That issue only applied if you installed the RC, not release. Interestingly enough, it was caused by a race condition during the build process and not the code.
If you use the same type of hardware the devs use, you'll have a mostly good time. If you use some obscure integrated hardware that you got with your $40 motherboard or some crazy exotic hardware with your $200 motherboard, it probably won't be supported. Most of FreeBSD devs are running Intel servers and similarly speced desktops.
They suspected that the person submitting the patch had already seen the Linux code, which means he has been contaminated. You can't use his code. Welcome to copyright and the GPL.
Words have meaning based on their usage. Anime is used to mean something different than "cartoons". Yes, anime is a type of cartoon, but definitely not the same thing. Remember, words only have any meaning because of how they're used. If you ignore how they're used, they're just a bunch of meaningless sounds.
A local non-profit has been providing enterprise grade 99.999% up-time 1Gb fiber for $300/m to schools, libraries, and hospitals for the past 15 years. Of course AT&T wanted to deny them right-of-way access because the non-profit was not a telcom or cable company. The state spent several million dollars in legal costs and won, by a hair, with the restriction that this non-profit could not sell to residential or private sector. The only reason the law makers allowed them access was because AT&T wanted to charge $100,000 for the exact same service, and that's ignoring all of the complaints of AT&T not fixing their network in a timely fashion. Lots of down time, even for the "enterprise" fiber connections.
That's an interesting problem. Googling it a bit, it seems it's an issue with HPET plus lower power CPU states. Load is relative to the current CPU speed. When my Haswell CPU clocks itself down to 200mhz, a load of 0.5 would indicate 100mhz of load. Pretty much every OS(Windows, Linux, FreeBSD) out there does this. The issue with FreeBSD is when a CPU core is put completely to sleep with 0mhz. There is a bias that when HPET interrupt occurs, it has to wake up the core, and the core has funny load characteristics when it comes out of sleep. This is exasperated by HPET waking up several cores at the same time and they're all counting each other's "load" by how responsive they are.
I did find someone who "fixed" the issue by changing their timing source from HPET to something else, but it meant that the CPU could not go into deep sleep, because that required HPET. Their CPU went from 7watts idle to 12watts idle, but the load issue went away.
Fiber to the node is 40% more expensive than fiber to the home. Nodes are very expensive. They're already talking about 25Tb/s fiber giving each customer 100Gb/100Gb of dedicated bandwidth. How long do you think it will be before they have copper nodes with a 25Tb/s upllink and able to handle 100Gb/s over your copper connection?
Not all PON is shared. Google Fiber uses WDM-PON, which is dedicated per lambda. NG-PON2 will eventually share 25Tb/s of bandwidth down the one fiber. At what point do you stop caring that it's "shared"? When a single fiber has more than 10% of the entire Internet's bandwidth, I don't see the issue. PON typically has great scheduling. You can set guaranteed minimum bandwidth along with maximum ping, also independently of each other. There are PONs that can guarantee that your ping will never go above 0.5ms under any situation, even if you swing from 1Gb/s to 10Mb/s. Of course that's a scheduling guarantee, not a bufferbloat guarantee.
Of course the lower you guarantee the latency, the less efficient the scheduler is at maximizing overall bandwidth. You might only be able to reach 85% peak bandwidth with 0.5ms, but 95% with 2.5ms.
Google is selling dedicated 1Gb to core bandwidth. Each customer has a 1.25Gb/1.25Gb lamda of light in a shared fiber. That fiber has 32 lambdas and 32 customers for a total of 40Gb/40Gb. That is a single port in an 8 port line-card that has 640Gb/s of bandwidth into the back-plane. The backplane has something like 8 line-cards and a total backplane bandwidth of 5.2Tb and 8+ 400Gb uplinks that plug directly into the core.
Google Fiber actually says you plug "directly" into the core. They call their GPON fiber chassis a "fiber aggregator" because it's really just a mostly dumb Layer 2+ whose sole job is to connection the customer into the core with dedicated bandwidth.
My ISP does a similar thing. A senior network admin said they could literally handle 100% of customers fully saturating all of their connections without issue in their core. They do not over-provision at all in the last-mile or core. They only thing they need to worry about is the trunk, which has at least 3x 95th percentile and 6x if they load-balance to their failover, which they have done when under a DDOS.
I have 3 desktop computers running 24/7, and even when I turned them off at night, my $50 electric bill did not change by a statistically significant amount. My electric bill is more affected by the weather or my wife's cooking habits. All of my lights are 10watt LEDs or 12.5watt CFLs, and the 2-3 lights that we run over 80% of the time are the LEDs in our living room with the computers. 80% of my electric bill comes from the stove, clothes drier, and fridge.
My wife gets pissed off if she has to reload Chrome. Screw OS up-time, she bases her system stability in Chrome up-time. She'll wait until Chrome's update icon color is red for a week before relaunching it. She used to go 2-3 months between rebooting her computer. Window's updates got mighty angry. Now Win10 reboots itselt at night, and I hear about it every time. "omg, my computer restarted! Now I gotta wait 15 seconds for Chrome to load my 50 tabs!". And I'm glad I have a low latency dedicated fiber connection with Level 3 as upstream. Opening Chrome pegs the connection for quite a bit with 50+ tabs. At least I got her a Samsung Pro M2. Freaking 2.5GiB/s and 300k IO/s, and still not fast enough.
I've spoiled her with good computers over the years. She has come to expect the computer to be instantly ready 24/7. I've been custom building computers since a wee lad.
What were we talking about again? Ohh yeah. I don't know anyone who doesn't leave their computer on 24/7, even my mom. Computers are appliances and need to be ready at a moment's notice.
Bodily functions don't bother people, bodily functions that spread disease do. Of course understanding this requires abstract thought from higher level thinking which seems to be lost on many.
You can have RAM that runs as fast as CPU cache, you just can't afford it.
Not really. Cache is only fast because of the huge number of traces. Unfortunately, you have to choose between one of two things for larger caches. n^2 increase in traces as your cache size increases or reduce the size of n at the expense collisions, which means your "memory" is now lossy. That also ignore the nasty increase of latency as cache gets larger. After a certain point, well within the range of current memory sizes, cache will have higher latency than DRAM.
Fiber in DCs is also nice to isolate surges or ground differentials. Copper may be faster for signal prorogation, but the complexity of processing the signal at high bandwidth adds much more latency than the signal over short runs.
I actually had an issue with NoScript in Chrome that caused my tabs to sometimes use a large amount of CPU and other times crash. I couldn't even go to Google.com without it crashing. Some pages with absolutely no external links or even javascript would take tens of second to load with it enabled and pretty much instant with it disabled. I was able to find people complaining about this starting many many years ago. I eventually built a new computer, fresh install of Windows. Gave it another try before I even installed any other software in order to see if there was a conflict with a program I had previously installed in my old computer. Same issue. uMatrix has worked fine for me.
Technology is a production multiplier. Every hour worked produces 10 hours of goods. We're at a point in history where we can out-produce consumption. The only reason prices are as high as they are is artificial limitations in order to make capitalism work. You make the assumption that the rules won't be changed and we'll just give more people jobs. We do need to make some changes, but not a whole lot. The bigger issue is psychological buy in. Money only has value because people think it has value.
AI also doesn't need to replace all tech workers. It just has to apply pressure at the margins.
I already do that in my job. I take an 8 hour process and turn it into a 5 minute one. My job is to automate away other people's jobs. Of course my job could be replaced by an AI, but by that point AI's will be both the producers and consumers.
Slashdot Mirror
PC-BSD is just a thin layer on top of FreeBSD. You can upgrade and downgrade FreeBSD to/from PC-BSD.
That issue only applied if you installed the RC, not release. Interestingly enough, it was caused by a race condition during the build process and not the code.
If you use the same type of hardware the devs use, you'll have a mostly good time. If you use some obscure integrated hardware that you got with your $40 motherboard or some crazy exotic hardware with your $200 motherboard, it probably won't be supported. Most of FreeBSD devs are running Intel servers and similarly speced desktops.
They suspected that the person submitting the patch had already seen the Linux code, which means he has been contaminated. You can't use his code. Welcome to copyright and the GPL.
It's not a double standard, Windows is the standard, Linux is not. The hope is the new standard is for all games to be cross platform.
Words have meaning based on their usage. Anime is used to mean something different than "cartoons". Yes, anime is a type of cartoon, but definitely not the same thing. Remember, words only have any meaning because of how they're used. If you ignore how they're used, they're just a bunch of meaningless sounds.
A local non-profit has been providing enterprise grade 99.999% up-time 1Gb fiber for $300/m to schools, libraries, and hospitals for the past 15 years. Of course AT&T wanted to deny them right-of-way access because the non-profit was not a telcom or cable company. The state spent several million dollars in legal costs and won, by a hair, with the restriction that this non-profit could not sell to residential or private sector. The only reason the law makers allowed them access was because AT&T wanted to charge $100,000 for the exact same service, and that's ignoring all of the complaints of AT&T not fixing their network in a timely fashion. Lots of down time, even for the "enterprise" fiber connections.
That's an interesting problem. Googling it a bit, it seems it's an issue with HPET plus lower power CPU states. Load is relative to the current CPU speed. When my Haswell CPU clocks itself down to 200mhz, a load of 0.5 would indicate 100mhz of load. Pretty much every OS(Windows, Linux, FreeBSD) out there does this. The issue with FreeBSD is when a CPU core is put completely to sleep with 0mhz. There is a bias that when HPET interrupt occurs, it has to wake up the core, and the core has funny load characteristics when it comes out of sleep. This is exasperated by HPET waking up several cores at the same time and they're all counting each other's "load" by how responsive they are.
I did find someone who "fixed" the issue by changing their timing source from HPET to something else, but it meant that the CPU could not go into deep sleep, because that required HPET. Their CPU went from 7watts idle to 12watts idle, but the load issue went away.
Fiber to the node is 40% more expensive than fiber to the home. Nodes are very expensive. They're already talking about 25Tb/s fiber giving each customer 100Gb/100Gb of dedicated bandwidth. How long do you think it will be before they have copper nodes with a 25Tb/s upllink and able to handle 100Gb/s over your copper connection?
Not all PON is shared. Google Fiber uses WDM-PON, which is dedicated per lambda. NG-PON2 will eventually share 25Tb/s of bandwidth down the one fiber. At what point do you stop caring that it's "shared"? When a single fiber has more than 10% of the entire Internet's bandwidth, I don't see the issue. PON typically has great scheduling. You can set guaranteed minimum bandwidth along with maximum ping, also independently of each other. There are PONs that can guarantee that your ping will never go above 0.5ms under any situation, even if you swing from 1Gb/s to 10Mb/s. Of course that's a scheduling guarantee, not a bufferbloat guarantee.
Of course the lower you guarantee the latency, the less efficient the scheduler is at maximizing overall bandwidth. You might only be able to reach 85% peak bandwidth with 0.5ms, but 95% with 2.5ms.
Google is selling dedicated 1Gb to core bandwidth. Each customer has a 1.25Gb/1.25Gb lamda of light in a shared fiber. That fiber has 32 lambdas and 32 customers for a total of 40Gb/40Gb. That is a single port in an 8 port line-card that has 640Gb/s of bandwidth into the back-plane. The backplane has something like 8 line-cards and a total backplane bandwidth of 5.2Tb and 8+ 400Gb uplinks that plug directly into the core.
Google Fiber actually says you plug "directly" into the core. They call their GPON fiber chassis a "fiber aggregator" because it's really just a mostly dumb Layer 2+ whose sole job is to connection the customer into the core with dedicated bandwidth.
My ISP does a similar thing. A senior network admin said they could literally handle 100% of customers fully saturating all of their connections without issue in their core. They do not over-provision at all in the last-mile or core. They only thing they need to worry about is the trunk, which has at least 3x 95th percentile and 6x if they load-balance to their failover, which they have done when under a DDOS.
IPv6 will make it first.
I have 3 desktop computers running 24/7, and even when I turned them off at night, my $50 electric bill did not change by a statistically significant amount. My electric bill is more affected by the weather or my wife's cooking habits. All of my lights are 10watt LEDs or 12.5watt CFLs, and the 2-3 lights that we run over 80% of the time are the LEDs in our living room with the computers. 80% of my electric bill comes from the stove, clothes drier, and fridge.
My wife gets pissed off if she has to reload Chrome. Screw OS up-time, she bases her system stability in Chrome up-time. She'll wait until Chrome's update icon color is red for a week before relaunching it. She used to go 2-3 months between rebooting her computer. Window's updates got mighty angry. Now Win10 reboots itselt at night, and I hear about it every time. "omg, my computer restarted! Now I gotta wait 15 seconds for Chrome to load my 50 tabs!". And I'm glad I have a low latency dedicated fiber connection with Level 3 as upstream. Opening Chrome pegs the connection for quite a bit with 50+ tabs. At least I got her a Samsung Pro M2. Freaking 2.5GiB/s and 300k IO/s, and still not fast enough.
I've spoiled her with good computers over the years. She has come to expect the computer to be instantly ready 24/7. I've been custom building computers since a wee lad.
What were we talking about again? Ohh yeah. I don't know anyone who doesn't leave their computer on 24/7, even my mom. Computers are appliances and need to be ready at a moment's notice.
Bodily functions don't bother people, bodily functions that spread disease do. Of course understanding this requires abstract thought from higher level thinking which seems to be lost on many.
Have you seen Lord of the Rings? You would take a 13 year old to that?
I saw Aliens and Predator around the age of 7. I was running around blowing up demons in Doom when I was 8. I don't get the issue.
How's this from my home PFSense firewall? Seems my $60 MSI motherboard is pretty good at keeping time. -0.023ms offset after 512 seconds.
remote refid st t when poll reach delay offset jitter
==removed junk chars==
-ra.steadfastdns 216.86.146.46 2 u 388 512 377 14.234 -0.706 0.632
+rb.steadfastdns 216.86.146.46 2 u 144 512 377 14.356 -0.533 0.546
*dns1.steadfast. 216.86.146.46 2 u 400 512 377 14.826 -0.023 0.183
+time1.google.co 133.45.56.223 2 u 386 512 377 24.719 0.007 0.181
-time2.google.co 222.174.180.252 2 u 362 512 377 34.593 -3.214 0.550
-time3.google.co 63.32.115.164 2 u 354 512 377 38.754 -0.978 0.568
-time4.google.co 226.42.99.180 2 u 318 512 377 23.877 -0.292 0.398
Who needs sleep when you can spin? CPU time is cheap. This works really great on single core systems.
http://www.extremetech.com/ext... 13.8 billion year archive media with ultra-high densities.
If I can't make a bootable ZFS volume from a bootdisk that I downloaded, it doesn't work.
You can have RAM that runs as fast as CPU cache, you just can't afford it.
Not really. Cache is only fast because of the huge number of traces. Unfortunately, you have to choose between one of two things for larger caches. n^2 increase in traces as your cache size increases or reduce the size of n at the expense collisions, which means your "memory" is now lossy. That also ignore the nasty increase of latency as cache gets larger. After a certain point, well within the range of current memory sizes, cache will have higher latency than DRAM.
Fiber in DCs is also nice to isolate surges or ground differentials. Copper may be faster for signal prorogation, but the complexity of processing the signal at high bandwidth adds much more latency than the signal over short runs.
I actually had an issue with NoScript in Chrome that caused my tabs to sometimes use a large amount of CPU and other times crash. I couldn't even go to Google.com without it crashing. Some pages with absolutely no external links or even javascript would take tens of second to load with it enabled and pretty much instant with it disabled. I was able to find people complaining about this starting many many years ago. I eventually built a new computer, fresh install of Windows. Gave it another try before I even installed any other software in order to see if there was a conflict with a program I had previously installed in my old computer. Same issue. uMatrix has worked fine for me.
Technology is a production multiplier. Every hour worked produces 10 hours of goods. We're at a point in history where we can out-produce consumption. The only reason prices are as high as they are is artificial limitations in order to make capitalism work. You make the assumption that the rules won't be changed and we'll just give more people jobs. We do need to make some changes, but not a whole lot. The bigger issue is psychological buy in. Money only has value because people think it has value.
AI also doesn't need to replace all tech workers. It just has to apply pressure at the margins.
I already do that in my job. I take an 8 hour process and turn it into a 5 minute one. My job is to automate away other people's jobs. Of course my job could be replaced by an AI, but by that point AI's will be both the producers and consumers.