I used to admin at a company in Mountain View that sat on the same fiber ring as Google, one that was operated by SBC, the local LEC. We had nothing but problems with throughput and such which our rep intimated was due to Google's needs. I'm not sure how much of Google's service is channeled in and out of their main headquarters but apparently it's significant. Towards the end of my time at that company our service improved dramatically, which I attribute to Google going a different route.
It makes more business sense to lease dark fiber, break it into waves with WDM and run parallel OCn circuits once you start pushing a certain amount of bandwidth, which clearly Google does. This as opposed to buying OCn circuits from the incumbent LEC which is very pricey. If you're going to go to all the trouble of leasing dark fiber and you have enough smarts in-house to do it, you might as well register yourself as a CLEC so you can buy your voice services wholesale. That would also give you the opportunity to position yourself in the VoIP market, if you were interested in doing so.
This "Internet in a box" thing definitely sounds like something but I'm dubious it has anything to do with dark fiber.
I've been a Time Warner digital cable subscriber for a while and my last few cable boxes have all been Scientific Atlanta. Maybe there's a preexisting contract of some kind they can use to get their foot in the door in the services-via-settop market? Otherwise, why bother...
It shouldn't make a difference unless Linux naturally hits the CPU cache more than Windows. AFAIK, the OS doesn't have any insight into whether pages are cached on the CPU or within main memory.
It makes sense that a app that consumes memory, I/O and CPU cycles like a database server would miss L1/L2 caches more often than not. The implication then is that it consumes more cycles for these "hyperthreads" to wait on a fetch than threads running on the main CPU. Is that right?
Unless I'm missing something, which is entirely possible, software and kernel developers are completely innocent here as hardware makes the choice whether an instruction is executed within a CPU's main core or hyperthreaded core.
If you've ever tried to implement an IPSEC VPN with numerous endusers that have DSL/CableModem gateways that default to 192.168.1.x, you'll know why NAT is so bad, particularly if you're using that address space internally already. Granted, there are workarounds to this.
That's dicey, but what's even more dicey is trying to interconnect corporate networks that use the same private address space. Companies that run virtual trading floors, for example, offer private line connections. You end up with multiple IP subnet conflicts and it's an incredible headache. That having been said, there are workarounds to that, too.
When NAT became popular way back when, I was part of a few really painful reIPing projects. The reason we went to NAT was because there was no way to get portable IP space and our ISP was being a complete dick, jacking their prices and refusing to run BGP with us. Moving to NAT meant portability and portability meant our ISP couldn't dick us. If I was to move away from NAT and put v6 addresses in my corp network, that's what I'd worry about more than anything.
It's more accurate to say "gaming fanaticism" and drug addiction both show characteristics of obsessive compulsive disorder. Both exhibit mental preoccupation (obsession) and a repetitive behavior pattern(s) that may or may not be reinforced with a positive stimulus (compulsion). If the individual tries to break the compulsive behavior pattern, he/she is pressed by the obsessive thoughts and becomes uncomfortable. It's as if the brain had imported a subroutine with an ill-placed GOTO loop.
Any behavior repeated constantly over a period of time that provides some sort of positive effect is going to be difficult to tear away from and may bring about OCD type symptoms. But comparing it to drug addiction is sensationalism... people don't kill each other or themselves on account of Sonic Hedgehog.
I don't have much experience with satellite-based networks but I've heard the latency makes them less-than-optimal. It doesn't matter how much pipe you have if it takes an eternity to get packets to and fro, especially if there's any packet loss (which assumedly there would be). Any interactive appliaction or protocol that used multiple simultaneous connections (like HTTP) would feel really slow.
If this network is no better than bonding 28.8 modems together, what's the point?
Not to be a Google sympathizer, but Alexa ranks websites based on where people go who have their toolbar installed.
Alexa-toolbar-who?
I don't think there's any way to get a good answer on how many visits anyone's getting.
Yahoo is investigating partnerships with other DVR makers in order to deliver content to the set-top box. Any DVR that has broadband capabilities would in theory be able to deliver vast catalogs of movies on demand, siphoning all the revenue left from the Blockbusters and NetFlixes of the world. That's serious bank.
It's not just the Yahoos and DVR makers who are maneuvering around this eventuality: broadband providers, most notably SBC, will be aggressively marketing DVRs with enhanced services like Yahoo's to their subscriber base when they roll out HDSL. SBC's HDSL service is called LightSpeed and it's expected to beta in central Texas within the next few months. Good stuff.
At least to some people.
Decency and humor are subjective. What offends you doesn't offend me and vice versa. There's a new trend in America that we have to conform to the standards of people who are easily offended. That's because in America we're under the influence of a population who insists that what offends them offends God and so must be eliminated. And being that most of these people are less than comfortable with physicality, they complain about meaningless things like this under the guise of being kind and serving The Lord (TM).
If I were doing something online that this administration or my community finds particularly heinous, and they tracked me down by my IP address and busted down my door, I could simply stipulate "fuck off, it wasn't me, somebody must have hacked into my access point." It would be up to them to prove otherwise. Now they have something new to go at me with.
Our elected officials' perverse fascination with the goings-on of its citizens and our willingness to let these things slide is going to be a really big deal in the near future. Oops, did I say that out loud? There'll be a van parked across the street sometime soon I expect.
If you were ISP in distress, wouldn't you find a way to circumvent this downed peering point? I'm sure other Tier 1s have jumped in and offered transit for a modest sum. If I were an evil ISP and I wanted to make sure that didn't happen, I'd blackhole all the netblocks (or any ASPATH with his AS) from the guy I wanted moolah from until he forked over the funds. Unless I'm missing something, that's the explanation for why networks who are purchasing transit from Cogent/L3 are having trouble.
Slashdot Mirror
I used to admin at a company in Mountain View that sat on the same fiber ring as Google, one that was operated by SBC, the local LEC. We had nothing but problems with throughput and such which our rep intimated was due to Google's needs. I'm not sure how much of Google's service is channeled in and out of their main headquarters but apparently it's significant. Towards the end of my time at that company our service improved dramatically, which I attribute to Google going a different route.
It makes more business sense to lease dark fiber, break it into waves with WDM and run parallel OCn circuits once you start pushing a certain amount of bandwidth, which clearly Google does. This as opposed to buying OCn circuits from the incumbent LEC which is very pricey. If you're going to go to all the trouble of leasing dark fiber and you have enough smarts in-house to do it, you might as well register yourself as a CLEC so you can buy your voice services wholesale. That would also give you the opportunity to position yourself in the VoIP market, if you were interested in doing so.
This "Internet in a box" thing definitely sounds like something but I'm dubious it has anything to do with dark fiber.
I've been a Time Warner digital cable subscriber for a while and my last few cable boxes have all been Scientific Atlanta. Maybe there's a preexisting contract of some kind they can use to get their foot in the door in the services-via-settop market? Otherwise, why bother ...
Doing coke off of stripper's tits? Geek alert. Only someone who watched Robocop more than a half-dozen times would think that qualified as cool.
It makes sense that a app that consumes memory, I/O and CPU cycles like a database server would miss L1/L2 caches more often than not. The implication then is that it consumes more cycles for these "hyperthreads" to wait on a fetch than threads running on the main CPU. Is that right?
Unless I'm missing something, which is entirely possible, software and kernel developers are completely innocent here as hardware makes the choice whether an instruction is executed within a CPU's main core or hyperthreaded core.
Shouldn't it be called AdminKit?
If you've ever tried to implement an IPSEC VPN with numerous endusers that have DSL/CableModem gateways that default to 192.168.1.x, you'll know why NAT is so bad, particularly if you're using that address space internally already. Granted, there are workarounds to this.
That's dicey, but what's even more dicey is trying to interconnect corporate networks that use the same private address space. Companies that run virtual trading floors, for example, offer private line connections. You end up with multiple IP subnet conflicts and it's an incredible headache. That having been said, there are workarounds to that, too.
When NAT became popular way back when, I was part of a few really painful reIPing projects. The reason we went to NAT was because there was no way to get portable IP space and our ISP was being a complete dick, jacking their prices and refusing to run BGP with us. Moving to NAT meant portability and portability meant our ISP couldn't dick us. If I was to move away from NAT and put v6 addresses in my corp network, that's what I'd worry about more than anything.
It's more accurate to say "gaming fanaticism" and drug addiction both show characteristics of obsessive compulsive disorder. Both exhibit mental preoccupation (obsession) and a repetitive behavior pattern(s) that may or may not be reinforced with a positive stimulus (compulsion). If the individual tries to break the compulsive behavior pattern, he/she is pressed by the obsessive thoughts and becomes uncomfortable. It's as if the brain had imported a subroutine with an ill-placed GOTO loop.
... people don't kill each other or themselves on account of Sonic Hedgehog.
Any behavior repeated constantly over a period of time that provides some sort of positive effect is going to be difficult to tear away from and may bring about OCD type symptoms. But comparing it to drug addiction is sensationalism
I don't have much experience with satellite-based networks but I've heard the latency makes them less-than-optimal. It doesn't matter how much pipe you have if it takes an eternity to get packets to and fro, especially if there's any packet loss (which assumedly there would be). Any interactive appliaction or protocol that used multiple simultaneous connections (like HTTP) would feel really slow.
If this network is no better than bonding 28.8 modems together, what's the point?
Not to be a Google sympathizer, but Alexa ranks websites based on where people go who have their toolbar installed. Alexa-toolbar-who? I don't think there's any way to get a good answer on how many visits anyone's getting.
Yahoo is investigating partnerships with other DVR makers in order to deliver content to the set-top box. Any DVR that has broadband capabilities would in theory be able to deliver vast catalogs of movies on demand, siphoning all the revenue left from the Blockbusters and NetFlixes of the world. That's serious bank.
It's not just the Yahoos and DVR makers who are maneuvering around this eventuality: broadband providers, most notably SBC, will be aggressively marketing DVRs with enhanced services like Yahoo's to their subscriber base when they roll out HDSL. SBC's HDSL service is called LightSpeed and it's expected to beta in central Texas within the next few months. Good stuff.
At least to some people. Decency and humor are subjective. What offends you doesn't offend me and vice versa. There's a new trend in America that we have to conform to the standards of people who are easily offended. That's because in America we're under the influence of a population who insists that what offends them offends God and so must be eliminated. And being that most of these people are less than comfortable with physicality, they complain about meaningless things like this under the guise of being kind and serving The Lord (TM).
If I were doing something online that this administration or my community finds particularly heinous, and they tracked me down by my IP address and busted down my door, I could simply stipulate "fuck off, it wasn't me, somebody must have hacked into my access point." It would be up to them to prove otherwise. Now they have something new to go at me with. Our elected officials' perverse fascination with the goings-on of its citizens and our willingness to let these things slide is going to be a really big deal in the near future. Oops, did I say that out loud? There'll be a van parked across the street sometime soon I expect.
If you were ISP in distress, wouldn't you find a way to circumvent this downed peering point? I'm sure other Tier 1s have jumped in and offered transit for a modest sum. If I were an evil ISP and I wanted to make sure that didn't happen, I'd blackhole all the netblocks (or any ASPATH with his AS) from the guy I wanted moolah from until he forked over the funds. Unless I'm missing something, that's the explanation for why networks who are purchasing transit from Cogent/L3 are having trouble.