PowerPath Is Projected To Improve Data Centers While Saving Over 1 Billion kWh Per Year
One of the fastest growing sectors of the U.S. energy market is computer data centers. According to an EPA report to Congress (1), the U.S. data center sector consumed about 61 billion kWh of electricity in 2006, which equates to about 1.5% of total US electricity consumption, at a cost of about $4.5 billion. This represents an increase of more than double in power usage since the year 2000. This report estimated a further increase of 100 billion kWh, rising to a current level of more than 2.5% of U.S. electricity consumption, at a cost of about $7.4 billion. Factoring in worldwide power consumption increases these numbers by about three times.
The Need for High Power Quality
Data centers require a very high level of power quality. This is driven by the high cost of downtime in mission-critical computer systems. For example, if the servers supporting the New York Stock Exchange were to suffer even a brief outage, the resulting loss of business would amount to more than $150 million per minute (2). This potential liability makes it easy to justify almost any expense that will materially improve the reliability of these power systems. Yet in spite of these measures, it has been estimated that over the next five years power losses will still halt data center operations at 90% of all businesses at one time or another (3). So there is still a strong need for significant improvement in power system reliability.
Maintaining an acceptable level of power quality for these critical applications has typically been accomplished through the use of redundant uninterruptible power supply (UPS) systems. In the event of a failure in one of these already reliable UPS systems, a fast automatic power switch known as a Static Transfer Switch (STS) is employed to automatically switch to an alternate UPS system. This makes the STS the “last line of defense” in protecting the data center power supply, and thus the single most critical element. These switches are fast enough to detect a power outage and transfer to the alternate power source within 4 milliseconds or less, fast enough to prevent any interruption in computer operations.
The STS employs semiconductor switches using Silicon Controlled Rectifiers (SCRs). These devices can switch on at microsecond speeds, but suffer from an inherent conduction voltage drop that results in significant heat lost and waste of power at all times that output power is required. They also cannot be electronically turned off instantaneously which makes the control system design critical to prevent catastrophic failure.
There is also a slower class of switch called the Automatic Transfer Switch (ATS), which is widely used for non-critical applications. These switches are smaller, simpler, cheaper, and more robust than an STS. This is due to their use of mechanical rather than semiconductor switching. This also results in lower conduction losses, making them much more efficient. But the downside is that they typically switch at speeds of 100mSec or slower, more than an order of magnitude too slow for critical data center applications.
Our New Technology
Our proposal is to develop a new type of electro-mechanical switch that we call the Fast Transfer Switch (FTS) that is capable of switching at very high speed, combining the speed, reliability, and simple control advantages of a mechanical switch with the high speed of a semiconductor switch. This new switch will be based on an entirely new type of linear actuator / motor that we have invented and filed a patent on, which is capable of significantly faster speeds than any existing switch technology. It is also very simple and robust, as the moving section is simply a piece of lightweight conductive metal, with no magnets or windings attached. Power is transferred to the moving section via a wireless magnetic induction method, so there are no power leads or brushes to wear out.
Although the STS is an obscure device, its impact on the nation’s power usage is substantial, since most data center power (accounting as we have seen for a significant amount of all power usage) passes through at least one of these devices. We estimate that the improvement in data center efficiency that would be seen by widespread adoption of our new technology would be about 0.5%, which would correspond to a savings of up to 500 million kWh per year in electricity in the U.S., saving $37 million annually using the EPA’s latest estimate.
Although these savings alone are enough to justify support for this project, the new invention also brings a number of additional advantages. The new switch will be much more reliable since it requires no cooling fans, is easier to control and maintains a good connection even if its control circuits lose power or fail entirely. This will result in a large economic benefit by preventing the business losses we described earlier. The new switch will also have a much smaller size which will save on expensive data center floor space, a substantial direct cost savings.
As we have seen, the power in a typical data center is supplied by UPS systems and then passes through an STS. The UPS systems also have an output power switch similar to an STS that bypasses the UPS in the event of a failure. By applying our technology to this internal UPS switch, we would produce a power savings similar to the savings we get by replacing existing STS's. This raises the total power savings we would be able to achieve by applying our technology to data centers, up to more than 1 Billion kWh per year.
In addition, the improved speed and efficiency of the new technology could bring sweeping improvements to other electrical power applications, including:
These applications would also result in significant additional power savings.
Footnotes:
The Need for High Power Quality
Data centers require a very high level of power quality. This is driven by the high cost of downtime in mission-critical computer systems. For example, if the servers supporting the New York Stock Exchange were to suffer even a brief outage, the resulting loss of business would amount to more than $150 million per minute (2). This potential liability makes it easy to justify almost any expense that will materially improve the reliability of these power systems. Yet in spite of these measures, it has been estimated that over the next five years power losses will still halt data center operations at 90% of all businesses at one time or another (3). So there is still a strong need for significant improvement in power system reliability.
Maintaining an acceptable level of power quality for these critical applications has typically been accomplished through the use of redundant uninterruptible power supply (UPS) systems. In the event of a failure in one of these already reliable UPS systems, a fast automatic power switch known as a Static Transfer Switch (STS) is employed to automatically switch to an alternate UPS system. This makes the STS the “last line of defense” in protecting the data center power supply, and thus the single most critical element. These switches are fast enough to detect a power outage and transfer to the alternate power source within 4 milliseconds or less, fast enough to prevent any interruption in computer operations.
The STS employs semiconductor switches using Silicon Controlled Rectifiers (SCRs). These devices can switch on at microsecond speeds, but suffer from an inherent conduction voltage drop that results in significant heat lost and waste of power at all times that output power is required. They also cannot be electronically turned off instantaneously which makes the control system design critical to prevent catastrophic failure.
There is also a slower class of switch called the Automatic Transfer Switch (ATS), which is widely used for non-critical applications. These switches are smaller, simpler, cheaper, and more robust than an STS. This is due to their use of mechanical rather than semiconductor switching. This also results in lower conduction losses, making them much more efficient. But the downside is that they typically switch at speeds of 100mSec or slower, more than an order of magnitude too slow for critical data center applications.
Our New Technology
Our proposal is to develop a new type of electro-mechanical switch that we call the Fast Transfer Switch (FTS) that is capable of switching at very high speed, combining the speed, reliability, and simple control advantages of a mechanical switch with the high speed of a semiconductor switch. This new switch will be based on an entirely new type of linear actuator / motor that we have invented and filed a patent on, which is capable of significantly faster speeds than any existing switch technology. It is also very simple and robust, as the moving section is simply a piece of lightweight conductive metal, with no magnets or windings attached. Power is transferred to the moving section via a wireless magnetic induction method, so there are no power leads or brushes to wear out.
Although the STS is an obscure device, its impact on the nation’s power usage is substantial, since most data center power (accounting as we have seen for a significant amount of all power usage) passes through at least one of these devices. We estimate that the improvement in data center efficiency that would be seen by widespread adoption of our new technology would be about 0.5%, which would correspond to a savings of up to 500 million kWh per year in electricity in the U.S., saving $37 million annually using the EPA’s latest estimate.
Although these savings alone are enough to justify support for this project, the new invention also brings a number of additional advantages. The new switch will be much more reliable since it requires no cooling fans, is easier to control and maintains a good connection even if its control circuits lose power or fail entirely. This will result in a large economic benefit by preventing the business losses we described earlier. The new switch will also have a much smaller size which will save on expensive data center floor space, a substantial direct cost savings.
As we have seen, the power in a typical data center is supplied by UPS systems and then passes through an STS. The UPS systems also have an output power switch similar to an STS that bypasses the UPS in the event of a failure. By applying our technology to this internal UPS switch, we would produce a power savings similar to the savings we get by replacing existing STS's. This raises the total power savings we would be able to achieve by applying our technology to data centers, up to more than 1 Billion kWh per year.
In addition, the improved speed and efficiency of the new technology could bring sweeping improvements to other electrical power applications, including:
- Faster and more reliable circuit breakers
- Fault current limiting during short circuit
- The sizable non-data center STS market, such as industrial processes, prisons, hospitals and clinics, etc.
- Replacing existing ATSs with better performance.
These applications would also result in significant additional power savings.
Footnotes:
- Report to Congress on Server and Data Center Energy Efficiency – Public Law 109-431, U.S. Environmental Protection Agency, August 2, 2007.
- NYSE Group Volume in All Stocks Traded, July 2009, as reported on the NYSEData.com Factbook.
- Five Bold Predictions for the Data Center Industry that will Change YOUR Future, Data Center Institute, AFCOM 2006.
