It doesn't give you any advantages over using a good SCSI solution, like triple channel Ultra160 or Ultra320 cards.
You make some valid points, but my main point was one of cost-efficiency.
Also, with PCI 64/66, triple Ultra 160/320 is pretty silly. A 8 disk ATA contoller already maxes out said bus, so does a single Ultra 320 channel. Until we get faster busses, all that speed is wasted anyway, the only thing you eliminate is some device contention, something that doesn't exist in SATA/PATA one-disk-per-channel controllers.
Also, about "low-end", do you consider a 5-10TB RAID low-end? As I said, the only drawback is in heavy transaction loads, latency sensitive applications, because of lack of intelligent command queuing (sometimes, as I said the integrated devices that have SCSI on the outside and ATA on the inside get around this in large part) and lower RPMs.
If I had the money, I'd love to see modern ATA put up against modern SCSI in real world tests. It's a hands down winner in cost/GB, performance is not as clear, but I bet it would be "good enough" for a lot of applications, and that is what matters.
As far as your comments on reliability... I see this as more of a non-issue, since the useful life of a RAID is far less than 5 years in a lot of applications. As I pointed out in my real world examples, the cost to maintain a 5 year old RAID outweighs the cost of newer technology in many cases. Immediate redundancy is taken care of, because one can easily afford completely redundant systems, in addition to RAID on each, and still have a cost advantage over SCSI.
As far as platform independance, the ATA-SCSI hardware RAID devices take care of that if you need it. If you don't, lower cost internal solutions can be used.
Also, I don't think SCSI is obselete, I just see ATA as the disruptive technology that will soon relegate it to highly specialized applications. In a way it is sad, I admit, SCSI/FC is a far better technology. The writing is on the wall it seems though, and cheap and "good enough" seems to have won again, just like VHS/Beta.
You are already guaranteed limited liability to $50 and chargeback rights by law. The credit card companies sell that fraud protection because they know it doesn't really cost them anything, since it's mostly what they have to provide anyway.
This is an article I wrote for another forum, I think it fits well in this discussion. ---------- Maxtor recently announced a 320GB ATA drive. This, in itself, isn't too remarkable, other than the quite large size jump. What is remarkable is that they are targetting it to somewhere that has never been specifically aimed at with ATA drives, the large corporate storage market.
About a year ago, my company was running out of hard disk space, again. We had filled up our 500GB main live SCSI RAID, and our 190GB work in process Fiber Channel RAID was looking pretty small and obselete.
For reference, we paid about $30,000 for the 500GB SCSI RAID only three years ago. The Fibre Channel RAID came about a year earlier and cost about $45,000. Support on both is over $5,000 a year.
Our main archive needs are for mostly dead storage, that is, speed is not critical, but availability is. The main archive was previously housed on a very slow 1TB tape robot. The tapes were becoming a bottleneck in our workflow, so about two years ago I designed a shell script system in IRIX to allow the files to be stored on live RAIDs without the risk of using older versions of files by accident.
This worked fine, but with the increased throughput, we also had increasing demands for more storage. Our archive has increased in size almost 100% since 2 years ago, and grows by nearly 5GB per week or more.
So, about a year ago, things were starting to get tight again. I looked at the options, and Maxtor 100GB ATA drives were out. I figured we could use Linux in this situation, and just attach the storage to the network, since speed was not critical. After some searching, I found 3ware, www.3ware.com, who makes a PCI ATA RAID controller that holds up to 8 drives and has good Linux support.
There were some issues at first, due to a hardware recall on 3ware's part, but after the bugs were worked out, I managed to build two NAS systems. These were a huge leap forward in terms of storage cost, each server cost about as much as we pay in one year of support on the old RAID systems, and were complete computers, compared to simple direct attached storage. We got 3 TB of storage for about $12,000.
These storage servers were pretty novel a year ago. Very few people had built multiple Terabyte systems from ATA drives at that point. Now they are becoming pretty common. For example, RaidZone (www.raidzone.com) makes a similar NAS (Network Attached Storage) product based on Linux. www.acnc.com is also embracing ATA as a viable platform for large RAID, though they are packaging it with special controllers that allow one to use the ATA RAID with any operating system, the OS sees the RAID as a single, directly attached storage device, through SCSI and fibre channel. This has the added benefit of allowing SCSI features like command reordering, and more robust command integrity checking.
The coming of serial ATA over the next year is serving to quickly push ATA into the mainstream low-to-mid end server market. For large data storage needs, tapes no longer offer a viable option. Linux based NAS and hardware based directly attached ATA storage are quickly becoming the de facto standard for large archival storage, at 25% or less of the cost of tapes, without the hassle that comes with the non-random access of tapes.
ATA is often viewed in the industry as a "consumer level" technology, or "not for serious use". This view is quickly becoming outdated with the wealth of new ATA based server products.
ATA in many ways has a history a lot like Linux's. It is a low cost technology, that is quickly eating away the market of more "serious" technologies like SCSI and Tape. This is the same way Linux is eating away at the old UNIX market, replacing so called "high end" servers with commodity hardware. It's no surprise that many ATA NAS products are Linux based, for an embedded system where cost is a primary concern, a Windows license would add unnecessarily to the cost, and reduce reliability while limiting the freedom of the developer to customize the system to the application.
Theere is a long history of lower cost disruptive technologies killing superior technologies. ATA isn't technologically better (yet). It is, however, good enough, and much much lower cost, and that's all it needs.
The lack of high speed 10,000 RPM ATA drives does tend to make it not currently viable for high speed, random access, low latency, applications at this point, such as heavily loaded databases. There is no reason a high RPM ATA drive could not be produced in the future though. In my experience, most high capacity storage needs are archival, or only require a low transaction rate with moderate to high data rates.
In terms of raw throughput, ATA RAID can easily outpace any SCSI RAID, do to the nature of the interface. ATA RAID is implemented like switched ethernet, each drive has its own channel to the host controller. SCSI and basic configurations of Fibre Channel are like old fashioned shared ethernet, ATA is like switched ethernet.
ATA is Linux's new partner in the server room. Together, they are drastically reducing the cost of operations, while at the same time, offering a much more capable solution for many common needs.
In most browsers, you can just hit reload on the "result" page to resubmit the form. Like on slashdot when I type a message too quickly and I am punished for being semi-literate, I get back at slashdot by hitting reload every couple seconds until the message goes through. I figure once they see that their lameness filter is chewing up tons of bandwidth, they will make it more reasonable.
i think businesses would do well to offer value-add services like wireless internet inside their stores free-of-charge just like they do restrooms. I think starbuck's plan will go the way of the pay-toilet. but this guy is going about demonstrating it the wrong way.
If it weren't for those brave rebels who dared to park their porta-potty in the parking lot of stores with pay toilets then we would all still be paying to use the toilet.
It looks like he has some sort of tubular antenna on there, wouldn't an omnidirectional be a better choice?
Maybe that antenna is the one for the fixed wireless service (which is is probably violating FCC laws by using mobile), and the 802.11 antenna is the thing on top of that?
You can technically make a beowulf cluster out of anything that supports openssh that you can compile the pieces of your program for.
A simple beowulf cluster is just a shell script that does some sshing to each client and compiling and running the job to be run, combine that with another trivial script to scp files over, and that's it.
MSN requires a shitload of ports to be forwarded in to a specific client, something like 1000 IIRC. I might be lumping that together with the MS Gaming Zone, the only time I used MSN messenger was when I used to play combat flight sim a couple years ago when I still had a Windows box.
ICQ uses UDP and I believe you have to play with UDP NAT and timeout options to get it to work reliably. Again, I havn't used ICQ since ipchains was the main linux firewall system, so things may have changed now.
It's possible that the "real" clients compensate for the firewall better than trillian, but I'm sure you could get it all working with some firewall/client tweaking.
Either way, my original point stands- important people are often crazy people.
Heh, OK. Sorry for jumping to conclusions. I've done my share of relaying information found in less than reliable books. (You see there's this mat, and you "jump" to "conclusions").
Red Hat also has more motivation to support older stuff, because I believe their specialized commercial distros usually trail their main distro by a few versions. Someone correct me if I am wrong.
OK... It's possible the book is incorrect however, considering the snippet I posted, it looks like they mixed up two seperate (but similar) events involving two different important figures in the history of electrical power.
You are correct, but it's just a matter of time until MS's glacial turn around time, and outright refusal to fix certain bugs, combined with a "windows update" that often doesn't apply all the needed fixes, or installs patches that undo other patches.... I could go on...
Anyway, it's going to bite them, in a big way. Recently some "combination attacks" have formed, i.e. a series of non-critical security flaws that can be combined to gain total system access.
This is combined with their aggressive end-of-life program which EOLs software that is still in widespread use, completely dropping even critical security bugfix support for said software. As Windows 2000 nears EOL in a couple years, that is when we will really see the shit hit the fan. Hell, my girlfriend got a contract job to migrate systems from NT4 to 2000 last week. With no compelling reasons to upgrade, a lot of people are going to be running unpatchable systems in a couple years. Of course this is MS's whole strategy, to force people to upgrade their software just to get critical bugfixes.
Read about Edison- he electrified the urinals at train stations and killed a hobo
I can't find one reference to this, at all. I think you just made it up.
The closest thing to it I could find was this:
1881 - Dr. Albert Southwick, a dentist and former steamboat engineer, sees elderly drunkard touch terminals of electrical generator in Buffalo, New York. He is amazed at how quickly and apparently painlessly the man is killed and describes episode to friend State Senator David McMillan.
Also, Edison was known to publicly demonstrate the "dangers of AC power" by electrocuting animals in public demonstrations in an attempt to discredit Tesla's AC power in favor of his DC system.
Don't start urban legends, get the fact right if you are going to spew something controversial like that.
We all know that Slashdot is whoring themselves out to MS now, with huge MS Visual Studio ads. I bet this is just an attempt to log the IP address of every Slashdot reader. Best to keep an eye on those troublemakers, right?
What are you going to compile the compiler verifier with? This just goes around in circles.
Assembly should be pretty easy to verify due to the close relation to machine code, but it would have to be a complete OS... And then you have to trust the hardware, which is another complete can of worms.
Honeypots are not for normal businesses to run. They are mostly of interest to security people who want to get insight into the latest tools and exploits.
I imagine there are honeypot applications for something like this. You could make a cracker totally believe they had broken in when in reality they are just in a UML.
For the ultra paranoid you could also make a backup copy of your whole UML partition and only run services in that, periodically restoring it from backup, and copying in the new data that is stored on the real OS. If you got broken into, it wouldn't really matter.
Slashdot Mirror
Also, with PCI 64/66, triple Ultra 160/320 is pretty silly. A 8 disk ATA contoller already maxes out said bus, so does a single Ultra 320 channel.
I correct myself, the bus is 528 MB/Sec in theory. This is mostly incorrect, Two 320s would still max it out.
It doesn't give you any advantages over using a good SCSI solution, like triple channel Ultra160 or Ultra320 cards.
You make some valid points, but my main point was one of cost-efficiency.
Also, with PCI 64/66, triple Ultra 160/320 is pretty silly. A 8 disk ATA contoller already maxes out said bus, so does a single Ultra 320 channel. Until we get faster busses, all that speed is wasted anyway, the only thing you eliminate is some device contention, something that doesn't exist in SATA/PATA one-disk-per-channel controllers.
Also, about "low-end", do you consider a 5-10TB RAID low-end? As I said, the only drawback is in heavy transaction loads, latency sensitive applications, because of lack of intelligent command queuing (sometimes, as I said the integrated devices that have SCSI on the outside and ATA on the inside get around this in large part) and lower RPMs.
If I had the money, I'd love to see modern ATA put up against modern SCSI in real world tests. It's a hands down winner in cost/GB, performance is not as clear, but I bet it would be "good enough" for a lot of applications, and that is what matters.
As far as your comments on reliability... I see this as more of a non-issue, since the useful life of a RAID is far less than 5 years in a lot of applications. As I pointed out in my real world examples, the cost to maintain a 5 year old RAID outweighs the cost of newer technology in many cases. Immediate redundancy is taken care of, because one can easily afford completely redundant systems, in addition to RAID on each, and still have a cost advantage over SCSI.
As far as platform independance, the ATA-SCSI hardware RAID devices take care of that if you need it. If you don't, lower cost internal solutions can be used.
Also, I don't think SCSI is obselete, I just see ATA as the disruptive technology that will soon relegate it to highly specialized applications.
In a way it is sad, I admit, SCSI/FC is a far better technology. The writing is on the wall it seems though, and cheap and "good enough" seems to have won again, just like VHS/Beta.
Yes, of course, it's the most abundant element in the universe. In theory it is limitless.
In practice, generating hydrogen takes more energy than it yields, therefore you really can't say it is an energy source, in practical terms.
excess of $50, you're paying for more than you're getting
I've never heard of the CC company making anyone pay the $50 either. It's apparently almost always waived.
That fraud protection is ironically a scam.
You are already guaranteed limited liability to $50 and chargeback rights by law. The credit card companies sell that fraud protection because they know it doesn't really cost them anything, since it's mostly what they have to provide anyway.
All that fuel is heavy
Batteries are heavier. That's the whole point of why this is "news", because batteries have a lot lower energy density than fuel.
There is a limitless supply of hydrogen,
Uhhh, what? Hydrogen is not an energy source! It is merely a storage medium for energy.
This is an article I wrote for another forum, I think it fits well in this discussion.
----------
Maxtor recently announced a 320GB ATA drive. This, in itself, isn't too remarkable, other than the quite large size jump. What is remarkable is that they are targetting it to somewhere that has never been specifically aimed at with ATA drives, the large corporate storage market.
About a year ago, my company was running out of hard disk space, again. We had filled up our 500GB main live SCSI RAID, and our 190GB work in process Fiber Channel RAID was looking pretty small and obselete.
For reference, we paid about $30,000 for the 500GB SCSI RAID only three years ago. The Fibre Channel RAID came about a year earlier and cost about $45,000. Support on both is over $5,000 a year.
Our main archive needs are for mostly dead storage, that is, speed is not critical, but availability is. The main archive was previously housed on a very slow 1TB tape robot. The tapes were becoming a bottleneck in our workflow, so about two years ago I designed a shell script system in IRIX to allow the files to be stored on live RAIDs without the risk of using older versions of files by accident.
This worked fine, but with the increased throughput, we also had increasing demands for more storage. Our archive has increased in size almost 100% since 2 years ago, and grows by nearly 5GB per week or more.
So, about a year ago, things were starting to get tight again. I looked at the options, and Maxtor 100GB ATA drives were out. I figured we could use Linux in this situation, and just attach the storage to the network, since speed was not critical. After some searching, I found 3ware, www.3ware.com, who makes a PCI ATA RAID controller that holds up to 8 drives and has good Linux support.
There were some issues at first, due to a hardware recall on 3ware's part, but after the bugs were worked out, I managed to build two NAS systems. These were a huge leap forward in terms of storage cost, each server cost about as much as we pay in one year of support on the old RAID systems, and were complete computers, compared to simple direct attached storage. We got 3 TB of storage for about $12,000.
These storage servers were pretty novel a year ago. Very few people had built multiple Terabyte systems from ATA drives at that point. Now they are becoming pretty common. For example, RaidZone (www.raidzone.com) makes a similar NAS (Network Attached Storage) product based on Linux. www.acnc.com is also embracing ATA as a viable platform for large RAID, though they are packaging it with special controllers that allow one to use the ATA RAID with any operating system, the OS sees the RAID as a single, directly attached storage device, through SCSI and fibre channel. This has the added benefit of allowing SCSI features like command reordering, and more robust command integrity checking.
The coming of serial ATA over the next year is serving to quickly push ATA into the mainstream low-to-mid end server market. For large data storage needs, tapes no longer offer a viable option. Linux based NAS and hardware based directly attached ATA storage are quickly becoming the de facto standard for large archival storage, at 25% or less of the cost of tapes, without the hassle that comes with the non-random access of tapes.
ATA is often viewed in the industry as a "consumer level" technology, or "not for serious use". This view is quickly becoming outdated with the wealth of new ATA based server products.
ATA in many ways has a history a lot like Linux's. It is a low cost technology, that is quickly eating away the market of more "serious" technologies like SCSI and Tape. This is the same way Linux is eating away at the old UNIX market, replacing so called "high end" servers with commodity hardware. It's no surprise that many ATA NAS products are Linux based, for an embedded system where cost is a primary concern, a Windows license would add unnecessarily to the cost, and reduce reliability while limiting the freedom of the developer to customize the system to the application.
Theere is a long history of lower cost disruptive technologies killing superior technologies. ATA isn't technologically better (yet). It is, however, good enough, and much much lower cost, and that's all it needs.
The lack of high speed 10,000 RPM ATA drives does tend to make it not currently viable for high speed, random access, low latency, applications at this point, such as heavily loaded databases. There is no reason a high RPM ATA drive could not be produced in the future though. In my experience, most high capacity storage needs are archival, or only require a low transaction rate with moderate to high data rates.
In terms of raw throughput, ATA RAID can easily outpace any SCSI RAID, do to the nature of the interface. ATA RAID is implemented like switched ethernet, each drive has its own channel to the host controller. SCSI and basic configurations of Fibre Channel are like old fashioned shared ethernet, ATA is like switched ethernet.
ATA is Linux's new partner in the server room. Together, they are drastically reducing the cost of operations, while at the same time, offering a much more capable solution for many common needs.
In most browsers, you can just hit reload on the "result" page to resubmit the form. Like on slashdot when I type a message too quickly and I am punished for being semi-literate, I get back at slashdot by hitting reload every couple seconds until the message goes through. I figure once they see that their lameness filter is chewing up tons of bandwidth, they will make it more reasonable.
i think businesses would do well to offer value-add services like wireless internet inside their stores free-of-charge just like they do restrooms. I think starbuck's plan will go the way of the pay-toilet. but this guy is going about demonstrating it the wrong way.
If it weren't for those brave rebels who dared to park their porta-potty in the parking lot of stores with pay toilets then we would all still be paying to use the toilet.
It looks like he has some sort of tubular antenna on there, wouldn't an omnidirectional be a better choice?
Maybe that antenna is the one for the fixed wireless service (which is is probably violating FCC laws by using mobile), and the 802.11 antenna is the thing on top of that?
You can technically make a beowulf cluster out of anything that supports openssh that you can compile the pieces of your program for.
...
...
./clustercopy.sh mysource.c '/home/cluster/work' ./clusterrun.sh 'gcc /home/cluster/work/mysource.c' ./clusterrun.sh '/home/cluster/work/a.out'
A simple beowulf cluster is just a shell script that does some sshing to each client and compiling and running the job to be run, combine that with another trivial script to scp files over, and that's it.
clusterrun.sh:
ssh cluster@192.168.0.1 $1
ssh cluster@192.168.0.2 $1
ssh cluster@192.168.0.n $1
clustercopy.sh
scp $1 cluster@192.168.0.1:$2
scp $1 cluster@192.168.0.2:$2
scp $1 cluster@192.168.0.n:$2
$
$
$
Anyway, that's all there is to it.
MSN requires a shitload of ports to be forwarded in to a specific client, something like 1000 IIRC. I might be lumping that together with the MS Gaming Zone, the only time I used MSN messenger was when I used to play combat flight sim a couple years ago when I still had a Windows box.
ICQ uses UDP and I believe you have to play with UDP NAT and timeout options to get it to work reliably. Again, I havn't used ICQ since ipchains was the main linux firewall system, so things may have changed now.
It's possible that the "real" clients compensate for the firewall better than trillian, but I'm sure you could get it all working with some firewall/client tweaking.
Either way, my original point stands- important people are often crazy people.
Heh, OK. Sorry for jumping to conclusions. I've done my share of relaying information found in less than reliable books. (You see there's this mat, and you "jump" to "conclusions").
You are correct, Red Hat still supports 6.2.
Red Hat also has more motivation to support older stuff, because I believe their specialized commercial distros usually trail their main distro by a few versions. Someone correct me if I am wrong.
OK... It's possible the book is incorrect however, considering the snippet I posted, it looks like they mixed up two seperate (but similar) events involving two different important figures in the history of electrical power.
You are correct, but it's just a matter of time until MS's glacial turn around time, and outright refusal to fix certain bugs, combined with a "windows update" that often doesn't apply all the needed fixes, or installs patches that undo other patches.... I could go on...
Anyway, it's going to bite them, in a big way. Recently some "combination attacks" have formed, i.e. a series of non-critical security flaws that can be combined to gain total system access.
This is combined with their aggressive end-of-life program which EOLs software that is still in widespread use, completely dropping even critical security bugfix support for said software. As Windows 2000 nears EOL in a couple years, that is when we will really see the shit hit the fan. Hell, my girlfriend got a contract job to migrate systems from NT4 to 2000 last week. With no compelling reasons to upgrade, a lot of people are going to be running unpatchable systems in a couple years. Of course this is MS's whole strategy, to force people to upgrade their software just to get critical bugfixes.
Or just run up2date -u or apt-get upgrade.
I know, I know, don't feed the trolls. I just had to a little this time.
Read about Edison- he electrified the urinals at train stations and killed a hobo
I can't find one reference to this, at all. I think you just made it up.
The closest thing to it I could find was this:
1881 - Dr. Albert Southwick, a dentist and former steamboat engineer, sees elderly drunkard touch terminals of electrical generator in Buffalo, New York. He is amazed at how quickly and apparently painlessly the man is killed and describes episode to friend State Senator David McMillan.
Also, Edison was known to publicly demonstrate the "dangers of AC power" by electrocuting animals in public demonstrations in an attempt to discredit Tesla's AC power in favor of his DC system.
Don't start urban legends, get the fact right if you are going to spew something controversial like that.
The one thing to prevent errors is design. I spet 2 months one time designing an application before I got aournd to writeing it, and no errors.
Yes, your meticulousattention to detail is obvious from your well written post.
I wrote a stirring review of "Who moved my cheese?", but it was rejected. :(
And the link is at research.microsoft.com.
We all know that Slashdot is whoring themselves out to MS now, with huge MS Visual Studio ads. I bet this is just an attempt to log the IP address of every Slashdot reader. Best to keep an eye on those troublemakers, right?
compilation verifier
What are you going to compile the compiler verifier with? This just goes around in circles.
Assembly should be pretty easy to verify due to the close relation to machine code, but it would have to be a complete OS... And then you have to trust the hardware, which is another complete can of worms.
Honeypots are not for normal businesses to run. They are mostly of interest to security people who want to get insight into the latest tools and exploits.
I imagine there are honeypot applications for something like this. You could make a cracker totally believe they had broken in when in reality they are just in a UML.
For the ultra paranoid you could also make a backup copy of your whole UML partition and only run services in that, periodically restoring it from backup, and copying in the new data that is stored on the real OS. If you got broken into, it wouldn't really matter.