't like VMS. It may be a stable system, but I can't help but wonder if that isn't because it is even uglier then Unix is, and thus no one uses it. The Unix command line at least appeals to geeks after they get to know it, but even the VMS advocates I know agree that it is Very Messy Syntax.
"I've used Mach; Mach is one of the reasons I think micro-kernels are a bad idea. I've used VMS; VMS is one of the reasons I think VMS is a bad idea." -- Linus Torvalds
I mean, if I had $ 1 million a year, I could run multiple redundant T3s to my own office and use them for my own personal/company use, too.
I'm sure you could. But Exodus doesn't deal with anything as slow as multiple T3's. They advertise the fact that each of their data centers has multiple OC-12 lines. Each one of those is capable of 622.08 megabits per second. A T3 gets you only 44.736 Mbps.
They have huge battery systems, power conditioners, and multiple disel generators at each site. They are usually connected to more then one power grid. They can function without commercial power indefiniately.
They have redundent everything. Network feeds. Routers. Switches. Power. Cooling.
They have very tight security. Armed guards. Biometric (e.g., hand print) locks. Cameras. Steel doors. Double-walls. Personnel locks.
You use Exodus if you absolutely, positively cannot afford downtime due to third-party service failures. If you have to ask, you can't afford it.
Exodus is one of the world's biggest (in terms of service capacity available) Internet Service Providers.
"We're going to need bandwidth. Lots of bandwidth."
Exodus specializes in having more bandwidth then most of the third world. They've got NAPs (Network Access Points, i.e., backbone connections) all over the continental United States, and a few outside the US as well. They link this all together using both external and internal networks. The end result is, most anywhere on the net that has a good connection, has a good connection to Exodus.
They provide servers. Do you need to host downloads for ten million users? Exodus can give you servers to do so.
They provide co-location space. If their standard server packages just won't cut it -- bring your own. They'll give you a rack, a dedicated co-loc cage, or a dedicated high security vault.
Their web page has a lot of graphics because they have a lot of pictures of their equipment and graphs of their capacity. It is actually justified. You may want to make a return trip.
Slashdot Mirror
... does anybody remember Geoworks? That was a cool piece of software...
GeoWorks Ensemble! AKA PC/GEOS! Now NewDeal. That program has had more names then Prince, but it was still cool. I still miss GeoWrite. Still the best WYSIWYG word processor I've used, to this day.
For those who don't know, PC/GEOS was a multi-tasking, multi-threading, object-oriented GUI operating environment that ran on a PC/XT with 640KB of RAM and 10 MB of hard disk. And it was pretty fast, too. A company is trying to resurrect it as NewDeal Office, but I don't put high hopes on it.
Ansible was talking about an 8-port 100mbit switch, not a hub.
I'm being truly pedantic here, but technically, a hub, a switch, and a repeater are all different things.
Hub is simply a generic term for anything that forms the center of a star topology cabling system. Key phone systems have hubs. Electrical equipment often has hubs. TP Ethernet has hubs.
Repeaters do just that: Repeat an Ethernet transmission (frame) on other ports. Repeaters operate at the electrical level; they don't have any intelligence. Originally used simply to work around cable length limits, multi-port repeaters are required for TP.
Switch is a term for a multi-port bridge. A bridge is a small computer with at least two network interfaces. It receives Ethernet frames, figures out which interface the destination is on, and sends it appropriately (or not at all, if the destination and source interface are the same). If the bridge doesn't know which interface the destination is on, it sends the frame out on all of them (except the source interface).
Now, the breakup of Microsoft will certainly cost the company an immense amount of money as it buys new lodgings for the spinoff company, duplicates support staff, reorganizes, etc. But none of that money will be going to the companies that were harmed by Microsoft's illegal actions. I'd have liked to see them get a few hundred million each.
Any company that feels it was harmed by Microsoft's business practices can sue them for damages. Given the evidence, they may well win.
Microsoft will likely be beset by several hundred rather large house flies in the next few years.
Trust me, there's enough of it already. Start clicking on random things on freshmeat and you'll see what I mean. Yeah, there's lots of good stuff, but OTOH there's a lot of bad stuff too.
Of course. Sturgeon's Law: Ninety percent of everything is crap.
Linux has been above the curve for some time now; I don't think we can expect it to stay there forever. Fortunately, the really good Open Source/Free Software that is out there now will still be out there when the signal-to-noise ratio drops.
First of all, as stated by some others, Windows has already had ATA 100 support.
...
Which I wasn't disputing.
IBM's latest IDE release, the 75GXP, maxes out at 37MB/s sustained
Ok, I stand corrected; ATA/66 does have some use then. ATA/100 is still completely worthless to anyone but marketing types, and is likely to remain so for its service lifetime.
SCSI 160 is bogus at this point, ignoring the fact you will need 4 of either of the above mentioned 2 drives to max the throughput.
And I can fly, ignoring the fact that gravity pulls me down.
Point being: That is a pretty big fact to ignore there! Wide SCSI supports up to sixteen devices on the bus at a time, all of which can have operations pending at once (unlike ATA). So, unlike ATA, you can very much use the full potential of SCSI Ultra160. In fact, we could use more bus bandwidth!
The problem is that it requires a 64bit PCI slot which no mainstream consumer level board carries.
That sounds an awful lot like "640K should be enough for anybody!" to me.
First, about five years ago, no mainstream consumer-level board had PCI slots at all. I am sure we will see 64-bit PCI in consumer systems soon enough.
Second, anyone buying four super-fast hard drives is probably not going to be putting them in their 80386 with 4MB of RAM. They will likely have either 66 MHz PCI slots, 64-bit PCI slots, or both.
(For example, one system we have in our integration room right now has four PCI buses with two slots each. Two of those buses are 64-bit. That same system has eight high-speed drives on two Ultra160 SCSI buses using only one IRQ. Do that with ATA!)
32bit PCI which we all have has a theoretical limit of 133MB/s for the whole bus.
First, don't use "all". Speak for yourself.
Second, if 32-bit PCI is such a limit, isn't ATA/100 kind of pushing things? Won't serial ATA be completely pointless?
Hey, let's all go back to Tandy TRS-80s and Apple ][e's, this new stuff is too fast!
The release of USB2 will eliminate almost all of SCSI's external peripherals, except for highend hard drives.
USB2 is still vaporware right now. SCSI Ultra160 you can buy today.
Personally, I think USB is appropriate for devices like scanners, cameras, and the like. I think SCSI is more appropriate for hard drives, tape drives, and other mass storage systems.
Most current gen IDE drives do support command queuing. Whether or not it actually works is anyone's guess...
Another thing to love about ATA: Nobody knows if you're actually getting the performance you should or not!
The reason ATA CDROM drives used to kill system performance was due to a lack of DMA, not because IDE could only access one device per bus at a time. The advent of DMA has elminated this bottleneck.
Then why does ATA CD-ROM multi-tasking performance still blow chunks, when SCSI systems barely even notice the load? Could it be because the whole ATA subsystem has to STOP and wait for the CD-ROM to complete its transaction before that bus can become free again?
ATA sucks, pure and simple. Rather then putting money into trying to extend a dead technology, the industry should just switch to SCSI, which is (1) here today (2) widely used (3) a standard and (4) actually works.
Win2k does indeed support USB. And win2k is essentially a marketing name for NT 5.0, so Windows NT has supported USB for about 6 months now.
Actually, NT 4.0 has had USB support since 1998, I believe. However, Microsoft decided not to release the updates which enabled that support, preferring to use it as a level to force you to upgrade to Windows 2000.
I'm sure the fact that Linux also lacked good USB support had something to do with it.
Got that? Unless they have competition, Microsoft cares nothing about you or your desire to use the technologies you want.
Give me Linux any day.
How does ATA/100 compare with SCSI in speed? Reliability? Price?
Reliability
These days, the HDA (Hard Disk Assembly) is usually the same for both SCSI and ATA (IDE) disk drives. They simply use a different PCB (Printed Circuit Board) for different bus interfaces. Thus, all other things being equal, reliability should be about the same.
Of course, all other things are not equal. Generally speaking, very-low-end systems will never see SCSI, and very-high-end systems will never touch ATA. Thus, some really cheap (i.e., unreliable) drives are only available in ATA versions, and some really expensive (i.e., more reliable) drives are only available in SCSI versions.
Price
Despite the fact that there is no practical difference in cost of materials, SCSI drives are almost always much more expensive then their ATA counter-parts. There are two reasons for this.
The first is volume. A lot more ATA drives get sold then SCSI. But the second reason is the more significant one: SCSI is still considered a "high end" technology. Such technologies command a premium price. Thus, manufactures charge more for SCSI drives, and people are willing to pay it.
Performance and Features
ATA does not support device disconnection. This means only one device at a time can be using the bus. Since your average hard drive is going to sustain maybe 15 MByte/sec throughput, if you're lucky, even ATA/33 is overkill. ATA/66 and ATA/100 are completely useless to everyone but marketing types.
SCSI is the clear winner here:I've heard people claim that "most users don't need" the features and performance of SCSI. I disagree at two levels: One, I think they do. Even if it is just wondering why all their applications slow down when the CD-ROM drive is busy, regular users encounter the problems of ATA. And second, if regular users don't need that kind of performance, why are they bothering to upgrade ATA at all?
There's a lot of discussion over whether digital or analog is the better medium to use. I want to point out a few things most people are missing.
It is true that, in theory and all other things being equal, analog gives you better reproduction. Digital (by definition) requires you to sample a signal periodically. Changes occur in discrete steps. Analog gives you smooth transitions, as it isn't limited to a particular rate of sampling.
However, digital has other advantages that, IMNSHO, outweigh the advantages of analog in practical use.
First, digital can be reproduced, stored and retransmitted, without limit, without experiencing any signal degradation. Digital signals can also be encoded with redundant data for error correction. Digital media is also generally more resistant to physical degradation from repeated use then analog media (although that is more by accident then through conscious design).
The end result is that while, in theory, an analog signal offers better reproduction, a digital signal will often have better quality, because analog media tends to get worn out quicker and more easily then digital. This is why I like CD over vinyl records; CDs don't pop and hiss like my records used to. This is why I like DVD over VHS; DVDs do not degrade with multiple viewings.
With proper care, you can generally prevent analog systems from degrading in this manner, but neither I, nor your average movie theater teenage projector jocky, treat media that well.
One other thing: The analog purists argue that digital is inferior because digital is sampled. It is interesting to note that motion video of any type is already sampled: What we perceive as motion is really a series of still frames. If a sampled signal is automatically disqualified, then all motion video is disqualified.
what is the technical/financial/legal difference between different long-distance phone companies both providing service to the same area and two cable companies doing the same thing?
You can switch long distance companies because they are providing a connection to distance sites. You still have the same LEC (Local Exchange Carrier), which actually brings the wire to your house. Now, there are (in places) CLECs (Competitive LECs) available, but what they do is essentially buy local lines in bulk from the primary LEC.
The limitation is basically the fact that physical wires need to be strung on poles. You can talk of competition among the phone, power, and cable companies, but there is still only one set of power lines, one set of phone lines, and one TV cable running down your street. That is what the technical limitation is.
The problem with a company like AT&T is that they have a vertical monopoly in particular locales. They don't own more then 30% of the service nationwide, but in the areas they do service, AT&T is the:
- Phone Local Exchange Carrier (runs wires to your house, connects you to your neighbor)
- Long-distance phone carrier (connects you to Aunt Marge)
- Cable TV wire carrier (runs cable to your house)
- Television programming producer (makes TV shows ("content"))
- Internet Service Provider (connects to the 'net backbone)
- Internet Content Provider
Aunt Marge may not have AT&T's monopoly, but she may have Time-Warner's instead (or whatever).
That's the objection these people have.
Reiser FS looks seriously cool. [However] it isn't well integrated.
.99 and its still not done?
That's mainly because ReiserFS is still considered to be very experimental (well, by everyone but Hans Reiser), and will not be part of the 2.4 production kernel. The earliest it can hope for adoption is the 2.5/2.6 series. This isn't because of flaws in ReiserFS; it is because of timing. ReiserFS started reaching usable stability too late for it to be integrated into the 2.4 kernel.
I have to wonder, though, what the hold-up is. I'm not trying to troll here, but what's taking them so damn long? They've been working on it since
First, they haven't been working on it since 0.99. That's just plain false.
Second, you have to understand something about the history of Linux. It is only relatively recently that the need for a journaling filesystem has arisen in the Linux world. Up until a year or two ago, Linux was still mostly the domain of hackers and geeks. There were some businesses using it, but largely "on the quiet" and for smaller tasks. Given that user community, the desired features in the filesystem were low complexity and fast general performance. ext2 fits the bill. In the event of a system crash, fsck can put the filesystem back together quickly enough.
Now that Linux is getting attention and being deployed on larger systems, though, fsck times have become an issue. For a really large array (say, 500 GB), without special tuning, fsck times in the eight to ten hour range are not unheard of. We can no longer wait for fsck, and so Linux needs a better solution.
Which leads to the next problem: Filesystems are important. Real important. Filesystem corruption is probably one of the worst possible software failures you can have. Even kernel crashes can be dealt with (just look at NT), but if your filesystem is trashed, then your system is dead. For good. You have to recreate the system (i.e., restore from backup) to continue.
Thus, when it comes to filesystems for general production use, the Linux Kernel Team(TM) makes damn sure they do it right. A filesystem has to be extraordinarily stable before it can be considered ready for the production kernel.
(Oh, and as an aside: "I'm not trying to troll here" should go right up there with "The check is in the mail" and "It's only a cold sore".)
... someone might kinda sorta get ready to start considering how to start thinking about looking at ext3 by 2.7 or so ...
When Ted Ts'o came and spoke at the local Linux User Group a couple weeks ago, he seemed much more optimistic about the ext3 filesystem making it into Linux for 2.5/2.6. He did indicate that there is little to no chance of it making it into 2.4. He also expressed concerns about ReiserFS and XFS being pushed in too quickly, though. (The argument basically being: Filesystems are damn important, so you have to make real sure you do 'em right.)
If the slashcode just forced the post to be previewed before submitting then it wouldn't be a problem.
Couldn't the bad-guy auto-post script just fake the HTML state that indicates you've previewed the comment at least once? You'd have to keep that state in a server-side database, which would add overhead to the existing system. How much overhead, I don't know.
Are you trying to say that there are not enough businesses in the entire world outside of the U.S. to keep a 'nation' about the size of a sports stadium in business?
Something like 90% of the world's Internet traffic goes though USA backbone sites. To get from Poland to France (on the net), you go though a USA site. The percentage of SSL traffic (the "important" stuff) is even higher. Even offshore business accounts often host on USA soil, because that's where the network is.
So, while the USA certainly doesn't have any magic control over the world economy, you cannot dispute the fact that most of the network traffic this thread is interested in goes through sites owned and operated by USA companies.
Now, consider that Sealand is a data haven. Without network connectivity, they are about as useful as a jet fighter without fuel. If the entire USA collectively decides to cut off Sealand, then they've lost most of the market they care about.
Is that going to happen any time soon? No. Too many laws against it, and too much popular opinion behind the laws. It would take a radical change in our culture and political climate. By the time the needed changes could be put into effect, the backbone situation will likely have changed.
Last time I heard a discussion about supercomputers, someone said that a supercomputer had to have a sustainable throughput of at least 1 Gigaflop.
I always liked the definition, "Any computer that is worth more then you are."
;-)
... the Thinking Machines CM-5 ... used Sun servers. I'm sure there are others that used less-powerful system to run mathematical behemoths.
Yup. The Sun Enterprise 10000 (AKA "Starfire") uses a dedicated Ultra 5 as the console/management station. It connects via dedicated ethernet to the Starfire.
If I was moderating that day, I would have just left it at its original +1. What really gets me is that two moderators thought it was worthy of their points.
;-)
Just FYI, but there was only one +1 moderation done to that comment. My karama is high enough that my posts get the "+1 Bonus" that many talk about.
Was that post worthy of the "+1 Bonus"? No. As I said, I expected to get moderated down, and posted accordingly. I can't be good all the time.
There are ways to reduce the impact of the clustering, but it will never be better than a parallel computer.
That's complete bunk. Whether a centralized multiprocessor machine or a massively-parallel distributed cluster would be faster depends completely on the task at hand. Specifically: How parallel is the task?
If the task can be broken up into many completely self-contained pieces, then a cluster will generally win. You can buy lots of low-end hardware cheaper then you can buy even very good high-end hardware.
If the task contains contention points or data access is very random, then you're better off with a single multiprocessor machine. An example of a contention point would be the locks in a database. An example of random data access would be logins to Slashdot.
Finally, it is worth pointing out that, after a certain point, most large machines have to move to a NUMA design, at which point you start to resemble a massively parallel cluster anyway.
Anyone who has submited a story and then had it posted w/o giving you credit -- or someone else posts something that you feel is much less important than your story will know exactly what you mean. People like CmdrTaco have an inside track and it seems like they can post whatever they want while sometimes it seems like it takes an act of congress for anyone else's story to get posted.
Okay, folks. Some people have a really warped idea of what this site is.
Please note that the button on the submit page is labeled "Submit Story". It is not labeled "Publish Story". It most definitely not labled "Splash my handle on the Slashdot home page and give me an ego boost".
There are a lot of people who seem to think getting a submission published on Slashdot is like loosing your virginity or getting your drivers license: Something you have to do to "be cool" or whatever.
Folks, that isn't the way it works.
This is Rob's site. The Great Taco can do whatever the hell he wants to. (Likewise, you're under no obligation to read this site. But that's another story...) The "Submit" page is so you can tell Rob & Co about something you think is cool. There is absolutely no obligation, of any kind, involved.
Anyone who thinks otherwise is having delusions of grandeur.
End of rant.
Come on, why the hell has this comment been moderated up to +3 insightful? ... There's nothing insightful about it, it's just pure flame bait, nothing more, nothing less.
Ya know, I was wondering that myself, and I posted the comment. I really expected to be moderated down for that. It isn't quite "pure" flame-bait, as that is my honest opinion, but it surely is a flame-worthy message.
I'll echo bungo: Who on Earth moderated that up?
Okay, I admit I'm not a Kerberos expert, but I've looking into this issue some, and it appears to me that everyone is up in arms for all the wrong reasons.
As I understand it: Microsoft took a field in Kerberos marked for "vendor-specific data" and used it for -- get ready for this -- vendor-specific data. (If that is wrong, please feel free to correct me.)
So there is nothing wrong with Microsoft's Kerberos implementation. Getting mad at them for that is incorrect.
However, Microsoft has done some things worth getting mad about: First, the vendor-specific data is in a closed, proprietary format, designed to lock-out non-Microsoft implementations; and second, they've threatened Slashdot for what are (IMO) silly reasons (the exact merits of their case have been debated to death elsewhere; let's not repeat all that here).
We should be after MSFT to open up their protocols and compete fairly, and not after them for using a field in Kerberos for what it is designed for.
Right. Then you ask him how much it cost, and you can laugh at him. Normal people can actually afford to make Beowulf or MOSIX clusters.
That ain't anything close to VMS clustering.
Beowulf and friends are distributing processing tools. They take an easily paralizable job and handle the mechanics of distributing it for you. Beowulf is mostly application-level software; the machines still function as seperate hosts.
A VMS cluster essentially turns a group of machines into a single machine. All resources are multiplexed into a single logical unit. If one of them fails, the others pick up all the work it left behind. Beowulf is nothing like it. Doesn't even come close.
(I say this as someone who would rather bash his head into the wall then use VMS. I prefer Unix, but I know where we still haven't beat the competition -- yet.)
It shows that VMS can take much more flaky hardware errors than can UNIX.
... kills an I/O processor, more likely than not, a UNIX will crash. Another bullet-proof OS, for example, could handle an I/O processor going out in the middle of operation and still keep running.
;-)
Well, I suppose VMS could implement some kind of ECC code in software, sort of a RAID-for-RAM, if you will, and that might help -- unless the bad RAM contains the ECC engine, of course. But really, this sort of thing is a hardware issue.
It's true. Even Win95 is less susceptible to flaky hardware than Linux is. Some severly overclocked systems will run on Win95, but crash on Linux.
Well, that's kind of a mis-truth. You often hear stories of Windows running where Linux does not. This is usually because of one of two reasons:
(1) Cheap OEMs designing hardware that works with Windows, rather then designing hardware that meets the specifications.
(2) Windows doesn't use as much of the hardware as it should. For example, in any SMP system, the extra processors could be defective and Win95 wouldn't notice. Linux would. Does that make Win95 fault tolerant?
(As an aside: I suspect BeOS is the same as Linux here. BeOS also is a much more sophisticated OS then '95.)
Take a PC running Win95 for example. If one
Er, not really. Most OSes I've seen will recover nicely enough if a non-critical system fails. Even Win9X will, assuming failing applications don't take out the kernel in their death. Now, if, say, the system drive fails, then you can bet that will kill most any system. The solution is redundant hardware -- it works better and faster then software solutions anyway.
(I don't know what an "I/O processor" is supposed to be. Severely failing hardware on a PC generally generates an NMI, which cannot be trapped, by any OS.)
I have no experiance with VMS...
But hey, lack of knowledge has never stopped you from posting before, right?
Sure, I like UNIX, but there are many advantages to VMS about which a UNIX only geek of today might never learn. VMS is a solid, highly secure OS -- to toss this technology away is plain folly.
I honestly think we'd be better off just devoting time and effort to fixing the (few) areas where the free Unixes are not as good as VMS then we would be trying to salvage anything useful from it.
I really, really don't like VMS. It may be a stable system, but I can't help but wonder if that isn't because it is even uglier then Unix is, and thus no one uses it. The Unix command line at least appeals to geeks after they get to know it, but even the VMS advocates I know agree that it is Very Messy Syntax.
"I've used Mach; Mach is one of the reasons I think micro-kernels are a bad idea. I've used VMS; VMS is one of the reasons I think VMS is a bad idea." -- Linus Torvalds
I mean, if I had $ 1 million a year, I could run multiple redundant T3s to my own office and use them for my own personal/company use, too.
I'm sure you could. But Exodus doesn't deal with anything as slow as multiple T3's. They advertise the fact that each of their data centers has multiple OC-12 lines. Each one of those is capable of 622.08 megabits per second. A T3 gets you only 44.736 Mbps.
They have huge battery systems, power conditioners, and multiple disel generators at each site. They are usually connected to more then one power grid. They can function without commercial power indefiniately.
They have redundent everything. Network feeds. Routers. Switches. Power. Cooling.
They have very tight security. Armed guards. Biometric (e.g., hand print) locks. Cameras. Steel doors. Double-walls. Personnel locks.
You use Exodus if you absolutely, positively cannot afford downtime due to third-party service failures. If you have to ask, you can't afford it.
Who/what exactly is Exodus?
Exodus is one of the world's biggest (in terms of service capacity available) Internet Service Providers.
"We're going to need bandwidth. Lots of bandwidth."
Exodus specializes in having more bandwidth then most of the third world. They've got NAPs (Network Access Points, i.e., backbone connections) all over the continental United States, and a few outside the US as well. They link this all together using both external and internal networks. The end result is, most anywhere on the net that has a good connection, has a good connection to Exodus.
They provide servers. Do you need to host downloads for ten million users? Exodus can give you servers to do so.
They provide co-location space. If their standard server packages just won't cut it -- bring your own. They'll give you a rack, a dedicated co-loc cage, or a dedicated high security vault.
Their web page has a lot of graphics because they have a lot of pictures of their equipment and graphs of their capacity. It is actually justified. You may want to make a return trip.