JP2015148914A - Power efficiency optimization information processing device due to selection of running power - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve efficiency of AC-DC power conversion when AC-DC conversion efficiency of an AC power unit used in an information processing device has a curve with a power load 50% as a peak and the AC-DC conversion efficiency thereof at a time of a low power load is low.SOLUTION: An information processing device 1 in which DC power units 2A to 2C and 3A to 3C different in a power source capacity are redundantly and selectably loaded, and power is supplied to a load from the selected DC power unit comprises a management processor that monitors an output power of each DC power unit and controls a running state of each DC power unit. The management processor is configured to: calculate power consumption of the information processing device on the basis of the output power obtained from each DC power unit; and select any of the DC power units of the power source capacity to be run in accordance with the calculated power consumption.

Description

  The present invention realizes optimization of power conversion efficiency by selecting an operating power source in an information processing apparatus.

2. Description of the Related Art In recent years, power saving technology has been attracting attention and developing in information processing apparatuses such as servers.
In particular, the power saving function of the CPU is remarkably developed, and the difference in power consumption between the standby state and the high load state is very large. Further, in order to guarantee stable operation of the information processing apparatus, it is necessary to select a DC power supply unit having a capacity that guarantees operation in a high load state. On the other hand, the AC-DC power conversion efficiency of the DC power supply unit is generally low power efficiency when the load is low, and sufficient power efficiency cannot be exhibited during standby.

  Patent Document 1 (Japanese Patent Application Laid-Open No. 2008-125349) discloses a method in which a plurality of power supply units having a low power supply capacity are mounted, and power is supplied from the plurality of power supply units when the power that can be supplied alone is exceeded. ing. However, it is only possible to install a plurality of power supplies with the same power supply capacity, and improvement in power efficiency cannot be expected. Patent Document 2 (Japanese Patent Laid-Open No. 2007-60847) discloses a method of mounting a battery, charging the battery at a low load, and supplying power from the battery when the power supply capacity exceeds the power supply capacity. However, the amount of electricity stored in the battery is limited, and stable power supply cannot be implemented.

JP 2008-125349 A JP 2007-60847

  Except for large-scale computing systems such as academic applications, a general system rarely always has a high computing load state, and the computing load varies depending on the time zone. Along with the difference in the calculation load factor, a difference between the power consumption of the system device at the time of high load and low load also occurs. Furthermore, in recent years, attention has been paid to a technology for reducing power consumption at low load, and a power saving function of the CPU and a technology for reducing the rotational speed control of the system FAN when the heat generation of the information processing device is low have been developed. . That is, the difference in power consumption between high load and low load is increasing.

Normally, it is necessary to select a DC power supply unit having a power supply capacity that sufficiently satisfies power consumption at a high load in consideration of operation at a high load as a DC power supply unit used in an information processing apparatus. On the other hand, the AC-DC power conversion efficiency of the DC power supply unit generally has a curve with a peak at 50% of the power load, and the AC-DC power conversion efficiency when the power load is low is extremely low. Even in recent high-efficiency DC power supply units, improvement in AC-DC power conversion efficiency with a low load factor is small.
In other words, at a low load that occupies a lot of time in actual operation, the DC power supply unit is operated with a low AC-DC power conversion efficiency, resulting in a large power loss.

  In the present invention, a method of selecting an appropriate operating state of the DC power supply unit according to the load status of the information processing apparatus is shown. It is assumed that a system having redundancy in the power supply configuration and a DC power supply unit having a different power supply capacity are mounted. It aims at the efficiency improvement of AC-DC power conversion by selecting the direct-current power supply unit which operates appropriately.

  DC power supply units having different power supply capacities are mounted redundantly and selectably, and an information processing apparatus that supplies power to the load from the selected DC power supply unit, monitors the output power of each DC power supply unit, and A management processor for controlling the operating state of the DC power supply unit; the management processor calculates the power consumption of the information processing apparatus based on the output power acquired from each DC power supply unit; and according to the calculated power consumption Select which power capacity of the DC power supply unit to operate.

  According to the present invention, in the system device having the redundancy of the DC power supply unit, it is possible to improve the AC-DC power conversion efficiency by controlling the operating state of the DC power supply unit according to the load of the system device.

It is an image figure about the power supply efficiency used as the objective of this invention. It is a power system conceptual diagram of the present invention. It is a power supply conceptual diagram at the time of power-off of the present invention. It is a power supply control signal system diagram of the present invention. It is a load limiting signal system diagram of the present invention. It is a flowchart which shows an example of selection of the operation power supply which is one Embodiment of this invention. It is a conceptual diagram which shows coefficient selection in this invention. It is a time chart at the time of switching operation | movement to the power supply from which a capacity | capacitance is one Embodiment of this invention. It is a flowchart which shows an example of the process at the time of abnormality detection of the power supply unit which is one Embodiment of this invention. It is the table which showed the operating state of the power supply unit in the main operating state and the state of the load control signal in this invention. It is an example of the register | resistor which stores the operating state of the power supply unit in this invention.

  Embodiments of the present invention will be described below with reference to the drawings.

  As shown in FIG. 1, the power efficiency generally has a curve that remarkably deteriorates in a low load state with a power load of 50% as a peak. FIG. 1 shows a power supply unit having a capacity of 400 [W] and a power supply unit having a capacity of 800 [W] as an example. From the general power efficiency curve, when the server device is in a low load state (around 200 [W]), a large-capacity power supply (power supply with a capacity of 800 [W]) has a low capacity power supply (capacity of 400 [W]). It can be seen that the power efficiency is very inferior to the power supply. On the other hand, it can also be seen that power in a high load state (about 700 [W] -800 [W]) cannot be supplied by a low-capacity power supply.

  FIG. 2 shows the power system configuration of the present invention. The information processing apparatus 1 has redundancy in the DC power supply unit, and the DC power supply unit is configured by two DC power supply units, power supply units A 2A to 2C and power supply units B 3A to 3C having different power supply capacities. The power supply unit B is assumed to have a larger power capacity than the power supply unit A. Similar to a general power supply unit for an information processing apparatus, the power supply unit has a main power supply system that operates only when the information processing apparatus is powered on, and a standby power supply system that supplies power even when the power is off.

  The power output from all the power supply units is supplied to the same power supply system of the information processing apparatus, transformed to an appropriate voltage value by the DC low voltage circuits 4A to 4C, and operates the loads 5B to 5D. That is, which power supply unit is the power supply source does not affect the operation of the information processing apparatus. Here, the load is defined as a device represented by a CPU, a memory, and an HDD that are necessary for the operation of the information processing apparatus. When the information processing apparatus is in operation, all loads are turned on.

  As shown in FIG. 3, standby power from the power supply units A and B is supplied to the BMC 6 and the chipset 7 that need to be operated for power-on processing described later, so that the operation can be performed before the power is turned on. It is in a state. Note that the DC low voltage circuits 4D to 4H that transform the voltage supplied by the standby power source to an appropriate voltage for use by the device operate by the standby power source.

  Subsequently, FIG. 4 shows a system of the power control signal 103. In the present invention, power on / off is controlled by the IO Expander 9A. The assertion / deassertion of the on / off signal of the DC power supply unit is controlled by rewriting the register of the IO Expander by an instruction from the BMC 6. When N + M is a small number such as two, the IO Expander is not essential, and a power control signal can be directly connected to the BMC.

  In the present invention, the power supply units 2A to 2C, 3A to 3C and the BMC 6 are connected by the PM bus 104 as shown in FIG. When the BMC reads the output power using PM bus, the power consumption of the information processing apparatus can be calculated. Furthermore, the rated power capacity of each power supply and the number of installed power supply units are acquired.

The rated power capacity of each power supply unit with respect to the power consumption P of the entire information processing apparatus that has been read via PM bus (hereinafter, power consumption P is the sum of the power consumption values of all operating power supply units) Using P _Amax and P _Bmax , a power supply unit that operates according to the flowchart of FIG. 6 is selected. When the power supply unit to be operated is determined, the state is transmitted to the IO Expander, and the IO Expander asserts the ON signal of the selected power supply unit.

  First, the power-on process of the information processing apparatus will be described. The information processing apparatus ON signals 101A and 101B are notified to the chip set 7 when the user presses the power supply SW6 (S601). The power-on notification is notified to the BMC (S602), and when one of the power supply units B is turned on by the BMC, the information processing apparatus starts to be activated (S603). Here, the value of the operating number n of the power supply unit B is set to 1 (S604).

The power consumption P measured by the BMC after activation is compared with a value obtained by multiplying the rated power capacity P _Amax of the power supply unit A by a coefficient k (S605). The coefficient k can be arbitrarily selected according to the rated power capacity difference between the power supply unit A and the power supply unit B. In the above determination, when there is a power unit A that can be operated with power consumption smaller than kP _Amax (S606) and is not currently operating with the power unit A (S607), a power source with a small power capacity It is assumed that the power efficiency is improved by switching the operation to the unit A (S608) and bringing the load factor of the power supply unit close to 50%. Here, the operation number m of the power supply unit A is set to 1, and the operation number n of the power supply unit B is set to 0 (S609).

On the other hand, when the power consumption is larger than kP _Amax , it is determined whether the number n of operation of the power supply unit B is 0 (S610), and the number of operations of the power supply unit B is 0, that is, the power supply unit A If it is in operation, the power supply unit to be operated is switched from the power supply unit A to the power supply unit B (S611), the operation number m of the power supply unit A is 0, and the operation number n of the power supply unit B is 1 (S612). ). When the number of operating power supply units B is not 0, that is, when the power supply unit B is operating, a value obtained by multiplying P _Bmax by the current number n of operating power supply units and an arbitrary coefficient l and power consumption P Are compared (S613). The arbitrary coefficient l here is a value of 1 or less multiplied to prevent the possibility of exceeding the rated power value, and an arbitrary value can be selected.

  When the power consumption P is large, it is determined whether or not the operation number can be increased by comparing whether the operation number (n + 1) when the operation number of the power supply unit B is increased is smaller than the maximum operable number Nmax (S614). ). If possible, one additional power supply unit B is added (S615), and the number n of operation of the power supply unit B is increased by 1 (S616). Increase efficiency. If the added number of requested operation power supplies exceeds the number of operable power supplies, exception processing is performed (S617).

If the power consumption P is small, the current number of operating power supply units is maintained and the next determination is made (S618). When the power consumption P is smaller than the value obtained by multiplying P _Bmax by a value (n-1) decremented from the current number of operating power supply units and an arbitrary coefficient, one operating power supply unit B is stopped (S619). ), 1 is subtracted from the operating number n of the power supply unit B (S620). As a result, the load factor of one power supply unit approaches 50%, and the power efficiency is improved.

  A supplementary explanation will be given for the coefficients k and l with reference to FIG. As described above, according to the present invention, the type of operating power supply unit and the number of operating power supply units are selected so as to obtain a power load factor with good power supply efficiency. Coefficients are provided to allow the user to select an arbitrary value for the timing for performing these switching operations. In FIG. 7, a power load factor of 60% to 90% (k, l = 0.6 to 0.9) is recommended. The closer to 100%, the higher the risk of exceeding the power rating, and the closer to 50%, the worse the power efficiency when the number of operating power supply units is increased.

  Here, when switching is performed, there is a delay in switching time according to the time chart of FIG. 8, thereby preventing the risk of voltage drop due to switching of the power supply unit and ensuring stable operation of the information processing apparatus. .

In the following, abnormal processing will be described.
There are three main types of abnormal systems. 1. 1. Failure of power supply unit B Operation request number of power (n + 1) the operable speed (N _max) is less than, 3. The number of power supply units B that can be operated is 0, and only a power supply unit with a small power supply capacity can be operated. An implementation of processing for the above abnormal system will be described with reference to FIG.
Example of abnormal system 2. Since 3 and 3 can be included in the case of Example 1, description will be made using Example 1.

First, due to failure of the power supply unit B, the failure is notified through PM Bus (S901), so that the BMC detects an abnormality and sets a power supply abnormality flag (S902). Since the number of power supply units B that can operate due to a failure decreases, the number of failure detections is decremented from N _max (S903). Moreover, BMC stores fault the number of working power supply unit B when occurring as N _abn (S904).

Subsequently, it is determined whether there is a power supply unit B that can operate as an alternative (S905). If the number _Nmax of operable power supply units is larger than the current number n of operational power supply units, the standby power supply unit is immediately operated as a substitute (S906), and the normal flow is returned (S907). However, the user can know that there is a failed power supply without releasing the power supply abnormality flag.

  On the other hand, a process when assignment of a substitute is impossible will be described. The subsequent processing is shown in Example 2.3. This is the same as the process. The BMC asserts the load limit signal 105 shown in FIG. 5 because the number of operating power sources cannot be secured, that is, the power capacity sufficient to supply the power consumption P cannot be secured. (S908). The load control signal 105 is connected to all loads capable of controlling power consumption through the IO Expander 9B. When the number of loads to be controlled is small, it is possible to connect a signal directly to the BMC without using the IO Expander 9B.

Judgment whether the number M _max of operable power supply units A is 1 or more with respect to the power consumption P after asserting the load control signal, and the power consumption P is smaller than a value obtained by multiplying the coefficient k by the rating of the power supply unit A That is, it is determined whether to switch to the power supply unit A (S909). When possible, the operation power supply unit is changed to the power supply unit A (S910), the operation number m of the power supply unit A is set to 1, and the operation number n of the power supply unit B is set to 0 (S911). In other cases, except for the failed power supply unit, the system continues to operate with one _Nabn- 1 (S912). At this time, the operating number n of the power supply units B is changed to _Nabn- 1 (S913). The operation of the information processing apparatus is continued in a state where the load is controlled, and the state of the abnormality detection signal is monitored (S914). When all the abnormality detection signals are released, the BMC releases the power supply abnormality flag (S915), and returns the operable number _Nmax of the power supply unit B to the initial value (S916). After setting the number of operating power supplies n to the value _Nabn before the occurrence of the failure (S917, S918), the power limit signal is canceled (S919). After the above processing is completed, the operation returns to the normal flow and continues to operate (S920).

FIG. 10 shows a plurality of main examples in the operation state table of each power supply unit.
At the time of startup, one power supply unit B is in an operating state. In general, when the initialization process is performed, all loads do not become high loads, so that sufficient power can be supplied.
Next, an operation state when switching to the power supply unit A is shown. In the user use state, n power supply units B are mainly in operation. As shown here, the power supply is basically turned on from the power supply unit with the smallest number. Next, the abnormal time will be described. When the power control of the information processing apparatus can be remarkably limited by turning on the load control, the operation is continued only with the power supply unit A. When the power supply unit A does not have enough power, a stable operation is provided by operating all of the power supplies that were operating when the failure occurred, except for the failed power supply unit. The table shows an example when a failure occurs in the power supply unit B [2]. Finally, a case where a failure occurs in the power supply unit A will be described. The M power supply units A are standby systems except for the power supply unit with the lowest number. When an abnormality occurs in the power supply unit A with the lowest number, the power supply unit A with the next number operates.

  FIG. 11 shows an example of a register map that stores the operating state of the power supply unit. Among the storage areas of the BMC, the user area is used to store the number of operations, the maximum number of operations possible (the number of implementations), and the number of operations when a failure occurs, using 8 bits each. Note that 8 bits is an example, and the minimum number of bits necessary for system requirements can be selected.

1 Information processing device 2, 2A to 2C Power supply unit A
3, 3A-3C Power supply unit B
4, 4A-4H DC low voltage circuit 5, 5A-5D Load 6 BMC
7 Chipset 8 Power switch 9, 9A, 9B I / O Expander
101 Power ON notification signal 102 I2C Bus
103 Power control signal 104 PM Bus
105 Load limit signal

Claims (3)

  DC power supply units having different power supply capacities are mounted redundantly and selectably, and an information processing apparatus that supplies power to the load from the selected DC power supply unit, monitors the output power of each DC power supply unit, and A management processor for controlling the operating state of the DC power supply unit; the management processor calculates the power consumption of the information processing apparatus based on the output power acquired from each DC power supply unit; and according to the calculated power consumption An information processing apparatus for selecting which power supply capacity of the DC power supply unit to operate.   2. The information processing apparatus according to claim 1, comprising two types of power supply units having a large and small power supply capacity, wherein the management processor selects a power supply unit having a small power supply capacity when the power consumption of the information processing apparatus is smaller than a predetermined value.   The management processor has a function capable of limiting the power consumption of the load. When there is no power supply unit that can be replaced in the event of a failure of the power supply unit, the management processor limits and limits the power consumption of the load. The information processing apparatus according to claim 1, wherein the power supply unit of which power capacity is to be operated is selected according to power consumption of the information processing apparatus later.

Images