For first-time callers, you need a little bit of an IVR front end, ideally some kind of TellMe system. Then you have additional information about a caller before it rings your extension, and if it is really advanced it can determine who the call actually goes to. If the caller is accepted as legitimate, it gets added to the whitelist, if it is rejected (by a human) it goes to the blacklist. Everything else stays greylisted.
Works great for individuals, not so well for businesses. You never know when a lead will come in, and you have to be careful how much effort you put a potential customer through.
Will deal with it in much the same way; known bad callers go directly to the honeypot, known good callers go through. Unknown callers will need some kind of probabilistic assessment as to how much IVR and call screening you put them through.
Since gas passed $3.50/gallon, I started to see other bicycles. At $4.00/gallon, I see a bunch of them. Now, at $4.50/gallon, I notice significantly fewer cars, more bikes, and people actually riding the bus.
I live in Los Angeles, possibly the most auto-centric and bicycle/pedestrian unfriendly city in the US.
I also live in a community with a median income over $150k, where people can afford to drive their big ass SUVs, and often prefer to not associate with the "kind of people that ride the bus."
Commuting to work via airplane to me is a different story. I've done it myself at different times. It just shows an imbalance between resources and demand, and many of those things can't be resolved overnight. (In my field, it looks like it is only getting worse over time.)
Mainly logistics and dependencies cause the delay. Hardware, software, and database failures must often be found before moving up the chain and starting the next system. Some functionality is usually capable of being restored within a couple hours.
Some of these things have been improved through automation, but the best-case example I have seen is reducing the time by 30-50%. We seem to have given up a lot of ground when you add back in virtualization.
But at the same time, there are fewer people tending to the equipment so you often have limited resources to recover when there are problems.
We usually don't worry too much about all the load coming back on at once from a power perspective, but standard operating procedure for many facilities is to do non-automatic restart for the power distribution units, so power must be manually restored in smaller blocks. It doesn't hurt to stagger things, and some of my clients prefer to switch each breaker back on one at a time with a little time delay.
An outage in a data center over about 2.2MW is a major hassle to re-start. Over about 5MW can be a 48 hour effort. When you get into these 20+ MW facilities, it can easily take weeks to get everything back up and running.
When a facility is properly compartmentalized, it isn't nearly as bad-- redundancies and fail-over mechanisms can continue to maintain most of the system operation, and hopefully extra load can be shifted to another site.
The problem is that historically data centers don't have fires. (In contrast, telco switch facilities have them all the time.) Electrically when we get over about 10-20MW of UPS in a single structure data center, the complexity of systems and maintenance provisions greatly increases the risk of fire. From a raised floor perspective, when we get over 20kW per rack, we have seen a couple small fires (out of thousands-- don't get me wrong, it isn't a huge widespread problem). With these changes brought on by the "mega-centers," it takes a lot to improve (electrical) reliability for the site.
So, in my book, it isn't the fact that you shouldn't be prepared for a data center to go down some times, it is that there is more concentration of facilities and they are being done at a larger scale which will impact the reliability in a major way. We advise most of our clients to keep under 6MW for a data center, and go for multiple facilities geographically isolated for the extra capacity. That approach isn't always commercially viable, but it is makes for a better long-term investment.
I don't disagree with the "truth about real-time quotes," but it often tells you a lot about how the market is reacting to news which can have some interesting implications.
As to the other subscription services, i was surprised that the WSJ is starting to go to real-time quotes as well. The more the merrier-- greater transparency for all.
Standard procedure on killing power to the site in a fire. All fire chiefs will require this, and given the situation it isn't a bad policy, especially if there is a structural failure or a major short. Only way around it is to make independent data halls with separate infrastructure.
As for your fire alarm story, your engineer or installing contractor didn't do their job; you coordinate it in advance and make sure everybody is on the same page. Same goes for the CO sensor-- you just use a cup around the sensor and inject the gas. (There are often alternate trigger gases that can be used as well.)
As much as it shouldn't happen, the law of unintended consequences rears its ugly head often enough when systems are designed for reliability first and safety a close second.
Per the NEC (National Electrical Code), this type of consideration is acceptable where losses associated with a failure could exceed the value provided by the safety improvement. The key is that only qualified people work on the system.
Ten to fifteen years ago, designing a main-tie-main switchboard for closed-transition operation, you would assume the duty rating to be that of the single-ended configuration, not parallel operation of both feeders. It was standard practice. Today, an overlapping transfer switch (performing the same basic function, only 1,000x faster) is considered to be a permanent tie between systems, so the short-circuit duty rating of the equipment must be considerably higher.
The reason for the change is that a theoretical potential for a problem turned into a real problem one or two times, and "best practices" changed.
This happens constantly. Underlying assumptions change in subtle ways, and an unforeseen (or purely academic) risk becomes real.
I have a client that won't do an upgrade that we have recommended, at least in a time-line I am comfortable with. I know that this failure is likely to lead to a major outage, and I cross my fingers that nobody gets hurt when it happens. Other clients have similar nagging problems that they are aware of, but you can't fix everything in a day. Some problems take several years to get repaired. Welcome to the world of capital planning.
This isn't that uncommon with a 200kAIC board with air-power breakers, if there is a bolted fault. Instantaneous delays. Newer insulated-case style breakers all have an instantaneous override which will limit fault energy,
The other possibility was that a tie was closed and the breakers over-dutied and could not clear the fault.
Odd that nobody was hurt though; spontaneous shorts are very rare-- most involve either switching or work in live boards, either of which would kill someone.
Insurance doesn't usually do much in these cases-- it might cover equipment but not labor at double overtime. It is easy to spend $500k just in engineering and stop-gap repairs on an electical-only incident. Fire would usually cause twice that much damage for stop-gap repairs.
Actually, someone is making the closed-circuit cooling towers with coils oversized for dry operation again. The one company doing it died in the dot-bomb, but someone more mainstream has picked it up again.
This type of system is really interesting when you look at optimizing power and water usage dynamically based on rates, approach, and plant efficiency, as there is no loss in reliability. (just cost...)
It sounds more like a cult of personality claiming they are high-tech. Their cold-aisle containment system has been around for years and produced by several vendors and engineers. It does work pretty well; our CFD models suggested a 30-50% reduction in fan energy.
There are a lot of neat ways to provide cooling, but unless you get heat directly off the chip to a coolant, it isn't easy to do anything but evaporative cooling in the desert... at least not without a huge thermal mass.
My calculation is closer to 2 million gallons, assuming they do everything they can to reduce drift and operate at a very high delta-t. You could also drop that down by half most of the year if you use the cooling towers as dry coolers at night.
Still, that is a huge amount of water storage even to run for a single day. Usually we try to have 72 hours of water storage on-site. You can use the dry coolers (or air cooled chillers), but the efficiency goes to hell in the Las Vegas climate.
For a more human scale of the water needs, it is about a 10" pipe running full out all day just to accommodate evaporation, drift, and blow-down.
The existing installer.app VT100 terminal is functional, and about as good as it will get with the touchscreen. I've used a Blackberry, Palm, N770, and my iPhone. The experience is comparable for command line on all of them, but none of them are worth a squat when I go to use nano or lynx for whatever reasons.
It really depends on what you need to do with it. If you just want to check security logs or do a manual backup, the iPhone is good enough.
We usually see 0.3-1% failure rates on servers in a data center powerdown (scheduled, soft shutdown). Bigger issues with blades and high power 1U servers, fewer with midrange and mainframe equipment. Hard drive and power supply failures are most common, generally attributed to thermal shock on re-start.
Would love to see any information that can debunk so we can hit the equipment manufacturers up for damages...
Personally, in my office I have a mad-scientist corner. We're engineers, and it is fun to actually make things rather than just design them all the time. I think over the past 30-some years of my life I have owned a dozen soldering irons and never really took a serious interest in learning how to actually do a damn thing with them. A couple years ago, we couldn't find something that met our needs, so we designed something to integrate different readily-available parts, and put it all together ourselves.
I'd say the key is to have enough cash to spend on stuff, not fearing breaking things (goes with part one), have a project that is important to you for starting off, and get a big old DigiKey and Allied Electronics catalog for ideas. Also, practice soldering on the cheap stuff if your natural talent for it is as bad as mine...
First off, the lineman will always ground out a medium voltage system prior to doing any work on it. That is the only safe way to work on the system.
This has the added benefit of shorting out your non-grid-tie inverter, which will eliminate the problem at the source.
On the low voltage side, any electrician (or geek) that doesn't have a voltage detector (neon bulb screwdriver or the nicer non-contact Fluke) units is a fool asking for a shock.
You can use a grid-tie system for independence from the utility's glitches. It is a conveinence thing, not a substitute for a UPS. Usually you will either need a contactor or a shunt-trip breaker that is manually reset when power is restored.
It isn't as energy efficient, but it provides a small base load rather than an impulse load to your personal "grid."
As to the article's fundamental question... it's called a blocking diode. Have a 120VDC or 48VDC power supply for the home that you tap off of for your needs. As long as you aren't trying to sell power back to the utility, this is the easiest way to make things work, and considerably cheaper. You do have an incentive to have some batteries in this configuration, though (on a separate DC:DC charging system ideally).
Power is generally not a huge issue, as containerized generators are fairly standardized and rapidly deployable. Cooling on the other hand is where most of these solutions fail. A large air cooled generator can be shipped in on skids, but will usually require an external pump skid to get things operational.
The problem with integrating cooling solutions is that it is too climate dependent to standardize. Tightly managed direct evaporative coolers can make a solid solution until you get to very hot, humid climates (over 75F wet bulb). At that point, you are going to have to have separate pre-cooling equipment to keep things working.
The only alternative is high-temperature heat rejection systems where you integrate a heat pipe into the chip and can work with 120F return water temperatures. Seems like fantasy at this point though...
While I used to love my old Nokia brick (most usable plain-jane cell phone I ever owned), I am at a complete loss of words with people's love of Symbian smart phones!
I tried the N95 and a couple other devices a year or two ago, and found everything to be abysmal hardware and clunky software as I went through Nokia and Sony-Ericson's flagship stores. The WinCE world was clearly worse, but offered better compatibility with the office.
But... what is great about the iPhone and some of HTC's models (by assumption rather than experience for the latter) is the huge, high-resolution screen. It really makes the difference in making the device usable on the road. I used to be a blackberry addict, and am quite happy with my transition to the iPhone.
Why are Europeans so passionate about their Symbian phones? Have they really gotten that much better, or is it just random evangelism?
My wife's bicycle was stolen at her work (directly across the street from the police station, with regular police officer foot-traffic in the building). We had fancy cameras and a close-up of the guy's face within an hour of the theft.
Did it help anything? No...
The cameras were also in plain sight, and he was especially brazen in how he went about it all.
Hard hat, clip board, tyvec suits, pointing a boom-shaped object at things while looking at a display, or rolling around a cart of strange looking tools will generally get you into plenty of places you shouldn't be going. Oddly enough, no matter how well secured the front door is (well after 9/11), there is usually a back door that can get you in.
The most surprising one is just juggling a bunch of crap and asking for the guard to let you through the man-trap.
While my purposes aren't nefarious, these things can easily be exploited.
Slashdot Mirror
For first-time callers, you need a little bit of an IVR front end, ideally some kind of TellMe system. Then you have additional information about a caller before it rings your extension, and if it is really advanced it can determine who the call actually goes to. If the caller is accepted as legitimate, it gets added to the whitelist, if it is rejected (by a human) it goes to the blacklist. Everything else stays greylisted.
Works great for individuals, not so well for businesses. You never know when a lead will come in, and you have to be careful how much effort you put a potential customer through.
It's called headhunters.
Will deal with it in much the same way; known bad callers go directly to the honeypot, known good callers go through. Unknown callers will need some kind of probabilistic assessment as to how much IVR and call screening you put them through.
Since gas passed $3.50/gallon, I started to see other bicycles. At $4.00/gallon, I see a bunch of them. Now, at $4.50/gallon, I notice significantly fewer cars, more bikes, and people actually riding the bus.
I live in Los Angeles, possibly the most auto-centric and bicycle/pedestrian unfriendly city in the US.
I also live in a community with a median income over $150k, where people can afford to drive their big ass SUVs, and often prefer to not associate with the "kind of people that ride the bus."
Commuting to work via airplane to me is a different story. I've done it myself at different times. It just shows an imbalance between resources and demand, and many of those things can't be resolved overnight. (In my field, it looks like it is only getting worse over time.)
Mainly logistics and dependencies cause the delay. Hardware, software, and database failures must often be found before moving up the chain and starting the next system. Some functionality is usually capable of being restored within a couple hours.
Some of these things have been improved through automation, but the best-case example I have seen is reducing the time by 30-50%. We seem to have given up a lot of ground when you add back in virtualization.
But at the same time, there are fewer people tending to the equipment so you often have limited resources to recover when there are problems.
We usually don't worry too much about all the load coming back on at once from a power perspective, but standard operating procedure for many facilities is to do non-automatic restart for the power distribution units, so power must be manually restored in smaller blocks. It doesn't hurt to stagger things, and some of my clients prefer to switch each breaker back on one at a time with a little time delay.
An outage in a data center over about 2.2MW is a major hassle to re-start. Over about 5MW can be a 48 hour effort. When you get into these 20+ MW facilities, it can easily take weeks to get everything back up and running.
When a facility is properly compartmentalized, it isn't nearly as bad-- redundancies and fail-over mechanisms can continue to maintain most of the system operation, and hopefully extra load can be shifted to another site.
The problem is that historically data centers don't have fires. (In contrast, telco switch facilities have them all the time.) Electrically when we get over about 10-20MW of UPS in a single structure data center, the complexity of systems and maintenance provisions greatly increases the risk of fire. From a raised floor perspective, when we get over 20kW per rack, we have seen a couple small fires (out of thousands-- don't get me wrong, it isn't a huge widespread problem). With these changes brought on by the "mega-centers," it takes a lot to improve (electrical) reliability for the site.
So, in my book, it isn't the fact that you shouldn't be prepared for a data center to go down some times, it is that there is more concentration of facilities and they are being done at a larger scale which will impact the reliability in a major way. We advise most of our clients to keep under 6MW for a data center, and go for multiple facilities geographically isolated for the extra capacity. That approach isn't always commercially viable, but it is makes for a better long-term investment.
I don't disagree with the "truth about real-time quotes," but it often tells you a lot about how the market is reacting to news which can have some interesting implications.
As to the other subscription services, i was surprised that the WSJ is starting to go to real-time quotes as well. The more the merrier-- greater transparency for all.
Standard procedure on killing power to the site in a fire. All fire chiefs will require this, and given the situation it isn't a bad policy, especially if there is a structural failure or a major short. Only way around it is to make independent data halls with separate infrastructure.
As for your fire alarm story, your engineer or installing contractor didn't do their job; you coordinate it in advance and make sure everybody is on the same page. Same goes for the CO sensor-- you just use a cup around the sensor and inject the gas. (There are often alternate trigger gases that can be used as well.)
As much as it shouldn't happen, the law of unintended consequences rears its ugly head often enough when systems are designed for reliability first and safety a close second.
Per the NEC (National Electrical Code), this type of consideration is acceptable where losses associated with a failure could exceed the value provided by the safety improvement. The key is that only qualified people work on the system.
Ten to fifteen years ago, designing a main-tie-main switchboard for closed-transition operation, you would assume the duty rating to be that of the single-ended configuration, not parallel operation of both feeders. It was standard practice. Today, an overlapping transfer switch (performing the same basic function, only 1,000x faster) is considered to be a permanent tie between systems, so the short-circuit duty rating of the equipment must be considerably higher.
The reason for the change is that a theoretical potential for a problem turned into a real problem one or two times, and "best practices" changed.
This happens constantly. Underlying assumptions change in subtle ways, and an unforeseen (or purely academic) risk becomes real.
I have a client that won't do an upgrade that we have recommended, at least in a time-line I am comfortable with. I know that this failure is likely to lead to a major outage, and I cross my fingers that nobody gets hurt when it happens. Other clients have similar nagging problems that they are aware of, but you can't fix everything in a day. Some problems take several years to get repaired. Welcome to the world of capital planning.
This isn't that uncommon with a 200kAIC board with air-power breakers, if there is a bolted fault. Instantaneous delays. Newer insulated-case style breakers all have an instantaneous override which will limit fault energy,
The other possibility was that a tie was closed and the breakers over-dutied and could not clear the fault.
Odd that nobody was hurt though; spontaneous shorts are very rare-- most involve either switching or work in live boards, either of which would kill someone.
Insurance doesn't usually do much in these cases-- it might cover equipment but not labor at double overtime. It is easy to spend $500k just in engineering and stop-gap repairs on an electical-only incident. Fire would usually cause twice that much damage for stop-gap repairs.
Actually, someone is making the closed-circuit cooling towers with coils oversized for dry operation again. The one company doing it died in the dot-bomb, but someone more mainstream has picked it up again.
This type of system is really interesting when you look at optimizing power and water usage dynamically based on rates, approach, and plant efficiency, as there is no loss in reliability. (just cost...)
It sounds more like a cult of personality claiming they are high-tech. Their cold-aisle containment system has been around for years and produced by several vendors and engineers. It does work pretty well; our CFD models suggested a 30-50% reduction in fan energy.
There are a lot of neat ways to provide cooling, but unless you get heat directly off the chip to a coolant, it isn't easy to do anything but evaporative cooling in the desert... at least not without a huge thermal mass.
My calculation is closer to 2 million gallons, assuming they do everything they can to reduce drift and operate at a very high delta-t. You could also drop that down by half most of the year if you use the cooling towers as dry coolers at night.
Still, that is a huge amount of water storage even to run for a single day. Usually we try to have 72 hours of water storage on-site. You can use the dry coolers (or air cooled chillers), but the efficiency goes to hell in the Las Vegas climate.
For a more human scale of the water needs, it is about a 10" pipe running full out all day just to accommodate evaporation, drift, and blow-down.
The existing installer.app VT100 terminal is functional, and about as good as it will get with the touchscreen. I've used a Blackberry, Palm, N770, and my iPhone. The experience is comparable for command line on all of them, but none of them are worth a squat when I go to use nano or lynx for whatever reasons.
It really depends on what you need to do with it. If you just want to check security logs or do a manual backup, the iPhone is good enough.
According to the IRS, all income is profit until you document expenses. Much like stock sales.
Umm... have you tried browsing /. with Lynx?!
People complain how bad it looks in Firefox... it manages to look worse in Lynx.
We usually see 0.3-1% failure rates on servers in a data center powerdown (scheduled, soft shutdown). Bigger issues with blades and high power 1U servers, fewer with midrange and mainframe equipment. Hard drive and power supply failures are most common, generally attributed to thermal shock on re-start.
Would love to see any information that can debunk so we can hit the equipment manufacturers up for damages...
Personally, in my office I have a mad-scientist corner. We're engineers, and it is fun to actually make things rather than just design them all the time. I think over the past 30-some years of my life I have owned a dozen soldering irons and never really took a serious interest in learning how to actually do a damn thing with them. A couple years ago, we couldn't find something that met our needs, so we designed something to integrate different readily-available parts, and put it all together ourselves.
I'd say the key is to have enough cash to spend on stuff, not fearing breaking things (goes with part one), have a project that is important to you for starting off, and get a big old DigiKey and Allied Electronics catalog for ideas. Also, practice soldering on the cheap stuff if your natural talent for it is as bad as mine...
First off, the lineman will always ground out a medium voltage system prior to doing any work on it. That is the only safe way to work on the system.
This has the added benefit of shorting out your non-grid-tie inverter, which will eliminate the problem at the source.
On the low voltage side, any electrician (or geek) that doesn't have a voltage detector (neon bulb screwdriver or the nicer non-contact Fluke) units is a fool asking for a shock.
You can use a grid-tie system for independence from the utility's glitches. It is a conveinence thing, not a substitute for a UPS. Usually you will either need a contactor or a shunt-trip breaker that is manually reset when power is restored.
It isn't as energy efficient, but it provides a small base load rather than an impulse load to your personal "grid."
As to the article's fundamental question... it's called a blocking diode. Have a 120VDC or 48VDC power supply for the home that you tap off of for your needs. As long as you aren't trying to sell power back to the utility, this is the easiest way to make things work, and considerably cheaper. You do have an incentive to have some batteries in this configuration, though (on a separate DC:DC charging system ideally).
Power is generally not a huge issue, as containerized generators are fairly standardized and rapidly deployable. Cooling on the other hand is where most of these solutions fail. A large air cooled generator can be shipped in on skids, but will usually require an external pump skid to get things operational.
The problem with integrating cooling solutions is that it is too climate dependent to standardize. Tightly managed direct evaporative coolers can make a solid solution until you get to very hot, humid climates (over 75F wet bulb). At that point, you are going to have to have separate pre-cooling equipment to keep things working.
The only alternative is high-temperature heat rejection systems where you integrate a heat pipe into the chip and can work with 120F return water temperatures. Seems like fantasy at this point though...
While I used to love my old Nokia brick (most usable plain-jane cell phone I ever owned), I am at a complete loss of words with people's love of Symbian smart phones!
I tried the N95 and a couple other devices a year or two ago, and found everything to be abysmal hardware and clunky software as I went through Nokia and Sony-Ericson's flagship stores. The WinCE world was clearly worse, but offered better compatibility with the office.
But... what is great about the iPhone and some of HTC's models (by assumption rather than experience for the latter) is the huge, high-resolution screen. It really makes the difference in making the device usable on the road. I used to be a blackberry addict, and am quite happy with my transition to the iPhone.
Why are Europeans so passionate about their Symbian phones? Have they really gotten that much better, or is it just random evangelism?
My wife's bicycle was stolen at her work (directly across the street from the police station, with regular police officer foot-traffic in the building). We had fancy cameras and a close-up of the guy's face within an hour of the theft.
Did it help anything? No...
The cameras were also in plain sight, and he was especially brazen in how he went about it all.
Technology won't solve the problem.
Hard hat, clip board, tyvec suits, pointing a boom-shaped object at things while looking at a display, or rolling around a cart of strange looking tools will generally get you into plenty of places you shouldn't be going. Oddly enough, no matter how well secured the front door is (well after 9/11), there is usually a back door that can get you in.
The most surprising one is just juggling a bunch of crap and asking for the guard to let you through the man-trap.
While my purposes aren't nefarious, these things can easily be exploited.