JP5712812B2 - Computer system and method of operating computer system - Google Patents

Description

  The present invention relates to a computer system having a plurality of processors.

  In general, a computer system such as a server apparatus is formed by housing a package board on which a processor and a memory are mounted in an apparatus rack. In this type of computer system, a fan is used to generate an air flow in the apparatus rack, and heat generated from the processor or the like is discharged to the outside. For example, the amount of air flowing through the device rack is adjusted by moving an adjustment valve in accordance with the temperature detected by the temperature sensor (see, for example, Patent Documents 1 and 2). When heat sinks of a plurality of processors are arranged side by side in the direction of the airflow, in order to prevent a decrease in cooling efficiency due to the heat sink, the interval between the fins of a part of the heat sink is widened (see, for example, Patent Document 3). A method has been proposed in which a load applied to a processor is predicted for each cycle, and the operating frequency of the processor, the number of rotations of a fan, and the like are controlled based on the prediction (for example, see Patent Document 4).

JP 2010-86450 A Japanese Patent Laid-Open No. 2-25497 JP 2010-152886 A JP 2008-198072 A

  For example, in a computer system equipped with a plurality of processors, it is necessary to arrange a temperature sensor for each processor in order to reliably cool the processors. However, in this case, the number of parts required for the cooling mechanism increases and the control becomes complicated, thereby increasing the cost of the computer system.

  An object of the present invention is to simplify the design of a cooling mechanism of a computer system by always making the heat generation amount of one processor larger than the heat generation amount of another processor.

  In one embodiment of the present invention, a computer system sequentially receives a plurality of processors that respectively process a plurality of sequentially generated tasks and task information indicating the contents of the tasks, and obtains a load applied to the processor to process each task. For each task group including a calculation unit and a predetermined number of consecutive tasks, a task with a relatively high load is assigned to the first processor, which is one of the processors, and a task with a relatively low load is assigned to the remaining processors. And an assigning unit.

  The heat generation amount of the first processor can always be larger than the heat generation amount of the remaining processors, and the design of the cooling mechanism of the computer system can be simplified.

1 illustrates an example of a computer system in one embodiment. 3 illustrates an example of a computer system in another embodiment. 3 shows an example of the operation of the calculation unit shown in FIG. 3 shows an example of the operation of the allocation unit shown in FIG. 3 shows an example of the operation of the computer system shown in FIG. 3 illustrates an example of a computer system in another embodiment. 7 shows an example of power consumption of each chip in the computer system shown in FIG. 7 shows an example of fan control shown in FIG. 6. 7 shows an example of setting the fan air volume of the computer system shown in FIG. 6. 3 illustrates an example of a computer system in another embodiment. 3 illustrates an example of a computer system in another embodiment. The example of the thermoactuator shown in FIG. 11 is shown. 3 illustrates an example of a computer system in another embodiment. 14 illustrates an example of fan control of the computer system illustrated in FIG. 13. 14 illustrates an example of operations of a processor, a shutter, and a fan in the computer system illustrated in FIG.

  Hereinafter, embodiments will be described with reference to the drawings.

  FIG. 1 shows an example of a computer system SYS in one embodiment. For example, the computer system SYS is a server system. The computer system SYS has a plurality of processor CPUs (CPU1, CPU2), a calculation unit CALC, and an allocation unit ASGN. For example, the processors CPU1 and CPU2 have the same functions and process tasks by executing programs. The program is stored in a built-in memory in each processor CPU1-2, or is stored in an external memory accessed by each processor CPU1-2.

  The calculation unit CALC sequentially receives task information TINF indicating the contents of the task TSK (TSKa1, TSKb1, TSKa2, TSKb2,...), And is applied to the processor CPU1-2 that processes each task TSK based on the task information TINF. Find the load. The calculation unit CALC outputs the obtained load to the allocation unit ASGN as load information LD. Note that the load information LD is common information that does not depend on the processor CPU1-2, and the calculation unit CALC does not recognize the processor CPU on which each task TSK is performed.

  For example, the task TSK supplied to the computer system SYS includes data to be processed by a program executed by the processor CPU1-2 and processing conditions (calculation type, data amount, etc.). The type of calculation is addition processing, multiplication processing, division processing, or the like. The task TSK may include a program executed by the processor CPU1-2. The task information TINF may be included in each task TSK, and may be supplied to the calculation unit CALC separately from each task TSK.

  The task TSK is sequentially supplied to the computer system SYS from another computer on the network connected to the host system or the computer system SYS. In this embodiment, one and the other of two tasks TSK (TSKa1, TSKb1, or TKa2, TSKb2, etc.) supplied in succession are processed by one and the other of the processors CPU1-2. In the following description, two continuous tasks TSK executed by the processor CPU1-2 are referred to as task groups TSKG (TSKG1, TSKG2).

  In this example, the number of tasks TSK included in each task group TSKG is the same as the number of processors CPU1-2. Note that the number of task TSKs included in each task group TSKG may be a multiple of the number of processor CPUs 1-2 (positive number; 2, 4, 6,...). For example, when each task group TSKG includes four tasks TSK, the two tasks TSK in the task group TSKG are executed by the processor CPU1, and the remaining two tasks TSK are executed by the processor CPU2.

  When the computer system SYS has three processor CPUs, each task group TSKG includes a task TSK that is a multiple of 3 (positive number). When the computer system SYS has four processor CPUs, each task group TSKG includes a task TSK that is a multiple of 4 (positive number). When the number of tasks TSK included in each task group TSKG is larger than the number of processor CPUs, the same number of tasks TSK are equally processed by the processor CPUs.

  Based on the load information LD from the calculation unit CALC, the allocation unit ASGN allocates a task TSK having a relatively large load to the processor CPU1 for each task group TSKG, and assigns the remaining task TSK having a relatively small load to the processor CPU2. Assign to. For example, when the load information LD indicates that the load of the task TKa1 is larger than the load of the task TSKb1, the assignment unit ASGN assigns the task TKa1 to the processor CPU1 and assigns the task TSKb1 to the processor CPU2. For example, the load corresponds to the number of clock cycles required when the processor CPU1 or CPU2 processes the task TSK. That is, the larger the number of clock cycles required for processing the task TSK, the greater the load.

  For example, task TSK allocation is performed by transferring data to be processed by task TSK and information indicating processing conditions (calculation type, data amount, etc.) to a memory area allocated to the corresponding processor CPU. The The memory area is allocated to the built-in memory or the external memory. Alternatively, the task TSK is assigned by transferring a program executed for processing the task TSK to a built-in memory or an external memory of the corresponding processor CPU.

  For example, the calculation unit CALC and the allocation unit ASGN are realized by a processor mounted in the system SYS executing a program. Note that the calculation unit CALC and the allocation unit ASGN may be realized by hardware.

  The processor CPU1-2 executes a program to process the task TSK (for example, TKa1 and TSKb1) allocated by the allocation unit ASGN. Tasks TKa1 and TSKb1 are executed overlapping each other. The allocation unit ASGN causes the processor CPU1 to always process a task TSK with a large load, so that the heat generation amount of the processor CPU1 is always larger than the heat generation amount of the processor CPU2. For this reason, the cooling mechanism of the computer system SYS may be designed in consideration of the processor CPU1, and the design of the cooling mechanism can be simplified. For example, in order to monitor the temperature of the processor CPU1, it is only necessary to arrange an intake air temperature sensor and an exhaust gas temperature sensor, and monitoring of the temperature of the processor CPU2 can be omitted. Moreover, since it is not necessary to provide an equivalent cooling mechanism in both processor CPU1-2, the cost of a cooling mechanism can be reduced.

  As described above, in this embodiment, the heat generation amount of the processor CPU1 can always be larger than the heat generation amount of the processor CPU2, and the design of the cooling mechanism of the computer system SYS can be simplified.

  FIG. 2 shows an example of a computer system SYS in another embodiment. The same elements as those described in the above-described embodiment are denoted by the same reference numerals, and detailed description thereof will be omitted. For example, the computer system SYS is a server system. The computer system SYS has a plurality of processors CPU (CPU1, CPU2), a memory MEM (MEM1, MEM2), and a control unit CNT.

  For example, the processor CPU1-2 is an independent semiconductor chip and has the same function as each other as in FIG. The processor CPU1 operates in synchronization with the clock CLK1, and processes a task by executing a program stored in the memory MEM1. The processor CPU2 operates in synchronization with the clock CLK2, and processes a task by executing a program stored in the memory MEM2.

  The memory MEM1 is a single semiconductor chip, and stores a program executed by the processor CPU1 and a task TSK (data, processing conditions, etc.) processed by the processor CPU1. The memory MEM2 is a single chip and stores a program executed by the processor CPU2 and a task TSK (data, processing conditions, etc.) processed by the processor CPU2. Note that the memories MEM1 and MEM2 may be built-in memories built in the processors CPU1 and CPU2, respectively. Alternatively, the memories MEM1 and MEM2 may be formed by one memory chip. Alternatively, the memories MEM1 and MEM2 are not limited to a single semiconductor chip, and may be a plurality of semiconductor chip groups.

  The control unit CNT includes a calculation unit CALC, an allocation unit ASGN, a clock control unit CCNT, a memory control unit MCNT, and a task control unit TCNT. For example, the control unit CNT is one semiconductor chip including a processor. Calculation unit CALC and allocation unit ASGN are realized by a processor included in control unit CNT executing a program. Note that the calculation unit CALC and the allocation unit ASGN may be realized by hardware.

  As in FIG. 1, the calculation unit CALC obtains the load applied to the processor CPU1-2 that processes each task TSK based on the task information TINF, and outputs the obtained load to the assignment unit ASGN as the load information LD. As in FIG. 1, each task TSK includes data to be processed by a program executed by the processor CPU1-2 and processing conditions (such as the type of operation and the amount of data). Each task TSK may include a program. The task TSK is sequentially supplied from the upper system or the like to the computer system SYS. The number of tasks TSK included in each task group TSKG is a multiple (positive number) of the number of processors CPU1-2, and is “2” in this example. The number of processor CPUs provided in the computer system SYS may be “3” or more. When the number of tasks TSK included in each task group TSKG is larger than the number of processor CPUs, the same number of tasks TSK are processed equally by the processor CPUs.

  Based on the load information LD from the calculation unit CALC, the allocation unit ASGN allocates a task TSK with a relatively large load to the processor CPU1 and allocates a task TSK with a relatively small load to the processor CPU2 for each task group TSKG. . In this embodiment, the task TSK is allocated by setting the operating frequency of the clock CLK1-2 supplied to the CPU 1-2, the allocation of the memory area of the memory MEM1-2 used by the CPU 1-2, and the allocated memory area. Transfer of task TSK (data, processing conditions or program).

  For example, when the load information LD indicates that the load of the task TSKa1 is larger than the load of the task TSKb1, the assignment unit ASGN assigns the task TSKa1 to the processor CPU1 and performs task control information for assigning the task TSKb1 to the processor CPU2. To the part TCNT. In addition, the allocation unit ASGN outputs information for setting the operating frequency of the clock CLK1-2 to the clock control unit CCNT, and outputs information for allocation of the memory area of the memory MEM1-2 to the memory control unit MCNT.

  The task control unit TCNT assigns data, processing conditions, or programs corresponding to the task TSK in order to allocate a pair of tasks TSK (for example, TKa1, TSKb1) to the processor CPU1-2 based on information from the allocation unit ASGN. The data is transferred to the memories MEM1-2. For example, the task control unit TCNT is a processor or a DMAC (Direct Memory Access Controller) that manages the task TSK.

  The clock control unit CCNT is, for example, a pair of PLL (Phase-locked loop) circuits, and sets the frequency of the clock CLK1-2 based on information from the allocation unit ASGN. At this time, the frequency of the clock CLK1 is set to be equal to or higher than the frequency of the clock CLK2 based on information from the allocation unit ASGN.

  The memory control unit MCNT allocates a memory area of the memory MEM1-2 used by the processor CPU1-2 based on information from the allocation unit ASGN. For example, the memory control unit MCNT has an MMU (Memory management Unit) function. When the memory MEM1-2 is a DRAM (Dynamic Random Access Memory), the memory control unit MCNT executes a refresh operation of the memory area used by the processor CPU1-2 and prohibits a refresh operation of other memory areas. When the memory area in which the refresh operation is executed is small, the refresh cycle, which is the execution interval of the refresh operation, can be lengthened, so that the power consumption of the memories MEM1-2 can be reduced.

  The allocation unit ASGN may include at least one of the functions of the clock control unit CCNT, the memory control unit MCNT, and the task control unit TCNT. Further, when the size of the memory area used by the processor CPU1-2 is fixed, the memory control unit MCNT may not be formed.

  The processor CPU1-2 executes the programs overlapping each other in order to process the task TSK (for example, TKa1 and TSKb1) allocated by the allocation unit ASGN. As in FIG. 1, the heat generation amount of the processor CPU1 is always larger than the heat generation amount of the processor CPU2 due to the operations of the calculation unit CALC and the allocation unit ASGN. For example, heat sinks are respectively attached to the surface of the processor CPU1-2 in order to release heat. The heat sink for the processor CPU1 that generates a large amount of heat is designed to have a higher cooling efficiency than the heat sink for the processor CPU2. Alternatively, a heat sink is attached only to the surface of the processor CPU1. Thereby, the design of the cooling mechanism of the computer system SYS can be simplified. Moreover, since it is not necessary to provide an equivalent cooling mechanism in both processor CPU1-2, the cost of a cooling mechanism can be reduced. For example, in order to monitor the temperature of the processor CPU1, it is only necessary to arrange an intake air temperature sensor and an exhaust gas temperature sensor, and monitoring of the temperature of the processor CPU2 can be omitted.

  FIG. 3 shows an example of the operation of the calculation unit CALC shown in FIG. The calculation unit CALC receives, as task information TINF, an information amount such as data to be processed by the task TSK and a unit load that is a load per unit information of the information amount. Here, in order to simplify the explanation, the information amount and the unit load are shown in relative sizes. The calculation unit CALC obtains load information LD (predicted load) indicating the load applied to the processor CPU1-2 by multiplying the information amount by a unit load. The calculation unit CALC outputs the obtained load information LD to the allocation unit ASGN for each task group TSKG.

  FIG. 4 shows an example of the operation of the allocation unit ASGN shown in FIG. For each task group TSKG, the allocation unit ASGN allocates a task TSK with a large load indicated by the load information LD to the processor CPU1, and allocates a task TSK with a small load to the processor CPU2. For example, in task group TSKG1, task TSKa1 with load information LD 100 is assigned to processor CPU1, and task TSKb1 with load information LD 60 is assigned to processor CPU2. In FIG. 4, the execution processor indicates a processor CPU that executes a task TSK.

  Allocation unit ASGN obtains frequencies Fa (Fa1, Fa2, Fa3,...) And Fb (Fb1, Fb2, Fb3,...) Of clock CLK1-2 according to the load indicated by load information LD. . The allocation unit ASGN associates the obtained frequencies Fa and Fb with the type of the CPU 1-2 and outputs the frequency to the clock control unit CCNT for each task group TSKG. For example, the frequencies Fa and Fb are set to values at which each processor CPU1-2 can process the task TSK within a predetermined processing time as the operating frequency of each processor CPU1-2. The frequency of each clock CLK1-2 is set to the lowest value among the settable frequencies in order to minimize the power consumption of the computer system SYS.

  Generally, the frequency is set high when the load is large, and is set low when the load is small. For this reason, the frequency Fa or Fb of the clock CLK1 for the processor CPU1 that performs processing with a relatively large load is higher than the frequency Fb or Fa of the clock CLK2 for the processor CPU2 that performs processing with a relatively small load. .

  Further, the allocation unit ASGN allocates memory areas MAa (MAa1, MAa2, MAa3,...), MAb (MAb1, MAb2, MAb3,...), Which are allocated to the processor CPU1-2 according to the load indicated by the load information LD. .) The allocation unit ASGN associates the obtained memory areas MAa and MAb with the type of the CPU 1-2, and outputs it to the memory control unit MCNT for each task group TSKG. Furthermore, the allocation unit ASGN outputs the data, processing conditions, or program processed by the tasks TSKa and TSKb to the task control unit TCNT for each task group TSKG in association with the type of the CPU 1-2.

  FIG. 5 shows an example of the operation of the computer system SYS shown in FIG. Each processor CPU1-2 receives the clock CLK1-2 having the frequency Fa or Fb obtained by the assignment unit ASGN, and performs the processing of the task TSK using the memory area MAa or MAb. The frequencies Fa and Fb of the clock CLK1-2 and the memory areas MAa and MAb are obtained by the assigning unit ASGN before the processor CPU1-2 starts processing of each task group TSKG, and the clock control unit CCNT and the memory control unit MCNT. Is set by As described above, also in this embodiment, the same effect as that of the above-described embodiment can be obtained.

  FIG. 6 shows an example of a computer system SYS in another embodiment. The same elements as those described in the above-described embodiment are denoted by the same reference numerals, and detailed description thereof will be omitted. For example, the computer system SYS is a server system.

  The computer system SYS has a package substrate PKGB on which the processor CPU1-2, the memory MEM1-2, and the control unit CNT illustrated in FIG. 2 are mounted. A heat sink HS is attached to the surface of each processor CPU1-2. The package substrate PKGB has temperature sensors TSI and TSE respectively attached to the intake side and the exhaust side of the airflow flowing through the heat sink HS attached to the processor CPU1.

  The number of processor CPUs provided in the computer system SYS may be “3” or more. When the number of tasks TSK included in each task group TSKG is larger than the number of processor CPUs, the same number of tasks TSK are processed equally by the processor CPUs.

  The package substrate PKGB is housed in the casing CS together with the power supply unit PSU, the hard disk drive HDD, and the fan FAN. In order to introduce the airflow AFI generated by the fan FAN into the casing CS and discharge the airflow AFE from the casing CS on the intake side (lower side in FIG. 6) and the exhaust side (upper side in FIG. 6) of the casing CS. Inlet holes and exhaust holes are respectively formed. The power supply unit PSU supplies power to the package substrate PKGB and the hard disk drive HDD. The hard disk drive HDD stores programs, data, and the like that are transferred to the memory MEM1-2 and executed by the processor CPU1-2.

  When the difference between the temperature detected by the temperature sensor TSE and the temperature detected by the temperature sensor TSI is relatively large, the fan FAN has an increased number of rotations (air volume) and a relatively small temperature difference. The number of rotations (air volume) is reduced. Further, the rotational speed (air volume) of the fan FAN increases as the heat generation amount of the processor CPU1 increases. The amount of heat generated by the processor CPU1 is proportional to the power consumption of the processor CPU1, and increases as the load on the processor CPU1 increases.

  For example, the rotation speed of the fan FAN based on the temperature detection by the temperature sensors TSI and TSE and the amount of heat generated by the processor CPU1 is controlled by a control circuit mounted on the package substrate PKGB or the control unit CNT. Note that only the temperature sensor TSE may be mounted on the package substrate PKGB, and the rotation speed of the fan FAN may be controlled based on the temperature detected on the exhaust side of the processor CPU1. The fan FAN and the control circuit or control unit CNT change the amount of airflow flowing on the package substrate PKGB according to the temperature detected by the temperature sensors TSI and TSE, and the cooling unit increases the amount of airflow as the temperature increases. It is an example.

  The computer system SYS of this embodiment operates in the same manner as in FIGS. The rotation speed of the fan FAN increases when the amount of heat generated (power consumption) of the processor CPU1 is large, and decreases when the amount of heat generated (power consumption) of the processor CPU1 is small. In this embodiment, the processor CPU1 performs processing of a task TSK with a large load. For this reason, the heat generation amount of the processor CPU1 is larger than the heat generation amounts of the processor CPU2 and other semiconductor chips.

  The rotational speed of the fan FAN is controlled based on the surface temperature of the processor CPU1 that generates the largest amount of heat on the package substrate PKGB. For this reason, other semiconductor chips such as the processor CPU2 having a relatively small heat generation amount are sufficiently cooled by the airflows AFI and AFE generated by the fan FAN. Therefore, the semiconductor chip mounted on the package substrate PKGB can be cooled with the minimum necessary air volume. As a result, the power consumption of the cooling system can be minimized without degrading the reliability of the computer system SYS.

  FIG. 7 shows an example of the power consumption Q of each chip in the computer system SYS shown in FIG. In this embodiment, the power consumption of the processor CPU1 is always larger than that of other chips due to load assignment by the control unit CNT. Further, the power consumption of the processor CPU1 is controlled by the fan FAN so as not to exceed the limit value Qlimit. The limit value Qlimit is a value of the power consumption Q corresponding to the maximum junction temperature allowed in the processor CPU1 (product specification value). The junction temperature is the chip temperature of the processor CPU1, and can be determined from the ambient temperature of the processor CPU1 and the power consumption if the thermal resistance of the package member molding the chip is known.

  FIG. 8 shows an example of fan control of the computer system SYS shown in FIG. The control shown in FIG. 9 is performed by a control circuit or control unit CNT mounted on the package substrate PKGB.

  First, in step S10, the power consumption Q of the processor CPU1 is estimated based on the load of the processor CPU1 obtained by the calculation unit CALC. Alternatively, the power consumption Q of the processor CPU1 is estimated based on power information that can be actually measured from a DC-DC converter on the printed circuit board. In step S12, a temperature difference ΔT between the temperature sensors TSI and TSE is detected. In step S14, the air volume V of the fan FAN is set based on the power consumption Q and the temperature difference ΔT. In step S16, after waiting for a predetermined time, steps S10 to S14 are repeatedly performed.

  FIG. 9 shows a setting example of the air volume of the fan FAN shown in FIG. The horizontal axis indicates the temperature difference ΔT detected by the temperature sensors TSE and TSI, and the vertical axis indicates the power consumption (heat generation amount) Q of the processor CPU1. Numerical values from 10 to 100 in the figure indicate the air volume V of the fan FAN.

  For example, the air volume V is substantially proportional to the rotational speed of the fan FAN. The relationship between the air volume V and the rotational speed is obtained in advance by actual measurement. The air volume V is set larger as the power consumption Q is larger, and is set larger as the temperature difference ΔT is larger. As shown in FIG. 7, since the power consumption Q is constantly changing, it is desirable to use the root mean square value for a predetermined period. In FIG. 7, the vertical axis is “power consumption”, but “vertical heat generation” may be plotted on the vertical axis.

  As described above, also in this embodiment, the same effect as that of the above-described embodiment can be obtained. Further, by controlling the rotation speed of the fan FAN in accordance with the heat generation amount (temperature) of the processor CPU1 having the largest heat generation amount, the cooling system can be achieved without degrading the reliability of the semiconductor chip mounted on the package substrate PKGB. Power consumption can be minimized.

  FIG. 10 shows an example of a computer system SYS in another embodiment. The same elements as those described in the above-described embodiment are denoted by the same reference numerals, and detailed description thereof will be omitted. For example, the computer system SYS is a server system.

  The computer system SYS includes a partition PT, a shutter SHT, and a stepping motor MTR between the package substrate PKGB and the fan FAN. Other configurations of the computer system SYS are the same as those in FIG. 6 except that the control unit CNT controls the stepping motor MTR instead of the rotation speed of the fan FAN.

  The number of processor CPUs provided in the computer system SYS may be “3” or more. When the number of tasks TSK included in each task group TSKG is larger than the number of processor CPUs, the same number of tasks TSK are processed equally by the processor CPUs.

  The partition wall PT has a through hole. The shutter SHT is movably disposed along the partition wall PT and has a through hole. The shutter SHT moves to the right or left side of FIG. 10 when the stepping motor MTR is driven. The hatched portions in the partition wall PT and the shutter SHT indicate plate-like members, and the white portions indicate through holes.

  In this embodiment, the rotation speed of the fan FAN is always constant. For this reason, the amount of airflow that is the amount of airflow flowing over the package substrate PKGB increases as the opening that is the overlapping portion of the through hole of the partition wall PT and the through hole of the shutter SHT is larger. The degree of overlap between the partition PT and the through hole of the shutter SHT is expressed by an aperture ratio (for example, 10% to 100%). As described above, the shutter SHT is arranged to be freely opened and closed on the downstream side of the airflow flowing on the package substrate PKGB. In FIG. 10, the partition wall PT and the shutter SHT are disposed on the downstream side of the airflow flowing on the package substrate PKGB, but may be disposed on the upstream side of the airflow.

  In addition to the functions shown in FIG. 2, the control unit CNT determines the aperture ratio of the shutter SHT based on the temperature difference detected by the temperature sensors TSI and TSE and the amount of heat generated by the processor CPU1, and controls the stepping motor MTR. It has a function to do. The determination of the aperture ratio of the shutter SHT and the control of the stepping motor MTR may be performed by another control circuit mounted on the package substrate PKGB. Alternatively, only the temperature sensor TSE may be mounted on the package substrate PKGB, and the aperture ratio of the shutter SHT may be determined based on the temperature detected on the exhaust side of the processor CPU1. The shutter SHT, the stepping motor MTR, the control unit CNT or other control circuit changes the amount of airflow flowing on the package substrate PKGB according to the temperature detected by the temperature sensors TSI and TSE, and the higher the temperature, the more the airflow It is an example of the cooling part which increases the quantity of the.

  The computer system SYS of this embodiment operates in the same manner as in FIGS. The processor CPU1 performs processing of the task TSK with a heavy load. For this reason, the heat generation amount of the processor CPU1 is larger than the heat generation amounts of the processor CPU2 and other semiconductor chips. The aperture ratio of the shutter SHT increases when the heat generation amount (power consumption) of the processor CPU1 is large, and decreases when the heat generation amount (power consumption) of the processor CPU1 is small.

  For example, the opening / closing control of the shutter SHT by the control unit CNT or the control circuit is exemplified by replacing step S14 in FIG. 8 with the setting of the aperture ratio of the shutter SHT. The aperture ratio is exemplified by replacing the value of the air volume V shown in FIG. 9 with the aperture ratio. Note that since the air volume flowing over the package substrate PKGB is proportional to the aperture ratio of the shutter SHT, the aperture ratio of the shutter SHT may be obtained based on the air volume V of FIG.

  Note that the amount of heat generated by the processor CPU1 is maintained substantially constant during processing of one task TSK. This is because the operating frequency of the processor CPU1 is set for each task TSK. The aperture ratio of the shutter SHT changes after the start of the task TSK until the heat generation amount of the processor CPU1 is stabilized, and does not change thereafter. Therefore, the operation frequency of the shutter SHT becomes higher as the processing time of the task TSK is shorter. In order to ensure the reliability of the shutter SHT and obtain a stable air volume, for example, the processing time of each task group TSKG1-3 shown in FIG. 5 is preferably 10 seconds or more.

  As described above, also in this embodiment, the same effect as that of the above-described embodiment can be obtained. Further, by controlling the aperture ratio of the shutter SHT in accordance with the heat generation amount (temperature) of the processor CPU1 having the largest heat generation amount, the cooling system can be achieved without reducing the reliability of the semiconductor chip mounted on the package substrate PKGB. Power consumption can be minimized.

  FIG. 11 shows an example of a computer system SYS in another embodiment. The same elements as those described in the above-described embodiment are denoted by the same reference numerals, and detailed description thereof will be omitted. For example, the computer system SYS is a server system.

  The computer system SYS has a thermoactuator SA, a wire WR, a pulley PL, and a spring SPR instead of the temperature sensors TSI and TSE and the stepping motor MTR shown in FIG. The positional relationship between the partition wall PT and the shutter SHT is the same as in FIG. 10, and the amount of airflow flowing on the package substrate PKGB is adjusted according to the aperture ratio that is the degree of overlap between the partition hole PT and the through hole of the shutter SHT. . Other configurations of the computer system SYS are the same as those in FIG. 10 except that the control unit CNT does not control the rotation speed of the fan FAN and the stepping motor MTR. For example, the rotation speed of the fan FAN may always be constant.

  The number of processor CPUs provided in the computer system SYS may be “3” or more. When the number of tasks TSK included in each task group TSKG is larger than the number of processor CPUs, the same number of tasks TSK are processed equally by the processor CPUs.

  The spring SPR is attached between one end of the shutter SHT and the housing CS in an extended state, and functions to pull the shutter SHT toward the spring SPR. The thermoactuator SA pushes out the wire WR when the amount of heat generated by the processor CPU1 is large. At this time, due to the contraction force of the spring SPR, the shutter SHT moves to the right side of the figure, and the aperture ratio increases. That is, the amount of air flowing over the package substrate PKGB is increased. On the other hand, the thermoactuator SA pulls the wire WR when the amount of heat generated by the processor CPU1 is small. At this time, the shutter SHT moves to the left side of the drawing while extending the spring SPR, and the aperture ratio becomes small. That is, the amount of air flowing over the package substrate PKGB is reduced.

  FIG. 12 shows an example of the thermoactuator SA shown in FIG. The thermoactuator SA has a heat transfer section 10 and a main body section 12. The heat transfer unit 10 is formed of aluminum, copper, or the like, and has a plate-like portion 10a that is sandwiched between the processor CPU1 and the heat sink HS. That is, the thermoactuator SA is in contact with the surface of the processor CPU1.

  The main body 12 includes a heat detection unit 12a, a storage unit 12b in which wax W (resin) is stored, and a piston 12d connected to a flange 12c that contacts the wax W. A spring 12e that presses the flange 12c toward the wax W is disposed in the storage portion 12b. The heat detection unit 12a is inserted into the opening 10b of the heat transfer unit 10, is thermally connected to the heat transfer unit 10, and heat transmitted from the processor CPU1 to the heat transfer unit 10 is transmitted. The heat detection part 12a and the opening part 10b are connected via thermal grease, or are connected by solder.

  The wax stored in the storage unit 12b increases in volume as the temperature rises as the temperature of the heat detection unit 12a increases, and decreases in volume as the temperature decreases in the heat detection unit 12a. The piston 12c moves following the change in the volume of the wax. The amount of movement of the piston 12c with respect to the temperature change is determined by the type of wax and the mixing ratio with the mixture. A wire WR is attached to the tip of the piston 12c.

  When the amount of heat generated by the processor CPU1 increases and the temperature of the heat transfer unit 10 rises and the volume of wax increases, the piston 12c moves in a direction protruding from the main body unit 12. As a result, the wire WR is pushed out, and the shutter SHT moves to the right in FIG. 11, and the aperture ratio of the shutter SHT increases. On the other hand, when the amount of heat generated by the processor CPU1 decreases and the temperature of the heat transfer unit 10 decreases and the volume of wax decreases, the piston 12c moves in the direction in which it is drawn into the main body unit 12. As a result, the wire WR is drawn to the main body 12 side, and the shutter SHT moves to the left side of FIG. 11, and the aperture ratio of the shutter SHT is reduced. In this embodiment, no electric power is required to open and close the shutter SHT. The piston 12c is an example of a movable part that moves according to the temperature of the processor CPU1.

  As described above, also in this embodiment, the same effect as that of the above-described embodiment can be obtained. Further, the heat generation amount (temperature) of the processor CPU 1 having the largest heat generation amount is detected by the thermoactuator SA and the aperture ratio of the shutter SHT is adjusted, thereby reducing the reliability of the semiconductor chip mounted on the package substrate PKGB. In addition, the power consumption of the cooling system can be minimized. In particular, since the shutter SHT can be opened and closed according to the amount of heat generated by the processor CPU1 without using power, the power consumption of the cooling system can be reduced.

  In this embodiment, the shutter SHT is moved by the thermoactuator SA. However, for example, the thermoactuator SA is arranged instead of the temperature sensors TSI and TSE of FIG. 6, and the fan FAN having an air volume adjusting lever is arranged. May be. Then, the air volume adjustment lever of the fan FAN may be moved by the piston PS or the wire WR.

  FIG. 13 shows an example of a computer system SYS in another embodiment. The same elements as those described in the above-described embodiment are denoted by the same reference numerals, and detailed description thereof will be omitted. For example, the computer system SYS is a server system.

  The computer system SYS has a device rack RACK in which a plurality of casings CS in which the package substrate PKGB shown in FIG. The internal structure of each casing CS is the same as that shown in FIG. 11 except that the fan FAN is not incorporated. That is, each package substrate PKGB adjusts the aperture ratio of the shutter SHT according to the power consumption (heat generation amount) of each processor CPU1.

  The fan FAN is attached in common to the plurality of casings CS above the exhaust space ES formed on the shutter SHT side of the casing CS in the device rack RACK. A pressure sensor PS for detecting a gauge pressure, which is a pressure difference between the pressure in the exhaust space ES in the device rack RACK and the pressure outside the device rack RACK, is attached at a position near the fan FAN in the exhaust space ES. ing. The fan control circuit FCNT controls the rotation speed of the fan FAN based on the gauge pressure detected by the pressure sensor PS. In the actual computer system SYS, the fan control circuit FCNT is mounted on the main control board stacked and housed in the device rack RACK together with the package board PKGB, or is mounted on one of the package boards PKGB. The operation of the fan control circuit FCNT is shown in FIG.

  FIG. 14 shows an example of fan control of the computer system SYS shown in FIG. First, the fan control circuit FCNT detects a gauge pressure based on information from the pressure sensor PS in step S20. In steps S22 and S24, the fan control circuit FCNT compares the pressure difference indicated by the gauge pressure with a reference value. When the pressure difference is equal to the reference value, the process returns to step S20. When the pressure difference is smaller than the reference value, the process proceeds to step S26, and when the pressure difference is greater than the reference value, the process proceeds to step S28.

  In step S26, the fan control circuit FCNT increases the rotational speed of the fan FAN and increases the amount of air flowing through the exhaust space ES. In step S28, the fan control circuit FCNT decreases the rotational speed of the fan FAN and decreases the amount of air flowing through the exhaust space ES.

  For example, in each package substrate PKGB, when the heat generation amount (power consumption) of the processor CPU1 is large and the aperture ratio of the shutter SHT is large, the pressure in the exhaust space ES is close to the pressure outside the device rack RACK, and the gauge pressure is small. . Further, in each package substrate PKGB, when the heat generation amount (power consumption) of the processor CPU1 is small and the aperture ratio of the shutter SHT is small, the pressure in the exhaust space ES becomes lower than the pressure outside the device rack RACK, and the gauge pressure becomes large. .

  As described above, the fan control circuit FCNT increases the rotation speed of the fan FAN in order to increase the air volume when the heat generation amount of the processor CPU1 is large and the gauge pressure is smaller than the reference value. The rotation speed of the fan FAN is increased until the gauge pressure reaches the reference value due to the increase in the air volume. On the other hand, the fan control circuit FCNT reduces the rotational speed of the fan FAN in order to reduce the air volume when the heat generation amount of the processor CPU1 is small and the gauge pressure is larger than the reference value. The rotation speed of the fan FAN is decreased until the gauge pressure reaches the reference value due to the decrease in the air volume. By repeating the flow shown in FIG. 14, the rotation speed of the fan FAN is controlled so that the gauge pressure is always constant regardless of the aperture ratio of the shutter SHT. Thereby, the semiconductor chip mounted on each package substrate PKGB can be efficiently cooled by the minimum number of fan FANs, and the power consumption of the fan FAN can be minimized.

  FIG. 15 shows an example of operations of the processor CPU1, the shutter SHT, and the fan FAN in the computer system SYS shown in FIG. FIG. 15 shows the operation of one package substrate PKGB in the device rack RACK.

  First, the power is turned on to the package substrate PKGB by the power-on PON of the computer system SYS (FIG. 15A). While the processor CPU1 is in an idle state, for example, the temperature of the processor CPU1 is relatively low (FIG. 15B). For this reason, the aperture ratio of the shutter SHT is maintained at, for example, 30% (FIG. 15C). When the aperture ratio of the shutter SHT is low, the gauge pressure is large, so the rotation speed of the fan FAN is set low (FIG. 15 (d)).

  When the processor CPU1 starts processing the task TSK, the temperature of the processor CPU1 rises (FIG. 15 (e)). For example, the power consumption of the processor CPU1 is maximized. As the temperature of the processor CPU1 rises, the aperture ratio of the shutter SHT increases and reaches 100% at the maximum power (FIG. 15 (f)). When the aperture ratio of the shutter SHT is high, the gauge pressure is small, so the rotational speed of the fan FAN is set high (FIG. 15 (g)). Note that the idle state and the operation state in FIG. 15 may be considered to be when the load of the task TSK processed by the processor CPU1 is small and when the load of the task TSK processed by the processor CPU1 is large, respectively.

  As described above, also in this embodiment, the same effect as that of the above-described embodiment can be obtained. Further, when a common fan FAN is provided for a plurality of package substrates PKGB, the reliability of the semiconductor chip mounted on the package substrate PKGB is controlled by controlling the rotation speed of the fan FAN based on the gauge pressure of the exhaust space ES. The power consumption of the cooling system can be minimized without lowering.

The invention described in the above embodiments is organized and disclosed as an appendix.
(Appendix 1)
A plurality of processors each processing a plurality of tasks that occur sequentially;
A calculation unit that sequentially receives task information indicating the content of the task, and calculates a load applied to the processor to process each task;
An assignment unit that assigns a task with a relatively large load to a first processor that is one of the processors and assigns a task with a relatively small load to the remaining processors for each task group including a predetermined number of consecutive tasks. And a computer system comprising:
(Appendix 2)
A board on which the processor is mounted;
A temperature sensor disposed on the substrate in a position proximate to the first processor;
The computer system according to claim 1, further comprising: a cooling unit that changes an amount of airflow flowing on the substrate according to a temperature detected by the temperature sensor, and increases the amount of airflow as the temperature increases. .
(Appendix 3)
The cooling part is
A fan for generating an airflow on the substrate;
The computer system according to claim 2, further comprising a control circuit that increases the rotational speed of the fan as the temperature detected by the temperature sensor increases.
(Appendix 4)
The cooling part is
A fan for generating an airflow on the substrate;
The computer system according to claim 2, further comprising: a shutter that is freely openable and closable on a downstream side or an upstream side of the airflow in the substrate, and that has a larger opening as the temperature detected by the temperature sensor increases. .
(Appendix 5)
The temperature sensor is a thermoactuator that is disposed in contact with the first processor, and a movable part moves according to the temperature of the first processor,
The computer system according to appendix 4, wherein the shutter has the opening adjusted in size according to movement of the movable part.
(Appendix 6)
An apparatus rack for storing a plurality of the substrates;
The computer system according to appendix 4, wherein the fan is provided in common to the plurality of boards.
(Appendix 7)
A pressure sensor for detecting a pressure difference between the atmospheric pressure inside the device rack and the atmospheric pressure outside the device rack;
The computer system according to claim 6, further comprising: a control circuit that changes the rotational speed of the fan according to the pressure difference, and increases the rotational speed of the fan as the pressure difference is smaller.
(Appendix 8)
The computer system according to any one of appendix 1 to appendix 7, wherein the allocating unit sets an operating frequency of the first processor to be higher than an operating frequency of the remaining processors.
(Appendix 9)
Each task group includes tasks that are multiples of “n” (positive numbers) when the number of processors is “n”;
The computer system according to any one of appendix 1 to appendix 8, wherein the assigning unit assigns an equal number of tasks in each task group to the processor.
(Appendix 10)
In order to receive the task information indicating the contents of a plurality of tasks that are sequentially generated to be processed by a plurality of processors,
Determining the load on the processor to process each task;
For each task group including a predetermined number of consecutive tasks, a task with a relatively high load is assigned to the first processor that is one of the processors, and a task with a relatively low load is assigned to the remaining processors. An operating method of a computer system.
(Appendix 11)
The amount of airflow flowing on the substrate is changed in accordance with the temperature detected by a temperature sensor arranged at a position close to the first processor on the substrate on which the processor is mounted. The operation method of the computer system according to appendix 10, characterized in that
(Appendix 12)
The computer system operating method according to appendix 11, wherein the higher the temperature detected by the temperature sensor, the more the number of rotations of the fan for generating an airflow on the substrate is increased.
(Appendix 13)
An air flow is generated on the substrate by a fan,
12. The method of operating a computer system according to claim 11, wherein the higher the temperature detected by the temperature sensor, the larger the opening of the shutter that can be opened and closed downstream or upstream of the airflow in the substrate. .
(Appendix 14)
The operating method of the computer system according to any one of appendix 10 to appendix 13, wherein an operating frequency of the first processor is set higher than an operating frequency of the remaining processors.
(Appendix 15)
Each task group includes tasks that are multiples of “n” (positive numbers) when the number of processors is “n”;
The operation method of the computer system according to any one of Supplementary Note 10 to Supplementary Note 14, wherein the assigning unit assigns an equal number of tasks in each task group to the processor.
(Appendix 16)
The computer system according to claim 4, wherein the airflow generated by at least one of the fans is distributed to a plurality of the shutters.

  From the above detailed description, features and advantages of the embodiments will become apparent. This is intended to cover the features and advantages of the embodiments described above without departing from the spirit and scope of the claims. Further, any person having ordinary knowledge in the technical field should be able to easily come up with any improvements and modifications, and there is no intention to limit the scope of the embodiments having the invention to those described above. It is also possible to rely on suitable improvements and equivalents within the scope disclosed in.

  ASGN allocator; CALC calculator; CCNT clock controller; CLK1, CLK2 clock; CNT controller; CPU1, CPU2 processor; CS casing; ES exhaust space; FAN fan; Control circuit; HDD, hard disk drive, HS, heat sink, LD, load information, MCNT, memory control unit, MEM1, MEM2, memory, MTR, stepping motor, PKGB, package substrate, PL, pulley, PS, pressure sensor, PSU Power unit; PT ... Bulkhead; RACK ... Equipment frame; SA ... Thermoactuator; SHT ... Shutter; SPR ... Spring; SYS ... Computer system; TCNT ... Task controller; TINF ... Task information; TSE, TSI ... Temperature sensor; Task; TSK ‥ Task Group

Claims (7)

A plurality of processors each processing a plurality of tasks that occur sequentially;
A calculation unit that sequentially receives task information indicating the content of the task, and calculates a load applied to the processor to process each task;
An assignment unit that assigns a task with a relatively large load to a first processor that is one of the processors and assigns a task with a relatively small load to the remaining processors for each task group including a predetermined number of consecutive tasks. When
A temperature sensor disposed at a position close to the first processor on a substrate on which the processor is mounted;
A computer system comprising: a cooling unit that changes an amount of airflow flowing on the substrate in accordance with a temperature detected by the temperature sensor, and increases the amount of airflow as the temperature increases . The cooling part is
A fan for generating an airflow on the substrate;
2. The computer according to claim 1 , further comprising: a shutter that is freely openable and closable on a downstream side or an upstream side of the airflow in the substrate, and that has a larger opening as the temperature detected by the temperature sensor is higher. system. The temperature sensor is a thermoactuator that is disposed in contact with the first processor, and a movable part moves according to the temperature of the first processor,
The computer system according to claim 2 , wherein a size of the opening of the shutter is adjusted according to movement of the movable portion. An apparatus rack for storing a plurality of the substrates;
The computer system according to claim 2 , wherein the fan is provided in common to the plurality of boards. A pressure sensor for detecting a pressure difference between the atmospheric pressure inside the device rack and the atmospheric pressure outside the device rack;
The computer system according to claim 4 , further comprising: a control circuit that changes a rotation speed of the fan according to the pressure difference, and increases the rotation speed of the fan as the pressure difference is smaller. The computer system according to any one of claims 1 to 5 , wherein the allocating unit sets an operating frequency of the first processor to be higher than an operating frequency of the remaining processors. In order to receive the task information indicating the contents of a plurality of tasks that are sequentially generated to be processed by a plurality of processors,
Determining the load on the processor to process each task;
For each task group including a predetermined number of consecutive tasks, a task with a relatively high load is assigned to the first processor that is one of the processors, and a task with a relatively low load is assigned to the remaining processors ,
Arranged at a position close to the first processor on the board on which the processor is mounted
The amount of airflow flowing on the substrate is changed according to the temperature detected by the temperature sensor, and the temperature
A method of operating a computer system, characterized by increasing the amount of airflow as the height increases .

Images