In rural USA, only fiber could do that, which will take a good part of a decade to upgrade to. Installing fiber is not nearly as expensive as many make it out to be, but we don't need to create an artificial employment bubble and waste money trying to roll fiber out faster. If we were installing it back in 2000, we'd be done by now, and everyone would have access to 1gb.
As for free space twisted light, 32gb is slow: http://spectrum.ieee.org/tech-... Jun 2012 "Scientist in California and Israel say they've transmitted data through the air at a rate of 2.56 terabits per second using beams of "twisted light.""
Some locking primitives map directly to instructions. I've done my own lock--free designs or even made my own "locks" with only basic atomic memory access instructions. I won't claim my designs are as well polished, but they were quickly made and knowing the more polished version would have helped. Helps to appreciate the minute differences between sync designs when you create your own.
Say a global company makes 9bil, but 3bil of that is from your country. Then assume 33% of their net pre-tax profit is to be paid in taxes, but have taxes paid to other countries as tax deductible. So the only real way for a company to get out of paying your country taxes is to pay taxes in another country. Also, assume that all forms of direct government subsidies a "tax deduction" from other countries. In case a government tries to "feed back" money paid to effectively reduce taxation.
The government's budget is planned spending. Because the military is "critical", they get to use huge amounts of "unplanned" spending. Back when I was in college, we went over actual spending, and based on average prices, the amount of money we spent on the entire military over a 5 year period was enough to cover everyone in the USA with free college and 100% coverage healthcare for 10 years. Actual military spending is, or at least was, more than social security, but if you look at the budget, social security is much more.
Depends on your customers. If you're a transit provider and your customer has an SLA that states 100% uptime and 1ms jitter and your insecure routing causes the route to become longer and the jitter goes above 1ms, suddenly you're paying your customer for not meeting the SLA.
You're just talking about BGP, which is done in software. A quick update will allow nearly all hardware that uses BGP to support the new protocol, assuming the code is small enough to fit in the firmware.
And what do you mean by edge routers? You mean the last mile or for peering? My ISP pays Level 3 to handle peering. If you're talking about last mile, then your ISP should have invested into fiber, which is easily and cheaply upgraded. At $100/port for a 500-1gb port chassis that can support 3tb/s, it's not that expensive. How long does it take to pay off $100? Actually, network equipment represents about 40% of an ISP's costs, the bulk of the cost is in customer support. Phone centers are expensive with an average cost of $1/minute that a customer is connected. A single truck roll can cost an ISP much much more.
Modern SSDs even move around data that isn't changing in order to keep an even wear. Assuming you're a normal user where most of your data doesn't change, then a 2x increase in storage with a 10% reduction in durability is a net gain. The issue of "wear" has technically already been solved for flash, but they have to figure out how to mass produce the changes. Flash is already getting replaced with mram soon. Several companies have system memory with mram slated within a few years, they've already retooled several plants.
HP has gone one step further and is creating a dynamically allocated mram system that works as both system memory and data storage, so your harddrive and memory is all from the same pool. This reduces power usage dramatically and increases performance dramatically. At least in their own load test, they've gotten about an 8x reduction in datacenter power usage and almost a 2x increase in average workload throughput.
They're currently working on custom Linux kernels that can dynamically allocate memory and storage instead of having to partition the pool between the two. A cool side effect is that "memory mapped files" are literally in memory all the time as storage is memory.
2. Somebody had to discover/invent those proper ways of solving those problems.
Nearly all solutions are simple given a well defined problem. I tend to "reinvent" solutions all the time. When I started playing with multi-threading, I was toying around with lots of different ways to handle locking and trying to safely handle sync without locks. Seems everything that I discovered on my own has already been done before, typically decades ago back in the 60s and 70s, but it doesn't mean I had to have someone else "discover" it for me. The solutions were blindingly obvious for anyone who spent a few hours thinking about the issues.
I have a BS, but in my few literature classes, we did a lot of critical thinking. Pretty much all of my 100 level classes emphasized a lot of critical thinking. The one thing that I did like more about critical thinking in such classes is you have to play devils advocate more and work with a lot more hypothetical situations that are "unnatural".
When doing critical thinking for stuff like programming, there is few "proper" ways of solving an issue, but in the more "arts" kind of classes, things were more open to debate or you had to identify an weigh different opinions. That requires more flexibility. I think both ways of critical thinking are important.
In 20 years data transmission will be 1000x where it is today.
Worse than that. Fiber tech has improved commercially available photonics by 1,000x in just the past 3 years. You should see the 16tb/fiber tech that uses a fraction of the power and has ranges nearing 1,000km without any signal regeneration. In just the past year, they have shown you can use off-the-shelf cheap parts to create 40gb/s interfaces for below $100/port, while getting over 10km ranges. In the next 8 years, data transmissions will be capable over more than 1,000x. Google fiber will look slow.
Currently, the fastest Core routers being sold can handle nearly 1pb/s, and support future 1tb/s line cards with a 2tb/s interface. but if you enable QoS on those 400gb ports, they slow down to about 150gb/s because they can't QoS line rate.
Minnesota has 1gb fiber to the cabin and many years back had 1gb fiber rolled out in an 30,000 square mile area with an average density of below 2 houses per square mile.
What's your excuse now?
In my town, my friend who is about 30 minutes outside the city limit has 1gb fiber, all under ground trenched fiber. I don't live in Minnesota, but I am close by.
Before you claim government subsidies, my ISP has turned down all government subsidies and is a small local ISP, did everything on its own dime, not to mention they use Level 3, which is one of the most expensive transit providers and all of my bandwidth is dedicated. I get my full 50/50 24/7 to nearly everywhere in the USA and Europe.
Most hardware disabled QOS when the link is below 80% utilization. QOS only works when there is congestion, otherwise it not only does nothing, but it adds more overhead. Many high end network devices cannot do QOS at line rate, and I'm not talking about 1gb, 10gb, or even 100gb ports. Those are low-end now days. It's those 400gb+ ports that start to have real issues doing QOS.
I pay $100/month for a 50/50 dedicated fiber connection which is entirely supplied by Level 3 for upstream. While my bandwidth is not "guaranteed", it is marketed as "dedicated" and I have called in at 2am complaining about 30ms pings to servers I normally get 8ms to, and they had a line tech out at my house at 8am that morning.
This is a "standard" residential connection around here. They don't differentiate between business and home users. Everyone gets the same packages offered and the same support. Mind you, they do have "enterprise" packages, and those are the "expensive" ones you're referring to.
Another time I was only getting 42mb/s on my speed test, and I have a 50mb connection. They got me in touch with an engineer and the problem was resolved on their end. They were in the process of upgrading from 1gb ports to 10gb for their line cards and my linecard's port was nearing saturation. They changed which linecard I was plugged into, it took them about 5 minutes and I had to power cycle my ONT.
I have been running a ping every 0.5sec to a datacenter in Chicago that peers with Level 3, which is 8 hops and 200+ miles from me, and I'm averaging a daily 0.005% packet-loss and sub 1ms of jitter.
You may want to ask what you're paying for. Midwest USA, ftw!
Yes, lets use the example of a bulk data transfer getting priority over low bandwidth latency sensitive gaming. Wait, doesn't priority only matter when there is congestion? Why is there congestion? Are they saying there is congestion on their internal network or do they mean they have congestion in the links to the other networks? If the congestion is in the external links, how do they propose "priority" to affect unrelated links?
How about they just deliver what they sell? F*ck "up to", they need to get rid of that legal loop hole. Our car gets "up to" 40mpg, but you'll never see more than 1mpg in normal usage.
You can detect corruption and it would probably cause a completely loss of network connection, which is still better than complete loss of some data and hardware.
Depends on your definition of "completely". It is at the lowest ice coverage in recorded history. The Antarctic on the other hand, which is what this article is about, has been gaining ice, but at a lesser rate than the Arctic has been losing it.
They'll just send you to collections where you can pay $10/month for the next 5 years. I've found collections easier to work with and typically reduce the amount owed. I prefer to work with collections. I've dragged some collections for several thousand dollars out with $10/month for several years until they contacted me and said they'll remove 20% of the amount remaining if I pay at all once and that's after they already removed 10% when they're were trying to get me to pay more per month.
Don't think I was trying to skirt paying, I had other more pressing bills and a lot of health issues during that time of my life.
Tellers at a bank could be completely rep[laced with ATMs or the like. You don't need to trust people much when their jobs are pointless. When you depend on someone's skill, and you don't trust them, that's when you have issues.
People don't work well as a team when there is no trust.
Slashdot Mirror
While we're at it, lets get rid of public education and public police and public courts.
In rural USA, only fiber could do that, which will take a good part of a decade to upgrade to. Installing fiber is not nearly as expensive as many make it out to be, but we don't need to create an artificial employment bubble and waste money trying to roll fiber out faster. If we were installing it back in 2000, we'd be done by now, and everyone would have access to 1gb.
As for free space twisted light, 32gb is slow: http://spectrum.ieee.org/tech-... Jun 2012 "Scientist in California and Israel say they've transmitted data through the air at a rate of 2.56 terabits per second using beams of "twisted light.""
Some locking primitives map directly to instructions. I've done my own lock--free designs or even made my own "locks" with only basic atomic memory access instructions. I won't claim my designs are as well polished, but they were quickly made and knowing the more polished version would have helped. Helps to appreciate the minute differences between sync designs when you create your own.
Say a global company makes 9bil, but 3bil of that is from your country. Then assume 33% of their net pre-tax profit is to be paid in taxes, but have taxes paid to other countries as tax deductible. So the only real way for a company to get out of paying your country taxes is to pay taxes in another country. Also, assume that all forms of direct government subsidies a "tax deduction" from other countries. In case a government tries to "feed back" money paid to effectively reduce taxation.
The government's budget is planned spending. Because the military is "critical", they get to use huge amounts of "unplanned" spending. Back when I was in college, we went over actual spending, and based on average prices, the amount of money we spent on the entire military over a 5 year period was enough to cover everyone in the USA with free college and 100% coverage healthcare for 10 years. Actual military spending is, or at least was, more than social security, but if you look at the budget, social security is much more.
There's a lot of loopholes.
Depends on your customers. If you're a transit provider and your customer has an SLA that states 100% uptime and 1ms jitter and your insecure routing causes the route to become longer and the jitter goes above 1ms, suddenly you're paying your customer for not meeting the SLA.
You're just talking about BGP, which is done in software. A quick update will allow nearly all hardware that uses BGP to support the new protocol, assuming the code is small enough to fit in the firmware.
And what do you mean by edge routers? You mean the last mile or for peering? My ISP pays Level 3 to handle peering. If you're talking about last mile, then your ISP should have invested into fiber, which is easily and cheaply upgraded. At $100/port for a 500-1gb port chassis that can support 3tb/s, it's not that expensive. How long does it take to pay off $100? Actually, network equipment represents about 40% of an ISP's costs, the bulk of the cost is in customer support. Phone centers are expensive with an average cost of $1/minute that a customer is connected. A single truck roll can cost an ISP much much more.
Yeah, thanks Verizon. They just had to suddenly public their huge number of internal routes and crash the Internet.
Modern SSDs even move around data that isn't changing in order to keep an even wear. Assuming you're a normal user where most of your data doesn't change, then a 2x increase in storage with a 10% reduction in durability is a net gain. The issue of "wear" has technically already been solved for flash, but they have to figure out how to mass produce the changes. Flash is already getting replaced with mram soon. Several companies have system memory with mram slated within a few years, they've already retooled several plants.
HP has gone one step further and is creating a dynamically allocated mram system that works as both system memory and data storage, so your harddrive and memory is all from the same pool. This reduces power usage dramatically and increases performance dramatically. At least in their own load test, they've gotten about an 8x reduction in datacenter power usage and almost a 2x increase in average workload throughput.
They're currently working on custom Linux kernels that can dynamically allocate memory and storage instead of having to partition the pool between the two. A cool side effect is that "memory mapped files" are literally in memory all the time as storage is memory.
2. Somebody had to discover/invent those proper ways of solving those problems.
Nearly all solutions are simple given a well defined problem. I tend to "reinvent" solutions all the time. When I started playing with multi-threading, I was toying around with lots of different ways to handle locking and trying to safely handle sync without locks. Seems everything that I discovered on my own has already been done before, typically decades ago back in the 60s and 70s, but it doesn't mean I had to have someone else "discover" it for me. The solutions were blindingly obvious for anyone who spent a few hours thinking about the issues.
I have a BS, but in my few literature classes, we did a lot of critical thinking. Pretty much all of my 100 level classes emphasized a lot of critical thinking. The one thing that I did like more about critical thinking in such classes is you have to play devils advocate more and work with a lot more hypothetical situations that are "unnatural".
When doing critical thinking for stuff like programming, there is few "proper" ways of solving an issue, but in the more "arts" kind of classes, things were more open to debate or you had to identify an weigh different opinions. That requires more flexibility. I think both ways of critical thinking are important.
In 20 years data transmission will be 1000x where it is today.
Worse than that. Fiber tech has improved commercially available photonics by 1,000x in just the past 3 years. You should see the 16tb/fiber tech that uses a fraction of the power and has ranges nearing 1,000km without any signal regeneration. In just the past year, they have shown you can use off-the-shelf cheap parts to create 40gb/s interfaces for below $100/port, while getting over 10km ranges. In the next 8 years, data transmissions will be capable over more than 1,000x. Google fiber will look slow.
Currently, the fastest Core routers being sold can handle nearly 1pb/s, and support future 1tb/s line cards with a 2tb/s interface. but if you enable QoS on those 400gb ports, they slow down to about 150gb/s because they can't QoS line rate.
Minnesota has 1gb fiber to the cabin and many years back had 1gb fiber rolled out in an 30,000 square mile area with an average density of below 2 houses per square mile.
What's your excuse now?
In my town, my friend who is about 30 minutes outside the city limit has 1gb fiber, all under ground trenched fiber. I don't live in Minnesota, but I am close by.
Before you claim government subsidies, my ISP has turned down all government subsidies and is a small local ISP, did everything on its own dime, not to mention they use Level 3, which is one of the most expensive transit providers and all of my bandwidth is dedicated. I get my full 50/50 24/7 to nearly everywhere in the USA and Europe.
Most hardware disabled QOS when the link is below 80% utilization. QOS only works when there is congestion, otherwise it not only does nothing, but it adds more overhead. Many high end network devices cannot do QOS at line rate, and I'm not talking about 1gb, 10gb, or even 100gb ports. Those are low-end now days. It's those 400gb+ ports that start to have real issues doing QOS.
I pay $100/month for a 50/50 dedicated fiber connection which is entirely supplied by Level 3 for upstream. While my bandwidth is not "guaranteed", it is marketed as "dedicated" and I have called in at 2am complaining about 30ms pings to servers I normally get 8ms to, and they had a line tech out at my house at 8am that morning.
This is a "standard" residential connection around here. They don't differentiate between business and home users. Everyone gets the same packages offered and the same support. Mind you, they do have "enterprise" packages, and those are the "expensive" ones you're referring to.
Another time I was only getting 42mb/s on my speed test, and I have a 50mb connection. They got me in touch with an engineer and the problem was resolved on their end. They were in the process of upgrading from 1gb ports to 10gb for their line cards and my linecard's port was nearing saturation. They changed which linecard I was plugged into, it took them about 5 minutes and I had to power cycle my ONT.
I have been running a ping every 0.5sec to a datacenter in Chicago that peers with Level 3, which is 8 hops and 200+ miles from me, and I'm averaging a daily 0.005% packet-loss and sub 1ms of jitter.
You may want to ask what you're paying for. Midwest USA, ftw!
Yes, lets use the example of a bulk data transfer getting priority over low bandwidth latency sensitive gaming. Wait, doesn't priority only matter when there is congestion? Why is there congestion? Are they saying there is congestion on their internal network or do they mean they have congestion in the links to the other networks? If the congestion is in the external links, how do they propose "priority" to affect unrelated links?
How about they just deliver what they sell? F*ck "up to", they need to get rid of that legal loop hole. Our car gets "up to" 40mpg, but you'll never see more than 1mpg in normal usage.
You can detect corruption and it would probably cause a completely loss of network connection, which is still better than complete loss of some data and hardware.
I would be more concerned about power lines or transformers melting.
completely covered in ice
Depends on your definition of "completely". It is at the lowest ice coverage in recorded history. The Antarctic on the other hand, which is what this article is about, has been gaining ice, but at a lesser rate than the Arctic has been losing it.
Rolling your own is only cheap if you can do hundreds of homes at a time. It's about 10x more expensive to do one-offs.
They'll just send you to collections where you can pay $10/month for the next 5 years. I've found collections easier to work with and typically reduce the amount owed. I prefer to work with collections. I've dragged some collections for several thousand dollars out with $10/month for several years until they contacted me and said they'll remove 20% of the amount remaining if I pay at all once and that's after they already removed 10% when they're were trying to get me to pay more per month.
Don't think I was trying to skirt paying, I had other more pressing bills and a lot of health issues during that time of my life.
Tellers at a bank could be completely rep[laced with ATMs or the like. You don't need to trust people much when their jobs are pointless. When you depend on someone's skill, and you don't trust them, that's when you have issues.
People don't work well as a team when there is no trust.
Disk storage is not base 10, it uses a base 10 number of base 2 blocks. Show me a single drive that optimally aligns data to a base 10 address.
You only *think* you can do all that because you don't understand how it's being done. ZFS fixes a lot of corner failure/corruption cases.