Two generators serving the load of one provides a 6x improvement in availability. 3/2 gives you about 3x, 4/3 2x, and 5/4 1.5x availability improvement.
So, the more generators you add, the harder it becomes to actually have an improved reliability. If you did the 5/4, to get "equal" reliability you would need to do at least 6/4. (Same math works for parallel UPS modules.) The other challenge is that to do the preventative maintenance on a running generator takes about 4 hours of cool-down, plus 6 hours of labor-- if you work fast.
Electrically, you really want to have 3MW generators for big facilities, operating in parallel at medium voltage (12.47-13.8 kV). These units are even harder to get.
Cat is just under 52 weeks for generators now, but Cummins is closer to 62 weeks right now. When you are building a big data center, this just kills you, unless all the planning has been done in advance.
The irony is that 6 years ago you could buy these units for 50% below list.
There are a couple good tricks for protecting them, stacking them like you would in a cargo facility is the first to come to mind, but Jersy Barriers on all sides would also complicate matters pretty well. Sandbagging the roof would also make it more work to get it out.
If all else fails, make a stand that they lock into on the parking lot. Those containers really are built for security and durability.
The only problem is that it is pretty small by most standards. APC has their "data center on demand" (I think it is branded Data Center Express now), but that is currently limited to 80kW. It should make it to 200kW before too long.
At 80kW, it is about $1M, and includes generator, UPS, and cooling. I forget if it is a 40' or 53' trailer, but you end up with just 8 rack-equivalents.
It's hard to believe that Sun has made a better solution in a smaller space, although the picture makes it look like the power and cooling (at least heat rejection) infrastructure are external.
While it does take two years to design and plan a Tier IV data center, most internet companies are only in Tier II facilities, which can usually be deployed in under 6 months. For one client now, we are going to just build inside a warehouse. You get most of the benefits of a container solution with less risk or vendor lock-in.
The incentive is the many companies that can't plan. Something that can be deployed quicker than a lease agreed to for CoLo, that can be sited in a spot that doesn't require a mile of trenching for a new OC12.
Today there is a dearth of quality data center space. A well-executed container-based system that allows for various equipment to be installed inside, and that can be pre-configured in a wharehouse and literally "dropped" into place (have you ever seen how they deliver containers in parking lots...) is a great infrastructure solution.
The biggest challenge is finding ways to make it scale from an application standpoint, and really maximizing the energy benefits. My company was going to use heat pipes to the chips to free-cool servers; the problem is that a solution like that doesn't meet most IT organizations needs. (Could work for a Google, but not Citibank.)
My concern is that if you go with 12V boards, you will end up needing a 380/580VDC:12VDC or 48:12V power converter for each computer, which complicates adoption of a DC standard.
If you go for the extra conversion step, you increase your losses by a percentage point at least, and there are reliability concerns (from what I am told) with the "HVDC" to "LVDC" equipment.
The best solution is to pick a voltage that can be used directly by the majority of the loads. Since 5 or 3.5 or 1.8V isn't practical, you have to evaluate the benefits of a higher voltage at a level beyond the individual computer.
I know Google's objective is to impact residential computers more than data center equipment, but history says that the best solution is one that can take advantage of both areas.
However, if you want to maximize efficiency, you are going to pick a motherboard input voltage of 42-48VDC, with large power converters at the rack (or 3-rack) level.
I fail to understand the logic of 12V for Google (although it would be nice for cars); their typical racks look like they run at about 6kW, so centralizing the power converter would result in a 500A DC bus at 12V. (250A if it is mounted at the center of the rack like their ethernet hubs).
The industry might move closer to a high-voltage DC (as in 380-480VDC or even 900VDC) over the next few years, and looking towards a 12V standard won't help matters.
Ideal voltage is all a function of what the maximum load is...
...and most importantly, for $1.5B, you could get a hell of a lot more starting from the ground up. Realistically, you could spend on the order of $500M and compete head-to-head. All YouTube is really worth is the equipment, time, and the marketing effort required to supplant their position. It's hard to imagine this having a value of more than $600M.
The term "infected" is used when people upload their own music into the player, and it is locked into DRM by default. In this instance, MS has no obligation to the labels nor any right in the consumers eyes to make their files inoperable with other machines.
Tell me, would YOU want to have a job where a mistake could cost you everything you own and 90% of what you earn for the next 20 years? We're not talking about high stakes gambling or high-risk investment here; we're talking about being a NOTARY.
I have professional liability insurance as an engineer. I pay about 8% of my gross income for the insurance, and yes, it covers errors and omissions up to what I would make over the next 30 years. If I ever make a mistake that endangers public safety, I would likely lose my engineering license.
Unfortunately, Notaries are a weak link in the system. They really don't have enough information to do their job, the cost for notarizing a document is not commensurate with the risks they could incur, and people aren't really willing to pay more for a formality that generally has little value. Either they need additional insurance, or identity verification needs to happen at multiple levels.
In this specific case, I don't know how you would avoid having multiple tiered documents that would make it impossible to prevent this type of fraud. It becomes a shell game quickly.
While I have a similar opinion to your answers, responding that way is a great way to get yourself in trouble.
It's really wrong how much power the immigration people have. I'm happy I spend most of my time in a country that is grossly corrupt and almost celebrates it, rather than a country that hides the obvious fact. It's nice to know where you stand.
I have found the easiest way to appease a spouse accusing you of paying more attention to the mistress than them is to buy them a mistress as well. They get hooked, no more complaints from them. Well, almost.
Show me a telephone number which you can dial and that, by the simple act of connection, results in the infiltration of your company's office such that your Intranet data (e.g., customer personal info, credit cards, etc) can be leaked out.
Have you not heard of Social Engineering? Same risk, possibly without the automation, but not dangerously far away.
The problem is that any kind of filtering is not likely to have the desired impact. A former employer had a Big Brother system installed, and the net result was that whenever I was in the office, I was tripped up in it regularly. The requests might have been "abnormal" in many respects, but still reasonably job-related.
And, ultimately, these systems aren't installed to protect security, they are there largely to prevent sexual harassment lawsuits based on people's lunch-hour behavior. Sad truth.
It all depends on what you need. I have had a 770 since January, and generally like it. Unfortunately, it is too slow, and data entry is painful. The keyboard options on screen are insufficient for me, and handwriting recognition is a pain. I do have a bluetooth keyboard, but it's an extra thing to have to carry around so I usually avoid it.
My wish for Nokia is to improve the browser, and port Maemo to the E61 (or is it E60). I know it would have lots of problems (maemo user interface issues at a different resolution/form factor for starters), but it would do a better job of showing support and options for the platform.
I was on a 16-hour flight two weeks ago. I was flying business class. I was given 16 oz of water by the flight attendants. I routinely bring 3-4 1.5L bottles of water with me for a long flight to help avoid the dehydration. Are the airlines going to improve service levels if they prohibit people from bringing water on board?
I admit, people (myself included) travel with way too much carry-on luggage, but banning that just destroys air travel as a means for business transportation. Short flights (say SF-LA) for a day trip suddenly become impossible, since you have an extra two hours lost to waiting for baggage, as well as additional check-in time.
Pelican boxes aren't airtight; the pressure changes would still impact a violin. It's apparently bad enough going from sea level to 8,000 ft in the pressurized cabin-- an unpressurized, unheated cargo hold would surely cause cracks. Moisture content might be able to be stabilized with the Pelican box to some degree though.
(Pelican doesn't make the boxes with manual purges anymore, do they-- just the automatic purges like on the laptop cases?)
The claims are clearly exaggerated, although understandable.
I personally wish Apple put some more effort into their power supplies, especially the new MagSafe units. At 220V, you can cook an egg on one. The DC strain relief is improved som, but they need more work on the mag-safe end of the wire. My cable has visible kinks already, after only 3 months.
They use a similar design to some of the servers today (HP); they appear to draw the same current at 120V as 220V, and dissipate the extra energy as heat. I expect more from Apple, although I understand the business drivers.
The purpose of this project was to test the viability of using DC power to eliminate multiple AC:AC, AC:DC, DC:AC, and DC:DC conversions, in order to boost efficiency for datacenters. 48V systems are limited in practical terms to under 200kW (as others have said, due to the copper involved in the higher currents). The typical datacenter they are targeting is ~2MW and above.
The main benefit of going DC today is that you can have UPS power without the terribly inefficient UPS, as DC:DC conversions can be done very efficiently. This wasn't possible before, which forced people into 48VDC or AC power.
What I never understood when this was first proposed is why not go with a higher voltage (say 960V) and use more of a 3N distribution, which would get you closer to a 400kW system size at 400A. With a 3N system, you would keep one system down for maintenance/upgrades, and rotate the systems as required for expansion, etc. Most companies today cringe at the idea of going down to "N" for service.
The biggest objection seems to be voltage regulation and limiting power consumption of each device. If your hard drive spins up, is that going to cause a problem for your iPod, etc. Proper lab-style power supplies that are lightly loaded aren't a problem, but it gets harder when you have multiple devices.
Likewise, if the wire length off the wall wart isn't known, you can't account for voltage drop-- what happens when someone tries to run 4V through a 25' extension cord?
Actually, you need more power if you are using AC, since you have to invert the DC voltage into AC.
Series/parallel combinations of batteries are a bad idea; internal resistance differences will casue significant circulating current losses.
The higher the voltage, the lower the current, so from that respect a series arrangement is better. Unfortunately, unless you are using "flooded" batteries (clear jars weighing about 300# that sit on racks 2 or 3 high), as a single cell in a single battery fails, it will fail open-circuit, and you won't get any voltage from the string. For this reason, you usually try to parallel at least two batteries (at the ends) for improved redundancy. (Exception being parallel-redundant UPS systems, where a single battery failure will only take down a single UPS module.)
What seems to be more common in expensive software is that the key is used in some kind of math within the program, so if you strip the key check, you will eventually get corrupted data in your files, things won't render properly, or whatever. This is much harder to just strip out; you end up needing to reverse engineer a significan portion of the program.
The dBA scale is weighted to higher frequency sounds, and is the basis for noise ordinances. Compare the readings with it on and off, and consult your local noise ordinance on what the limits at the property line are for above ambient. Some cities also have limits on noise levels for any specific frequency, so you could filter for a single frequency and determine the intensity for that. Usually, this is just around light industrial areas mixed with housing, but it could be worth a shot.
I saw the first Tablet that would be compelling to me last week - one of the little Fujitsu units at ~$2k. If it had built-in (or internal accessory) EDGE/HSCD capability, I would seriously consider it. If it was a Mac, I'd buy it for fun.
As the smart phones get more expensive and more (crippled) features, the market for an ultraportable tablet grows. WinCE just doesn't do it for my needs-- intrinsic limitations of battery, screen, and processor keep it from being a "laptop replacement."
I won't hold my breath, though. It has taken a long time for the tablet market to mature to this point, and I don't think Apple is ready to jump in any time soon.
Running an absorption chiller off a turbine "waste" heat is much more efficient than running a reciprocating or centrifugal chiller. This isn't that novel, aside possibly from the fact that it is a single unit. Pre-cooling coils are fairly common on co-gen turbines, and are often run off an absorption "pony" chiller.
The military is in dire need of a good APU source; they try everything. The existing turbines they use are a mess, so anything better is an improvement. This is why they have poured so much money into fuel cells over the past half-century.
Personally, I'm more impressed by Active Power's new UPS that runs off compressed air and also provides cooling.
Slashdot Mirror
Two generators serving the load of one provides a 6x improvement in availability. 3/2 gives you about 3x, 4/3 2x, and 5/4 1.5x availability improvement.
So, the more generators you add, the harder it becomes to actually have an improved reliability. If you did the 5/4, to get "equal" reliability you would need to do at least 6/4. (Same math works for parallel UPS modules.) The other challenge is that to do the preventative maintenance on a running generator takes about 4 hours of cool-down, plus 6 hours of labor-- if you work fast.
Electrically, you really want to have 3MW generators for big facilities, operating in parallel at medium voltage (12.47-13.8 kV). These units are even harder to get.
Cat is just under 52 weeks for generators now, but Cummins is closer to 62 weeks right now. When you are building a big data center, this just kills you, unless all the planning has been done in advance.
The irony is that 6 years ago you could buy these units for 50% below list.
There are a couple good tricks for protecting them, stacking them like you would in a cargo facility is the first to come to mind, but Jersy Barriers on all sides would also complicate matters pretty well. Sandbagging the roof would also make it more work to get it out.
If all else fails, make a stand that they lock into on the parking lot. Those containers really are built for security and durability.
The only problem is that it is pretty small by most standards. APC has their "data center on demand" (I think it is branded Data Center Express now), but that is currently limited to 80kW. It should make it to 200kW before too long.
At 80kW, it is about $1M, and includes generator, UPS, and cooling. I forget if it is a 40' or 53' trailer, but you end up with just 8 rack-equivalents.
It's hard to believe that Sun has made a better solution in a smaller space, although the picture makes it look like the power and cooling (at least heat rejection) infrastructure are external.
While it does take two years to design and plan a Tier IV data center, most internet companies are only in Tier II facilities, which can usually be deployed in under 6 months. For one client now, we are going to just build inside a warehouse. You get most of the benefits of a container solution with less risk or vendor lock-in.
The incentive is the many companies that can't plan. Something that can be deployed quicker than a lease agreed to for CoLo, that can be sited in a spot that doesn't require a mile of trenching for a new OC12.
Today there is a dearth of quality data center space. A well-executed container-based system that allows for various equipment to be installed inside, and that can be pre-configured in a wharehouse and literally "dropped" into place (have you ever seen how they deliver containers in parking lots...) is a great infrastructure solution.
The biggest challenge is finding ways to make it scale from an application standpoint, and really maximizing the energy benefits. My company was going to use heat pipes to the chips to free-cool servers; the problem is that a solution like that doesn't meet most IT organizations needs. (Could work for a Google, but not Citibank.)
Apparently they berthed at Pier 70 and hooked up to a power plant to provide steam to the power plant there. Interesting.
Not much on details, but some .
Only if the break horsepower is off an electric motor. Presumably, the steam is converted to mechanical power through a turbine.
Reasonably close, but at least a 10-20% penalty.
My concern is that if you go with 12V boards, you will end up needing a 380/580VDC:12VDC or 48:12V power converter for each computer, which complicates adoption of a DC standard.
If you go for the extra conversion step, you increase your losses by a percentage point at least, and there are reliability concerns (from what I am told) with the "HVDC" to "LVDC" equipment.
The best solution is to pick a voltage that can be used directly by the majority of the loads. Since 5 or 3.5 or 1.8V isn't practical, you have to evaluate the benefits of a higher voltage at a level beyond the individual computer.
I know Google's objective is to impact residential computers more than data center equipment, but history says that the best solution is one that can take advantage of both areas.
However, if you want to maximize efficiency, you are going to pick a motherboard input voltage of 42-48VDC, with large power converters at the rack (or 3-rack) level.
I fail to understand the logic of 12V for Google (although it would be nice for cars); their typical racks look like they run at about 6kW, so centralizing the power converter would result in a 500A DC bus at 12V. (250A if it is mounted at the center of the rack like their ethernet hubs).
The industry might move closer to a high-voltage DC (as in 380-480VDC or even 900VDC) over the next few years, and looking towards a 12V standard won't help matters.
Ideal voltage is all a function of what the maximum load is...
...and most importantly, for $1.5B, you could get a hell of a lot more starting from the ground up. Realistically, you could spend on the order of $500M and compete head-to-head. All YouTube is really worth is the equipment, time, and the marketing effort required to supplant their position. It's hard to imagine this having a value of more than $600M.
But, there's a sucker born every minute...
The term "infected" is used when people upload their own music into the player, and it is locked into DRM by default. In this instance, MS has no obligation to the labels nor any right in the consumers eyes to make their files inoperable with other machines.
I have professional liability insurance as an engineer. I pay about 8% of my gross income for the insurance, and yes, it covers errors and omissions up to what I would make over the next 30 years. If I ever make a mistake that endangers public safety, I would likely lose my engineering license.
Unfortunately, Notaries are a weak link in the system. They really don't have enough information to do their job, the cost for notarizing a document is not commensurate with the risks they could incur, and people aren't really willing to pay more for a formality that generally has little value. Either they need additional insurance, or identity verification needs to happen at multiple levels.
In this specific case, I don't know how you would avoid having multiple tiered documents that would make it impossible to prevent this type of fraud. It becomes a shell game quickly.
While I have a similar opinion to your answers, responding that way is a great way to get yourself in trouble.
It's really wrong how much power the immigration people have. I'm happy I spend most of my time in a country that is grossly corrupt and almost celebrates it, rather than a country that hides the obvious fact. It's nice to know where you stand.
Regime Change for the USA!!!
(My Blackberry.)
I have found the easiest way to appease a spouse accusing you of paying more attention to the mistress than them is to buy them a mistress as well. They get hooked, no more complaints from them. Well, almost.
Show me a telephone number which you can dial and that, by the simple act of connection, results in the infiltration of your company's office such that your Intranet data (e.g., customer personal info, credit cards, etc) can be leaked out.
Have you not heard of Social Engineering? Same risk, possibly without the automation, but not dangerously far away.
The problem is that any kind of filtering is not likely to have the desired impact. A former employer had a Big Brother system installed, and the net result was that whenever I was in the office, I was tripped up in it regularly. The requests might have been "abnormal" in many respects, but still reasonably job-related.
And, ultimately, these systems aren't installed to protect security, they are there largely to prevent sexual harassment lawsuits based on people's lunch-hour behavior. Sad truth.
It all depends on what you need. I have had a 770 since January, and generally like it. Unfortunately, it is too slow, and data entry is painful. The keyboard options on screen are insufficient for me, and handwriting recognition is a pain. I do have a bluetooth keyboard, but it's an extra thing to have to carry around so I usually avoid it.
My wish for Nokia is to improve the browser, and port Maemo to the E61 (or is it E60). I know it would have lots of problems (maemo user interface issues at a different resolution/form factor for starters), but it would do a better job of showing support and options for the platform.
I was on a 16-hour flight two weeks ago. I was flying business class. I was given 16 oz of water by the flight attendants. I routinely bring 3-4 1.5L bottles of water with me for a long flight to help avoid the dehydration. Are the airlines going to improve service levels if they prohibit people from bringing water on board?
I admit, people (myself included) travel with way too much carry-on luggage, but banning that just destroys air travel as a means for business transportation. Short flights (say SF-LA) for a day trip suddenly become impossible, since you have an extra two hours lost to waiting for baggage, as well as additional check-in time.
Pelican boxes aren't airtight; the pressure changes would still impact a violin. It's apparently bad enough going from sea level to 8,000 ft in the pressurized cabin-- an unpressurized, unheated cargo hold would surely cause cracks. Moisture content might be able to be stabilized with the Pelican box to some degree though.
(Pelican doesn't make the boxes with manual purges anymore, do they-- just the automatic purges like on the laptop cases?)
The claims are clearly exaggerated, although understandable.
I personally wish Apple put some more effort into their power supplies, especially the new MagSafe units. At 220V, you can cook an egg on one. The DC strain relief is improved som, but they need more work on the mag-safe end of the wire. My cable has visible kinks already, after only 3 months.
They use a similar design to some of the servers today (HP); they appear to draw the same current at 120V as 220V, and dissipate the extra energy as heat. I expect more from Apple, although I understand the business drivers.
The purpose of this project was to test the viability of using DC power to eliminate multiple AC:AC, AC:DC, DC:AC, and DC:DC conversions, in order to boost efficiency for datacenters. 48V systems are limited in practical terms to under 200kW (as others have said, due to the copper involved in the higher currents). The typical datacenter they are targeting is ~2MW and above.
The main benefit of going DC today is that you can have UPS power without the terribly inefficient UPS, as DC:DC conversions can be done very efficiently. This wasn't possible before, which forced people into 48VDC or AC power.
What I never understood when this was first proposed is why not go with a higher voltage (say 960V) and use more of a 3N distribution, which would get you closer to a 400kW system size at 400A. With a 3N system, you would keep one system down for maintenance/upgrades, and rotate the systems as required for expansion, etc. Most companies today cringe at the idea of going down to "N" for service.
The biggest objection seems to be voltage regulation and limiting power consumption of each device. If your hard drive spins up, is that going to cause a problem for your iPod, etc. Proper lab-style power supplies that are lightly loaded aren't a problem, but it gets harder when you have multiple devices.
Likewise, if the wire length off the wall wart isn't known, you can't account for voltage drop-- what happens when someone tries to run 4V through a 25' extension cord?
Actually, you need more power if you are using AC, since you have to invert the DC voltage into AC.
Series/parallel combinations of batteries are a bad idea; internal resistance differences will casue significant circulating current losses.
The higher the voltage, the lower the current, so from that respect a series arrangement is better. Unfortunately, unless you are using "flooded" batteries (clear jars weighing about 300# that sit on racks 2 or 3 high), as a single cell in a single battery fails, it will fail open-circuit, and you won't get any voltage from the string. For this reason, you usually try to parallel at least two batteries (at the ends) for improved redundancy. (Exception being parallel-redundant UPS systems, where a single battery failure will only take down a single UPS module.)
What seems to be more common in expensive software is that the key is used in some kind of math within the program, so if you strip the key check, you will eventually get corrupted data in your files, things won't render properly, or whatever. This is much harder to just strip out; you end up needing to reverse engineer a significan portion of the program.
The dBA scale is weighted to higher frequency sounds, and is the basis for noise ordinances. Compare the readings with it on and off, and consult your local noise ordinance on what the limits at the property line are for above ambient. Some cities also have limits on noise levels for any specific frequency, so you could filter for a single frequency and determine the intensity for that. Usually, this is just around light industrial areas mixed with housing, but it could be worth a shot.
I saw the first Tablet that would be compelling to me last week - one of the little Fujitsu units at ~$2k. If it had built-in (or internal accessory) EDGE/HSCD capability, I would seriously consider it. If it was a Mac, I'd buy it for fun.
As the smart phones get more expensive and more (crippled) features, the market for an ultraportable tablet grows. WinCE just doesn't do it for my needs-- intrinsic limitations of battery, screen, and processor keep it from being a "laptop replacement."
I won't hold my breath, though. It has taken a long time for the tablet market to mature to this point, and I don't think Apple is ready to jump in any time soon.
Running an absorption chiller off a turbine "waste" heat is much more efficient than running a reciprocating or centrifugal chiller. This isn't that novel, aside possibly from the fact that it is a single unit. Pre-cooling coils are fairly common on co-gen turbines, and are often run off an absorption "pony" chiller.
The military is in dire need of a good APU source; they try everything. The existing turbines they use are a mess, so anything better is an improvement. This is why they have poured so much money into fuel cells over the past half-century.
Personally, I'm more impressed by Active Power's new UPS that runs off compressed air and also provides cooling.