Forecasting Failure of Power Supply in a Data Center

Abstract

An apparatus comprising a power supply, which has first and second components, and actuarial circuitry in communication with the power supply. The actuarial circuitry forecasts the power supply's life expectancy based on real-time measurements of operational parameters of said first and second components.

Claims

1. An apparatus comprising a power supply and actuarial circuitry in communication with said power supply, said power supply comprising first and second components, wherein said actuarial circuitry is configured to forecast life expectancy of said power supply based at least in part on real-time measurements of operational parameters of said first and second components.

2. The apparatus of claim 1, wherein said first and second components comprise a fan and an electrolytic capacitor.

3. The apparatus of claim 1, wherein said first and second components comprise electrolytic capacitors.

4. The apparatus of claim 1, further comprising databases that provide models of said first and second components.

5. The apparatus of claim 1, wherein said power supply is configured to provide first and second data streams to said actuarial circuitry, said first and second data streams comprising said real-time measurements.

6. The apparatus of claim 1, wherein said actuarial circuitry comprises first and second mortality forecasting circuits that receive real-time data from said power supply and forecast life expectancies of said corresponding first and second components.

7. The apparatus of claim 1, wherein said actuarial circuitry is configured to receive a capacitor data-stream and a fan data-stream from said power supply, said power supply having an electrolytic capacitor and a fan, wherein said actuarial circuitry comprises a real-time fan-mortality forecasting circuit that receives a stochastic fan model and a real-time capacitor-mortality forecasting circuit that receives a stochastic capacitor model, wherein, based on outputs of said real-time fan-mortality forecasting circuit and said real-time capacitor-mortality forecasting circuit, said actuarial circuitry outputs a plurality of voltages selected from the group consisting of a first voltage and a second voltage, said plurality of voltages forming a pattern that is indicative of a power-supply life expectancy that serves as a basis for scheduling replacement of said power supply.

8. The apparatus of claim 1, wherein said actuarial circuitry comprises a comparator that receives forecasted life expectancies of components of said power supply, said components comprising said first and second components, and outputs the smallest of said forecasted life expectancies.

9. The apparatus of claim 1, further comprising a first sensor system coupled to said first component for measuring said operational parameters of said first component and a second sensor system coupled to said second component for measuring said operational parameters of said second component.

10. The apparatus of claim 1, wherein said first component is a fan, said apparatus further comprising a thermometer for measuring temperature of said fan, a Hobbs meter that is coupled to measure accumulated time during which said fan has been used, and a tachometer to measure the number of rotations said fan makes per unit time.

11. The apparatus of claim 1, wherein said second component is an electrolytic capacitor, said apparatus further comprising a thermometer for measuring temperature of said electrolytic capacitor, a voltmeter configured to measure voltage across said electrolytic capacitor, and an ammeter configured to measure current through said electrolytic capacitor.

12. The apparatus of claim 1, wherein said power supply is one of a plurality of power supplies within a data center and wherein actuarial circuitry is configured to forecast said life expectancy based at least in part on information obtained from other power supplies within said data center.

13. The apparatus of claim 1, wherein said actuarial circuitry is configured to forecast said life expectancy based at least in part on information obtained from a manufacturer of said power supply, said information being indicative of life expectancies of power supplies similar to said power supply.

14. The apparatus of claim 1, wherein said first and second components are the components with the shortest expected lifetimes.

15. The apparatus of claim 1, wherein said actuarial circuitry comprises a non-generic computer system.

16. A non-abstract method comprising obtaining real-time measurements of first and second sets of operational parameters, wherein said first set comprises operational parameters of a first component of said power supply and said second set comprises operational parameters of a second set of a power supply that supplies power to IT equipment, forecasting an expected lifetime of said first component, forecasting an expected lifetime of said second component, based at least in part on said expected lifetimes of said first and second components, forecasting said life expectancy of said power supply, determining that said life expectancy of said power supply is lower than a threshold, and providing an alert to draw attention to said life expectancy.

17. The method of claim 16, further comprising selecting the first component to be a fan and selecting the second component to be an electrolytic capacitor.

18. The method of claim 16, wherein all steps wherein all steps are carried out in a non-abstract manner outside the human mind and in a non-generic computer system.

Description

DESCRIPTION OF DRAWINGS

[0024] FIG. 1 shows a power supply with an accompanying actuarial circuit;

[0025] FIG. 2 shows the actuarial circuit of FIG. 1 installed in a data center;

[0026] FIG. 3 shows power supplies in the data center of FIG. 2 connected to the actuarial circuit of FIG. 1; and

[0027] FIG. 4 shows a method carried out by the actuarial circuit of FIG. 1

DETAILED DESCRIPTION

[0028] FIG. 1 shows a power supply 10 having a fan 12, an electrolytic capacitor 14, a fan sensor-system 16, which is coupled to the fan 12, and a capacitor sensor-system 18, which is coupled to the capacitor 14.

[0029] The fan sensor-system 16 measures operational parameters associated with the fan's operation. Such parameters include how long the fan 12 has been running, the temperature in the fan's vicinity, and the fan's speed. These parameters define measured time-varying functions that are provided as a fan data-stream 20 to an actuarial circuit 22.

[0030] To sense these operational parameters, the fan sensor-system 16 includes a fan thermometer 54 that measures the fan's temperature, a Hobbs meter 56 that measures the accumulated time of the fan's usage, and a tachometer 58 to measure the number of revolutions made by the fan 12 per unit time.

[0031] Similarly, the capacitor sensor-system 20 measures operational parameters associated with the capacitor's operation. Such parameters include the voltage across the capacitor 14, including the sign of that voltage, the current into the capacitor 14, including the sign of that current, and the temperature of the capacitor's casing These parameters define three additional measured time-varying functions that are provided as a capacitor data-stream 24 to the actuarial circuit 22.

[0032] To sense these operational parameters, capacitor sensor-system 18 includes a capacitor thermometer 60 that measures the capacitor's temperature, a voltmeter 62 that measures voltage across the capacitor 14, and an ammeter 64 that that measures current into the capacitor 14, thus providing a way to determine the capacitor's internal resistance.

[0033] The actuarial circuit 22 includes a real-time fan-mortality forecasting circuit 26 and a real-time capacitor-mortality forecasting circuit 28. The real-time fan-mortality forecasting circuit 26 receives both the fan data-stream 20 and a stochastic fan-model 30 from a fan database 32. The real-time capacitor-mortality forecasting circuit 28 receives both the capacitor data-stream 24 and a stochastic capacitor-model 34 from a capacitor database 36.

[0034] The stochastic fan-model 30 is typically supplied by the fan's manufacturer. Such a model reflects the manufacturer's efforts at predicting the fan's remaining lifetime based on parameters associated with the fan's operation. Similarly, the stochastic capacitor-model 34 is typically supplied by the capacitor's manufacturer and reflects the manufacturer's efforts at predicting the capacitor's remaining lifetime based on parameters associated with the capacitor's operation.

[0035] The real-time fan-mortality forecasting circuit 26 uses the fan data-stream 20 and the stochastic fan-model 30 to dynamically adjust the fan's life expectancy 38 in real time. Similarly, the real-time capacitor-mortality forecasting circuit 28 uses the capacitor data-stream 24 and the stochastic capacitor-model 34 to dynamically adjust the capacitor's life expectancy 40 in real time.

[0036] In some embodiments, the stochastic fan-model 30 comprises data collected from similar fans that have failed in the past. This data is keyed to life expectancy. Thus, by comparing the fan data-stream 20 with the data stored that comprise the stochastic fan-model 30, it is possible for the real-time fan-mortality forecasting circuit 26 to forecast life expectancy for the fan 12 with reasonable accuracy.

[0037] In some embodiments, the stochastic capacitor-model 34 comprises data collected from similar capacitors that have failed in the past. This data is keyed to life expectancy. Thus, by comparing the capacitor data-stream 24 with the data stored that comprise the stochastic capacitor-model 34, it is possible for the real-time capacitor-mortality forecasting circuit 28 to forecast life expectancy for the capacitor 14 with reasonable accuracy.

[0038] In other embodiments, the stochastic capacitor-model 34 forecasts the capacitor's life expectancy based on observations of voltage applied across the capacitor 14, its temperature, and detection of either ripples in the current flowing into the capacitor 14, and/or current surges into the capacitor 14. These observed while the power supply 10 is in operation and are used as a basis for determining a deduction in life expectancy from the capacitor's initial expected lifetime.

[0039] The real-time capacitor-mortality forecasting circuit 28 provides the capacitor's life expectancy 40 to a comparator 42. Similarly, the real-time fan-mortality forecasting circuit 26 provides the fan's life expectancy 38 to the comparator 42. The comparator outputs a plurality of voltages, the pattern of which is indicative of the smaller of the fan's life expectancy 38 and the capacitor's life expectancy 40. This pattern of voltages thus serves as a forecasted power-supply life expectancy 44. This forecasted power-supply life expectancy 44 provides a basis for scheduling replacement of power supplies at an opportune time that is before the probability of failure rises past a selected threshold.

[0040] FIG. 2 shows an internet data center 46 having a data room 48 that holds IT equipment that comprises power consumers 50, such as servers 52 and routers 54. These are often mounted on racks 56. In some embodiments, the power supply 10 and its accompanying actuarial circuit 22 operate within the internet data center 46.

[0041] The data center 46 receives ac power transmitted from a utility 54. This ac power is generally unsuited to operating the power consumers 50 and must be converted into suitable dc power. The power supply 10 carries out this function. In a typical data center 46, there are many such power supplies 10, each of which converts ac power into suitable dc power. Such power supplies 10 are often referred to as ac/dc converters.

[0042] In some embodiments, the actuarial circuit 22 is local to the power supply 10. Thus, each power supply 10 would have its own actuarial circuit 22. However, in other embodiments, the actuarial circuit 22 is a shared resource that connects to both the power supply 10 and to a plurality of other power supplies. Among these are embodiments in which the actuarial circuit 22 is shared by all power supplies 10 in the data center 46, as shown in FIG. 3.

[0043] In some embodiments, the actuarial circuit 22 receives information indicative of other properties of the power supply, such as its date of manufacture or its location in the data center 46. Information concerning a power supply's date of manufacture may be useful if, for example, it is known that power supplies manufactured during a particular interval possess lifetimes that differ from those manufactured in other intervals. Knowledge of location within a data center 46 is useful for recognizing flaws in the data center's cooling system. For example, if it is determined that all power supplies in a particular wing of the data center 46 are running hot, one can infer that something may be wrong with the cooling system in that wing and therefore dispatch maintenance personnel accordingly.

[0044] FIG. 4 shows a method for announcing a recommendation to replace a power supply 10 that begins with obtaining measurements of operational parameters of two or more components of the power supply (step 66) and using those measurements as a basis for dynamically forecasting their respective expected lifetimes (steps 68, 70). This is followed by forecasting the power supply's life expectancy (step 72) based on the forecasted expected lifetimes and a comparison between the power supply's life expectancy and some user-defined threshold (step 74). If the power supply's life expectancy falls short of the threshold, an alert is sounded (step 76). Otherwise, further measurements of operational parameters are obtained (step 66).

[0045] The embodiments described herein illustrate the principles of the invention without limiting its scope, which is defined by the scope of the appended claims.