Actually, yes it should, as long as you also set your bit torrent to use a maximum of download bandwidth, and report as choked to the "supplier".
My parent was talking about QoS at the router, not at the application. There's a difference. In any event, bit torrent is one protocol. The original question was more general and mentions "other high-bandwidth applications." Most protocols have no BCN (backwards congestion notification) nor is there any link layer method for allocating bandwidth a la fibre channel, which uses buffer credits. TCP is an inherently greedy protocol. It takes all the bandwidth it can until it starts dropping packets and then backs off a bit until it does the same thing over and over again. This does not play well with VoIP and the only way to truly control it is end to end QoS, like what is done in the enterprise.
No, it shouldn't. QoS only works on egress. You can't do any meaningful ingress QoS. All you can do is stop packets from coming past the router. That doesn't address the real problem which is that the ISP link is maxed out. In fact, if you start dropping TCP data that's in a lower priority queue than your UDP voice, it will cause even further issues as the sender will retransmit those lost TCP packets.
But doesn't all that convertion from AC to DC create heat? Wouldn't a rack of servers (equipment) be far more efficient, power-wise, if they all just accepted the DC direct from the rack?
You can order servers and networking equipment with a DC power input instead of AC input. It poses other problems though. In most situations, AC is still more efficient. Such DC systems are 48V, whereas AC in a datacenter is 200-240V. Roughly speaking, to obtain the same wattage with 48V DC, you need 4 times the current. Current carrying capacity is directly related to what gauge copper you can use. More current means you need more copper, and it's typically not worth the uptick in cost.
Someone needs to sit down with some of these people and explain to them what the GPL actually says. It doesn't require software written to run on Linux to be GPL'd. Even if you had some reason why you wanted to modify the Linux kernel itself (and why the hell would a Wall Street firm want to!?), you wouldn't need to GPL your modifications unless you were turning around and selling or distributing the modified version publicly.
Maybe 2nd tier Wall St. firms need help. 1st tier do not. They are already using Linux in full force. And to be perfectly honest, kernel modifications are required. Whether a Wall St. coder or a Red Hat coder is writing the change is irrelevant. Wall St. drives a lot of kernel changes. Trust me.
Wall Street has always been home to some of Sun's and IBM's largest corporate accounts. I don't doubt Linux and/or BSD can do the job that Solaris can in some cases (with caveats), but it will take years for that to happen. A "Linux stronghold" is misleading at best, TFA doesn't even support the claim.
I can say, without a doubt, Wall St. is a Linux stronghold. Buying Sun hardware is not popular as it used to be. Linux on blades or Linux on VMWare ESX on blades is becoming the most common solution.
If only the US had responsible taxes on fuel to begin with this wouldn't have been a problem in the first place.
Taxes? Come on. I can't believe anyone can honestly make a statement that we need more taxes. It really astounds me. Let me give you a bit of wisdom. Government does not exist to help you. Period. End of story. If it seems like they're helping you, look a bit closer, because they are *most certainly* screwing you.
yeah, swell, let's have amazon's servers encrypt all traffic, that'll reduce their server load.
Most likely they're using something like a NetScaler to do load balancing. In the case of NetScaler, they have built in hardware to do the public key encryption portion of SSL (which is the expensive part). Symmetric encryption, be it Triple-DES or AES is very easy to do in CPU.
If these algorithms actually worked, why would you need to be working for somebody else?
I hate to reply twice to the same post, but even if you have great algorithms, they're not the sort of thing you can put into place in your basement to get any sort of real gain. You get into issues regarding processing power and latency. For instance, the speed of light is pretty slow. It's slow enough that you can't have servers too far from an exchange to make use of some sorts of algorithms. No one ever said this stuff is easy; it requires massive infrastructure that typically only corporations can manage. In addition to the financial mathematics talent, it requires expertise in implementing the algorithms in code, and expertise in building networks that can support low latency and lots of resiliency. It's not exactly something one person can manage. Quantitative trading isn't as profitable as it was in its infancy, but it still works and still makes money.
If these algorithms actually worked, why would you need to be working for somebody else?
I never said I was creating these algorithms. While that could be fun and interesting, I only have a recreational interest in financial mathematics. Those that actually work on methods to make money in this manner typically have PhDs in varying fields (engineering, math, physics).
I said that this sort of thing pays my bills because I work for a Wall St. firm that makes some of its money doing this.
The best you can do is make a psuedo-ai that can make guesses based on data. And again, no one's made a computer good enough at guessing that it makes money.
Hahahahahahahahahaha. PLEASE keep thinking that. How do you think companies like D.E. Shaw & Co. exist? Not to mention Goldman Sachs, etc.
The reason I get a paycheck twice is month, in part, is because you can create efficient algorithms to make money in financial markets. But please don't let that dissuade you from your obviously very informed opinion.
My Leopard install is showing "OpenSSL 0.9.7l 28 Sep 2006" while my Debian machine is showing "OpenSSL 0.9.8g 19 Oct 2007". I imagine there might be a few bugs there, and it's late enough that it wouldn't have been released close enough to be included in 10.5.0.
Actually, the bug there is in Debian, not OS X. It's rather serious, too, so you might want to check it out. If your machine is publically exposed and running SSH, it might have already been rooted.
It would be even nicer if you went and looked for the facts yourself instead of expecting them to be handed to you on a silver platter.
If you write any sort of research paper, you can't go and write, "and I'll leave it as an exercise to the reader to confirm all my supposed 'facts.'" There is an accepted procedure that the burden of proof of claiming anything lies on he who claims it, at least in professional and academic circles.
will they be competitive with mid range priced hard drives? You can get 500GB for $100 these days.
In a few years. Right now SSDs perform incredibly in terms of IOPS (I/O operations Per Second) that enterprise storage type folks are eying them longingly. They just need a little bit more space for the money. Until such time, it's very possible that we'll see more in terms of using SSDs as caching components in front of more antiquated spinning media.
This is a very good exposition on why minimum wage is bad. I usually just explain it with the fact that even though it's hard to believe, some people are just not skilled enough to deserve minimum wage for the job they do. This realization is easy to experience when dealing with some service workers.
The right way to delivery a symmetric keys is using asymmetric key, like RSA. Where no hand delivery is needed and is very secure.
Uh, only if you have public key infrastructure (i.e. pre-trusted authorities). I can generate shared secrets all day long with Diffie Hellman, but it really only helps me if I know that the recipient is not a man in the middle.
(but, apply to D.E. Shaw Research on the off chance you can get a job there, they're building a supercomputer for this)
Quite simply put, do this. You probably would have more luck winning the lottery (literally), but it's worth a shot. Take a look. If you haven't heard of D.E. Shaw before, they're a hedge fund founded by former computer science professor David Shaw and are universally respected as one of the best hedge funds around.
What should be interesting are the results of some of their "risky" network algorithm setups - removing, for instance, acknowledgments (ACK packets) from each individual sent packet.
Well this is easy, and it can be done on the existing Internet. Use UDP.
Internet2 was never supposed to replace "The Internet." It's best thought of as a collection of private peering and transit agreements between research institutions.
This is an interesting argument, but really -- what are you getting with that over-the-top equipment? Bizarre drugs with side effects that kill you. A lot can be achieved without that fancy gear.
Then forget about doing meaningful research on viruses. After all, you can't see them in enough detail without an electron microscope.
I completely understand your point. My point is that adding disks increases the likelihood of a single drive failing, or in other words, a data loss event. That may or may not be acceptable.
I always looked at it this way: If you have one really nice/fast drive and it fails, you *still* lose everything you had. I'd rather spend the same (or less) cash on two slightly smaller, slower drives and throw them into an array...
Yes but the point is that with N drives striped without parity (i.e. RAID 0), you increase your probability of disastrous failure proportionally to N.
Then I asked the grad students lots of questions because the professors were not available much. They are very busy.
Here's where I'll disagree. Yes, professors are not available much. No, they're not very busy. Professors with tenure have very easy lives. They're professors for a reason. They could go make more money in industry if they're half decent, but they choose not to because staying at a university affords them the flexibility of not showing up on a daily basis and working on what they want when they want. So yea, they're hard to find, but not necessarily because they're huddled away doing tons of work. I didn't fully appreciate this contrast until I started working on Wall St. Only then did I fully understand what "busy" means.
Slashdot Mirror
Actually, yes it should, as long as you also set your bit torrent to use a maximum of download bandwidth, and report as choked to the "supplier".
My parent was talking about QoS at the router, not at the application. There's a difference. In any event, bit torrent is one protocol. The original question was more general and mentions "other high-bandwidth applications." Most protocols have no BCN (backwards congestion notification) nor is there any link layer method for allocating bandwidth a la fibre channel, which uses buffer credits. TCP is an inherently greedy protocol. It takes all the bandwidth it can until it starts dropping packets and then backs off a bit until it does the same thing over and over again. This does not play well with VoIP and the only way to truly control it is end to end QoS, like what is done in the enterprise.
QOS should work if you set it up properly.
No, it shouldn't. QoS only works on egress. You can't do any meaningful ingress QoS. All you can do is stop packets from coming past the router. That doesn't address the real problem which is that the ISP link is maxed out. In fact, if you start dropping TCP data that's in a lower priority queue than your UDP voice, it will cause even further issues as the sender will retransmit those lost TCP packets.
But doesn't all that convertion from AC to DC create heat? Wouldn't a rack of servers (equipment) be far more efficient, power-wise, if they all just accepted the DC direct from the rack?
You can order servers and networking equipment with a DC power input instead of AC input. It poses other problems though. In most situations, AC is still more efficient. Such DC systems are 48V, whereas AC in a datacenter is 200-240V. Roughly speaking, to obtain the same wattage with 48V DC, you need 4 times the current. Current carrying capacity is directly related to what gauge copper you can use. More current means you need more copper, and it's typically not worth the uptick in cost.
Someone needs to sit down with some of these people and explain to them what the GPL actually says. It doesn't require software written to run on Linux to be GPL'd. Even if you had some reason why you wanted to modify the Linux kernel itself (and why the hell would a Wall Street firm want to!?), you wouldn't need to GPL your modifications unless you were turning around and selling or distributing the modified version publicly.
Maybe 2nd tier Wall St. firms need help. 1st tier do not. They are already using Linux in full force. And to be perfectly honest, kernel modifications are required. Whether a Wall St. coder or a Red Hat coder is writing the change is irrelevant. Wall St. drives a lot of kernel changes. Trust me.
Wall Street has always been home to some of Sun's and IBM's largest corporate accounts. I don't doubt Linux and/or BSD can do the job that Solaris can in some cases (with caveats), but it will take years for that to happen. A "Linux stronghold" is misleading at best, TFA doesn't even support the claim.
I can say, without a doubt, Wall St. is a Linux stronghold. Buying Sun hardware is not popular as it used to be. Linux on blades or Linux on VMWare ESX on blades is becoming the most common solution.
In a world where most of the inhabitants are irresponsible, arrogant and self-centered assholes, it just doesn't work that well.
Two words: Invisible Hand. The market is doing just what it's supposed to do.
If only the US had responsible taxes on fuel to begin with this wouldn't have been a problem in the first place.
Taxes? Come on. I can't believe anyone can honestly make a statement that we need more taxes. It really astounds me. Let me give you a bit of wisdom. Government does not exist to help you. Period. End of story. If it seems like they're helping you, look a bit closer, because they are *most certainly* screwing you.
If you disagree, please provide at least two examples.
Two examples:
http://www.physics.nyu.edu/faculty/sokal/lingua_franca_v4/lingua_franca_v4.html
http://www.theregister.co.uk/2002/11/01/physics_hoaxers_discover_quantum_bogosity/
yeah, swell, let's have amazon's servers encrypt all traffic, that'll reduce their server load.
Most likely they're using something like a NetScaler to do load balancing. In the case of NetScaler, they have built in hardware to do the public key encryption portion of SSL (which is the expensive part). Symmetric encryption, be it Triple-DES or AES is very easy to do in CPU.
If these algorithms actually worked, why would you need to be working for somebody else?
I hate to reply twice to the same post, but even if you have great algorithms, they're not the sort of thing you can put into place in your basement to get any sort of real gain. You get into issues regarding processing power and latency. For instance, the speed of light is pretty slow. It's slow enough that you can't have servers too far from an exchange to make use of some sorts of algorithms. No one ever said this stuff is easy; it requires massive infrastructure that typically only corporations can manage. In addition to the financial mathematics talent, it requires expertise in implementing the algorithms in code, and expertise in building networks that can support low latency and lots of resiliency. It's not exactly something one person can manage. Quantitative trading isn't as profitable as it was in its infancy, but it still works and still makes money.
If these algorithms actually worked, why would you need to be working for somebody else?
I never said I was creating these algorithms. While that could be fun and interesting, I only have a recreational interest in financial mathematics. Those that actually work on methods to make money in this manner typically have PhDs in varying fields (engineering, math, physics).
I said that this sort of thing pays my bills because I work for a Wall St. firm that makes some of its money doing this.
The best you can do is make a psuedo-ai that can make guesses based on data. And again, no one's made a computer good enough at guessing that it makes money.
Hahahahahahahahahaha. PLEASE keep thinking that. How do you think companies like D.E. Shaw & Co. exist? Not to mention Goldman Sachs, etc.
The reason I get a paycheck twice is month, in part, is because you can create efficient algorithms to make money in financial markets. But please don't let that dissuade you from your obviously very informed opinion.
This may be the first free source for real-time quotes.
No. There you are.
My Leopard install is showing "OpenSSL 0.9.7l 28 Sep 2006" while my Debian machine is showing "OpenSSL 0.9.8g 19 Oct 2007". I imagine there might be a few bugs there, and it's late enough that it wouldn't have been released close enough to be included in 10.5.0.
Actually, the bug there is in Debian, not OS X. It's rather serious, too, so you might want to check it out. If your machine is publically exposed and running SSH, it might have already been rooted.
It would be even nicer if you went and looked for the facts yourself instead of expecting them to be handed to you on a silver platter.
If you write any sort of research paper, you can't go and write, "and I'll leave it as an exercise to the reader to confirm all my supposed 'facts.'" There is an accepted procedure that the burden of proof of claiming anything lies on he who claims it, at least in professional and academic circles.
will they be competitive with mid range priced hard drives? You can get 500GB for $100 these days.
In a few years. Right now SSDs perform incredibly in terms of IOPS (I/O operations Per Second) that enterprise storage type folks are eying them longingly. They just need a little bit more space for the money. Until such time, it's very possible that we'll see more in terms of using SSDs as caching components in front of more antiquated spinning media.
This is a very good exposition on why minimum wage is bad. I usually just explain it with the fact that even though it's hard to believe, some people are just not skilled enough to deserve minimum wage for the job they do. This realization is easy to experience when dealing with some service workers.
The right way to delivery a symmetric keys is using asymmetric key, like RSA. Where no hand delivery is needed and is very secure.
Uh, only if you have public key infrastructure (i.e. pre-trusted authorities). I can generate shared secrets all day long with Diffie Hellman, but it really only helps me if I know that the recipient is not a man in the middle.
(but, apply to D.E. Shaw Research on the off chance you can get a job there, they're building a supercomputer for this)
Quite simply put, do this. You probably would have more luck winning the lottery (literally), but it's worth a shot. Take a look. If you haven't heard of D.E. Shaw before, they're a hedge fund founded by former computer science professor David Shaw and are universally respected as one of the best hedge funds around.
What should be interesting are the results of some of their "risky" network algorithm setups - removing, for instance, acknowledgments (ACK packets) from each individual sent packet.
Well this is easy, and it can be done on the existing Internet. Use UDP.
Internet2 was never supposed to replace "The Internet." It's best thought of as a collection of private peering and transit agreements between research institutions.
Are you also amongst the group of people that think Extended Validation certificates are anything more than something to make Verisign more money?
This is an interesting argument, but really -- what are you getting with that over-the-top equipment? Bizarre drugs with side effects that kill you. A lot can be achieved without that fancy gear.
Then forget about doing meaningful research on viruses. After all, you can't see them in enough detail without an electron microscope.
You miss the point of my post.
I completely understand your point. My point is that adding disks increases the likelihood of a single drive failing, or in other words, a data loss event. That may or may not be acceptable.
I always looked at it this way: If you have one really nice/fast drive and it fails, you *still* lose everything you had. I'd rather spend the same (or less) cash on two slightly smaller, slower drives and throw them into an array...
Yes but the point is that with N drives striped without parity (i.e. RAID 0), you increase your probability of disastrous failure proportionally to N.
Then I asked the grad students lots of questions because the professors were not available much. They are very busy.
Here's where I'll disagree. Yes, professors are not available much. No, they're not very busy. Professors with tenure have very easy lives. They're professors for a reason. They could go make more money in industry if they're half decent, but they choose not to because staying at a university affords them the flexibility of not showing up on a daily basis and working on what they want when they want. So yea, they're hard to find, but not necessarily because they're huddled away doing tons of work. I didn't fully appreciate this contrast until I started working on Wall St. Only then did I fully understand what "busy" means.