Yes, c is a fundamental constant. It is, by definition, 299792458 m/s. But that's not the point.
Rather than invoking any grand unified theory theories [sic], how about just reading what the AC said? Amazingly enough, this is a case where the AC is exactly right. The speed of light in diamond is slower than the speed of light in vacuum. 1.6 times slower, IIRC. So one might expect the speed of light in a diamond-fiber to be about 200,000 km/s.
That said, I have no idea WTF they're talking about with the 120 km/h stuff. $10 says they don't know either, or that it was a typo.
Why bother? Just type "suceptible" into google, and it says "Did you mean: susceptible". Click there, and it brings up stuff for that word. And in the top right gives a link you can click for the definition.
They *do* reduce it on-site. That's why they "only" need to transfer 500TB/week. You'd be amazed by the amount of data that gets taken, and thrown out, at a modern accelerator experiment. But I'll leave that for an experimentalist to talk about. (I'm just a lowly theorist.)
Let's see... disks can be taken out by vibration, electric shock, or just by exceeding their shelf life. Failures can be from the drive electronics, the drive motor, a head crash, or simply getting unfiltered air in the drive. Life expectancy is 5 years if you get high-quality drives. They are typically kept online, which means they spend 100% of their lifetime vulnerable to threats from intruders, mistakes, and power surges.
On the other hand, tape has NONE of these flaws. Even if your tape drive explodes in a fireball from a lightning strike, the tapes will be fine, and can be read in another drive.
And yet here's a company trying to encourage people to migrate away from tape backups? Yes, they have a good offering -- remote backups. But that can (and should) be done with tape (possibly to disk first, but with a second copy on tape or other offline media).
Sorry to make the only on-topic post in the entire story, but I just couldn't handle the FUD.
Unfortunately they're part of university inventory. So "throw out" means the university takes them and stores them in a warehouse for 10 years or so (just in case anyone else happens to want to use them). After the 10 years, they probably crush them. It's silly, but I don't think I have any control over the process (or I'd take them home). About the best I can do is try to find another research group within the university that can use them. You know, the "free to a good home" type ads? With the constraint that the home has to be on-campus and accept responsibility for the inventory numbers.:P
You sure you're thinking about the O2? As another poster said, they weren't that expensive (though more expensive than the Indy, for sure). Too bad they're past their prime. We're about to throw out 4 of them to make space for more linux peecees.:(
Also, the 10-15 year time estimate seems off. When was the O2 introduced? Sometime around 1996?
Not trying to pick nits -- just curious if you were doing this on an Indigo or somesuch.
Uhh, X can cause heavy network traffic, especially if the user is browsing the web or something else that is graphics-intensive. If your CPU can handle the encryption at a high enough rate, it might as well be a desktop machine.
Of course, this is how I use windows now. It's just a cheap desktop that provides a web browser and an X server for my real work (all done on unix).
Just in case anyone actually reads this, I should make sure I was clear that I do NOT think multitasking is bad. It's just important for the individual to recognize when they've reached their limit on what they can efficiently handle. If you find yourself wasting time due to context switches, it's time to cut back. Until you reach that point, being able to background jobs is a big plus.
I've had to cut back recently. No IM at work. And IRC only when needed for tech support (rather than trying to continually monitor a dozen channels). I still read email compulsively at all hours but that's part of my job.
And, I'd have to say my attention span has probably suffered as a result. I find myself not even able to focus on a single/. article. Instead, I open all the interesting ones for the day in different tabs, and flip between them.
Now for the point -- even though I think it's great to multitask, it can sometime become too much to handle. So here's a recommendation for those who have too many context switches to get anything done: don't ever procrastinate anything. That will just mean another context switch each time you remember you need to do it. If you feel overwhelmed by the context switches, the trick is to actually complete some of the tasks.
Do these hypothetical thin clients support encryption? Because I don't particularly like the idea of people's passwords getting sent in plaintext between the thin client and the server.
We threw away a bunch of X terminals back in the late '90s for exactly that reason. Kinda scary to see it's being suggested in 2005.
Yeah... I actually looked at the guy's posting history, since it was so unexpected. I don't know... I think I kinda assumed that someone skript kiddie had stolen the password of someone with a low UID. Of course, it's also possible that/. isn't really decaying -- it's always been this bad. The real problem is that I read at +4 with -1 for "funny" (it never is) so I'm sheltered from most of the crap that gets posted here. Makes me more sensitive to it when I do see it. Kinda like how the 150 spams/day that spamassassin catches for me don't bother me, but the one or two it misses each day drive me batty. Before spamassassin, I was perfectly content to receive 5-10 spams/day. Funny the way your perspectives change.
Latency is measured in units of time. Lightyears are a measure of distance.
TCP's no good using standard broadcast methods
Huh? If I knew what you meant to say, it'd be easier to show you were wrong...
We need something that'll be as fast as fiber, but will stretch way way longer in distance.
So, like, line-of-sight laser communication?
Current radio's a broadcast. Can't do that, especially with package leakage.
How do you think we're communicating with the Mars rovers now? Or other planetary explorers?
I belive there was some experiments in quantum transmissio of data, in which an electron was split and one half sent to Munich, the other sent to Venice, and transmissions where near-instantaneous.
You can instantaneously determine what the other side received, but no information can be transmitted this way.
I see you have a low user-id, and therefore have learned to get modded up for saying stuff that is nonsensical and wrong. I must admit I'm impressed. I earn all my mod points the hard way.
"Information on the tapes was compressed, so viewing it would require special equipment, Kush said. It was not encrypted, she added."
It's worth noting that all tape drives are considered "special equipment", and compression isn't anything unusual. About the best thing they have going for them is that there's no label on the tape, so it's not obvious what goodies are on it.
For those saying the package was "lost", they're only partly correct. Actually, there were several tapes in a shipping box, which was damaged and had tapes fall out. Three of the four tapes that had fallen out have been recovered. The last one tape was lost, and we can only hope it was lost in a way that ensured its destruction.
We weren't seeing problems that severe (also using RHEL3 on a megaraid machine), though we did have to run a cron job to restart the megaraid daemon (for email notifications of drive failures). It had a habit of crashing after exactly 8 hours of execution. At least they've fixed that annoying bug.
In July 2004 we had a disk fail, and the entire partition got knocked offline. Not exactly what you'd expect from a raid5 array, is it? We had to rebuild the array offline (ie, system downtime) and then fsck (where there was plenty of corruption) to get it back up. We didn't lose anything important (hooray for tape backups) but it doesn't leave you with a warm fuzzy feeling when something like that happens.
I'd mod you up myself, but then people would just think you were just modded up for saying something that *might* be true. Instead, I'd rather say that it *is* true. We've been rather pissed off at the pitiful performance from our megaraid arrays. Stuff like 35M/s writes. You could almost beat that with a single disk. Sheesh.
For cheap setups, I go with 3ware. For more expensive ones, we use an external raid array with a scsi uplink to the computer. The cache, battery backup, and simplicity of host communication are important advantages to consider. Software raid, while fast and, just scares me.
Actually, factorial(x) is defined as gamma(x+1). And the gamma function is defined for negative values of x also. So computing the "factorial" of a negative number isn't really a problem.
Huh? Just to prove you wrong, I downloaded the latest kernel (2.6.11.7) and started a make -j on it. Top showed 80% usr, 20% sys. So io/wait was down under 1%. Cool part was the load average shooting up over 3000. Then top stopped refreshing, and I can't seem to ssh in anymore. Oops.
For the curious, this was on a dual opteron with 8 gig of ram.
Ah, doing a ^C got it back... here's top output on its way down:
Can't believe nobody's suggested just doing kernel compiles in a loop...?
I was about to, and saw your post when making sure I wouldn't be modded redundant. It's the best method I've found to stress-test your CPU and memory subsystem. In particular, do a make -j to compile in parallel, and use all your CPUs. Be careful, though, the 2.6 kernel will drive your load up to about 80 if you don't restrict it.
I've previously used the kernel-compile trick to test processor stability (when overclocking). Compile 100 kernels, do a diff/cmp on them, and see what percentage of the time your processor works. Of course, that doesn't work so well anymore, since now there are timestamps and other things compiled in. Oh well. Still, you can get a pretty good idea of whether things worked by comparing the size of the binary it produces. Not to mention the coolness factor of telling your friends you've compiled over 1000 kernels.;)
For generating heat, it would also be a good idea to beat on your hard drives a bit. I generally like to use bonnie or bonnie++ for that. Preferably multiple copies running simultaneously on different partitions.
Since you said these are blade servers, I'll assume we can ignore graphics-intensive ideas.
If all DNS records had 0 lifetime, the load on the core DNS servers would cause them to melt. Nice if you want a DDoS, not so nice if you want the internet to work.
Ever heard of a monoculture? It's dangerous. That's the primary reason Microsoft has so many security issues. To guard against this, the DNS infrastructure of the internet is intentionally made to be heterogeneous. They use different DNS software on different operating systems as much as possible.
Top security consultant? Doubtful. More likely an AC trying (and failing) to impersonate someone with a clue.
I don't consider myself to be a programmer, just a sysadmin. We use lots of open-source software on our machines, but don't really have funding to contribute to the projects. So, I give back in the form of answering questions on mailing lists, submitting patches, etc.
Recently, though, there was some functionality I wanted added to ClamAV, an open-source virus scanner. Basically, I wanted to make sure the milter was running. So, I wrote clmilter_watch, a tool to monitor the functionality of clamav-milter. Of course, I don't trust my own programming skills enough to know if it's stable for production use. So, it gets released to the world. A few downloads later, I get a couple of suggested patches, and the thing is pretty solid. Everyone wins.
Ok, this has me curious. Last time I experimented with breaking WEP, I broke a 128-bit key using aircrack in a few minutes. The catch was, it took a couple hundred meg of captured traffic (I initiated a ping flood to generate the data). So that meant taking hours to gather the data, not minutes. Anyone happen to know what changed since aircrack 2.1 (the most recent release)? Or was I just doing something wrong back then? Perhaps ping floods aren't a good way to generate the necessary data?
Your point would make sense if not for the fact that they said windows was scanned thousands of times, and redhat 9 not at all. Or maybe just nobody ever randomly picks the IP of a linux box? Yeah, that's it.
Maybe they think of attempts to ssh in as root and guess the password as attacks?
Slashdot Mirror
Rather than invoking any grand unified theory theories [sic], how about just reading what the AC said? Amazingly enough, this is a case where the AC is exactly right. The speed of light in diamond is slower than the speed of light in vacuum. 1.6 times slower, IIRC. So one might expect the speed of light in a diamond-fiber to be about 200,000 km/s.
That said, I have no idea WTF they're talking about with the 120 km/h stuff. $10 says they don't know either, or that it was a typo.
Or you could just learn to spell. ;)
They *do* reduce it on-site. That's why they "only" need to transfer 500TB/week. You'd be amazed by the amount of data that gets taken, and thrown out, at a modern accelerator experiment. But I'll leave that for an experimentalist to talk about. (I'm just a lowly theorist.)
On the other hand, tape has NONE of these flaws. Even if your tape drive explodes in a fireball from a lightning strike, the tapes will be fine, and can be read in another drive.
And yet here's a company trying to encourage people to migrate away from tape backups? Yes, they have a good offering -- remote backups. But that can (and should) be done with tape (possibly to disk first, but with a second copy on tape or other offline media).
Sorry to make the only on-topic post in the entire story, but I just couldn't handle the FUD.
Unfortunately they're part of university inventory. So "throw out" means the university takes them and stores them in a warehouse for 10 years or so (just in case anyone else happens to want to use them). After the 10 years, they probably crush them. It's silly, but I don't think I have any control over the process (or I'd take them home). About the best I can do is try to find another research group within the university that can use them. You know, the "free to a good home" type ads? With the constraint that the home has to be on-campus and accept responsibility for the inventory numbers. :P
Also, the 10-15 year time estimate seems off. When was the O2 introduced? Sometime around 1996?
Not trying to pick nits -- just curious if you were doing this on an Indigo or somesuch.
Of course, this is how I use windows now. It's just a cheap desktop that provides a web browser and an X server for my real work (all done on unix).
Just in case anyone actually reads this, I should make sure I was clear that I do NOT think multitasking is bad. It's just important for the individual to recognize when they've reached their limit on what they can efficiently handle. If you find yourself wasting time due to context switches, it's time to cut back. Until you reach that point, being able to background jobs is a big plus.
And, I'd have to say my attention span has probably suffered as a result. I find myself not even able to focus on a single /. article. Instead, I open all the interesting ones for the day in different tabs, and flip between them.
Now for the point -- even though I think it's great to multitask, it can sometime become too much to handle. So here's a recommendation for those who have too many context switches to get anything done: don't ever procrastinate anything. That will just mean another context switch each time you remember you need to do it. If you feel overwhelmed by the context switches, the trick is to actually complete some of the tasks.
We threw away a bunch of X terminals back in the late '90s for exactly that reason. Kinda scary to see it's being suggested in 2005.
Hmmm, I'm rambling. Must be getting old.
Latency is measured in units of time. Lightyears are a measure of distance.
TCP's no good using standard broadcast methods
Huh? If I knew what you meant to say, it'd be easier to show you were wrong...
We need something that'll be as fast as fiber, but will stretch way way longer in distance.
So, like, line-of-sight laser communication?
Current radio's a broadcast. Can't do that, especially with package leakage.
How do you think we're communicating with the Mars rovers now? Or other planetary explorers?
I belive there was some experiments in quantum transmissio of data, in which an electron was split and one half sent to Munich, the other sent to Venice, and transmissions where near-instantaneous.
You can instantaneously determine what the other side received, but no information can be transmitted this way.
I see you have a low user-id, and therefore have learned to get modded up for saying stuff that is nonsensical and wrong. I must admit I'm impressed. I earn all my mod points the hard way.
"Information on the tapes was compressed, so viewing it would require special equipment, Kush said. It was not encrypted, she added."
It's worth noting that all tape drives are considered "special equipment", and compression isn't anything unusual. About the best thing they have going for them is that there's no label on the tape, so it's not obvious what goodies are on it.
For those saying the package was "lost", they're only partly correct. Actually, there were several tapes in a shipping box, which was damaged and had tapes fall out. Three of the four tapes that had fallen out have been recovered. The last one tape was lost, and we can only hope it was lost in a way that ensured its destruction.
Well, at least the codes are right... not sure about the colors, though. It's been a few years. :P
In July 2004 we had a disk fail, and the entire partition got knocked offline. Not exactly what you'd expect from a raid5 array, is it? We had to rebuild the array offline (ie, system downtime) and then fsck (where there was plenty of corruption) to get it back up. We didn't lose anything important (hooray for tape backups) but it doesn't leave you with a warm fuzzy feeling when something like that happens.
For cheap setups, I go with 3ware. For more expensive ones, we use an external raid array with a scsi uplink to the computer. The cache, battery backup, and simplicity of host communication are important advantages to consider. Software raid, while fast and, just scares me.
Yes... though it blows up for non-positive integers. See mathworld for details.
Actually, factorial(x) is defined as gamma(x+1). And the gamma function is defined for negative values of x also. So computing the "factorial" of a negative number isn't really a problem.
For the curious, this was on a dual opteron with 8 gig of ram.
Ah, doing a ^C got it back... here's top output on its way down:
Gotta love killing machines.
I was about to, and saw your post when making sure I wouldn't be modded redundant. It's the best method I've found to stress-test your CPU and memory subsystem. In particular, do a make -j to compile in parallel, and use all your CPUs. Be careful, though, the 2.6 kernel will drive your load up to about 80 if you don't restrict it.
I've previously used the kernel-compile trick to test processor stability (when overclocking). Compile 100 kernels, do a diff/cmp on them, and see what percentage of the time your processor works. Of course, that doesn't work so well anymore, since now there are timestamps and other things compiled in. Oh well. Still, you can get a pretty good idea of whether things worked by comparing the size of the binary it produces. Not to mention the coolness factor of telling your friends you've compiled over 1000 kernels. ;)
For generating heat, it would also be a good idea to beat on your hard drives a bit. I generally like to use bonnie or bonnie++ for that. Preferably multiple copies running simultaneously on different partitions.
Since you said these are blade servers, I'll assume we can ignore graphics-intensive ideas.
Ever heard of a monoculture? It's dangerous. That's the primary reason Microsoft has so many security issues. To guard against this, the DNS infrastructure of the internet is intentionally made to be heterogeneous. They use different DNS software on different operating systems as much as possible.
Top security consultant? Doubtful. More likely an AC trying (and failing) to impersonate someone with a clue.
Recently, though, there was some functionality I wanted added to ClamAV, an open-source virus scanner. Basically, I wanted to make sure the milter was running. So, I wrote clmilter_watch, a tool to monitor the functionality of clamav-milter. Of course, I don't trust my own programming skills enough to know if it's stable for production use. So, it gets released to the world. A few downloads later, I get a couple of suggested patches, and the thing is pretty solid. Everyone wins.
Ok, this has me curious. Last time I experimented with breaking WEP, I broke a 128-bit key using aircrack in a few minutes. The catch was, it took a couple hundred meg of captured traffic (I initiated a ping flood to generate the data). So that meant taking hours to gather the data, not minutes. Anyone happen to know what changed since aircrack 2.1 (the most recent release)? Or was I just doing something wrong back then? Perhaps ping floods aren't a good way to generate the necessary data?
Uhh, 1/r^4 might be correct for 5-dimensional space. But in the 3 (large) spatial dimensions, it's just 1/r^2.
Maybe they think of attempts to ssh in as root and guess the password as attacks?