JP2005141669A - Grid computing and load distribution method in grid computing - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To distribute a load according to the margin of exhaust heat of each node computer in grid computing. <P>SOLUTION: Each node computer 1 is provided with an ambient temperature acquisition part 11 for acquiring ambient temperature, and an inside temperature acquisition part 12 for acquiring inside temperature; and a process distribution calculation device 2 for acquiring the ambient temperature and the inside temperature along with a CPU load from each node computer 1 to calculate, on a node computer basis, an allowable processing amount without exceeding a limit temperature specific to each node computer based on them, and a processing distribution device 3 for acquiring the allowable processing amount for every node computer from the calculation device 2 to distribute work within the allowable processing amount to each node computer are installed on a network 5. Thereby, each node computer 1 is allowed to execute work without exceeding the limit temperature. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to grid computing and a load balancing method in grid computing.

A technology that builds a huge virtual computer using the surplus CPU capability (surplus processing capability) of a plurality of computers (hereinafter referred to as node computers) connected to a network and obtains processing performance that cannot be realized by a single computer. A grid computing has been proposed and is currently being tested.
"Cell computing large-scale demonstration experiment final report (June 26, 2003)", NTT DATA Corporation [searched October 30, 2003], Internet <URL: http://www.cellcomputing.jp/test /index2.html>

  In conventional grid computing, node computers with low heat exhaustion, such as node computers in low-latitude regions and node computers in summer, are distinguished, and work is distributed according to the surplus CPU capacity of each node computer. Since the load was applied, the total amount of energy for cooling the node computer became enormous, and the effect on the global environment could not be ignored.

  In the present invention, in order to solve the above problems, each node computer is provided with an ambient temperature acquisition unit and an internal temperature acquisition unit, and a processing distribution calculation device, a processing distribution device, and a processing result collecting device are provided on the network.

  Alternatively, in the present invention, in order to solve the above problem, each node computer includes an ambient temperature acquisition unit, an internal temperature acquisition unit and an allowable processing amount calculation unit, or an ambient temperature acquisition unit, an internal temperature acquisition unit, a battery remaining amount acquisition unit, and In addition to providing an allowable processing amount calculation unit, a processing distribution device and a processing result collecting device were provided on the network.

  According to the present invention, since more work can be distributed to a computer with sufficient heat exhaustion, the total energy required for cooling compared to conventional grid computing that simply performs work distribution considering only surplus CPU capacity. Is less.

  FIG. 1 shows a first embodiment of grid computing according to the present invention, here, an embodiment corresponding to claims 1, 3 and 4 (11, 13 and 14). Node computer (# 1, # 2,...), 2 is a processing distribution calculation device, 3 is a processing distribution device, 4 is a processing result collecting device, 5 is each node computer 1, processing distribution calculation device 2, and processing distribution device. 3 and the processing result collecting apparatus 4.

  Each node computer 1 includes an ambient temperature acquisition unit 11 that acquires an ambient temperature and an internal temperature acquisition unit 12 that acquires an internal temperature, in addition to the function as a normal computer. The ambient temperature and internal temperature acquired by the acquisition unit 12 are transmitted to the processing distribution calculation device 2 via the network 5 together with the CPU load.

  The processing distribution calculation device 2 calculates an allowable processing amount for each node computer based on the ambient temperature, the internal temperature, and the CPU load received from each node computer 1 for each node computer. To the processing distribution device 3.

  The processing distribution device 3 distributes the work within the allowable processing amount for each received node computer to each node computer 1 via the network 5.

  The processing result collecting apparatus 4 acquires the work result from each node computer 1 via the network 5.

  There are two types of allowable processing amount calculation algorithms used in the processing distribution calculation device 2 depending on the presence or absence of temperature prediction, which are shown in flowcharts in FIGS. 2 and 3, respectively. The allowable processing amount calculation and the work distribution are performed at an arbitrary cycle within the range allowed by the calculation capacity of the processing distribution calculation device 2 and the bandwidth of the network 5.

  FIG. 2 is a flowchart showing an allowable processing amount calculation algorithm without temperature prediction.

  First, it is determined whether the internal temperature is higher or lower than the limit temperature. If the internal temperature is high, the allowable processing amount = 0 and the process ends. The limit temperature is a unique numerical value determined for each node computer. If the internal temperature is low, the CPU load (%) is determined. If the internal temperature is 100% or more, the allowable processing amount = 0 and the process ends. Furthermore, if the CPU load is less than 100%, the processing is terminated as allowable processing amount = (100% −current CPU load) × CPU capacity. The CPU capacity is a unique numerical value determined for each node computer.

  FIG. 3 is a flowchart showing an allowable processing amount calculation algorithm with temperature prediction.

  First, it is determined whether the internal temperature is higher or lower than the limit temperature. If the internal temperature is high, the allowable processing amount = 0 and the process ends. The limit temperature is a unique numerical value determined for each node computer. If the internal temperature is low, the CPU load (%) is determined. If the internal temperature is 100% or more, the allowable processing amount = 0 and the process ends. Furthermore, if the CPU load is less than 100%, the allowable processing amount is calculated by the equation (4) described later, and the process is terminated. The CPU capacity is a unique numerical value determined for each node computer.

  Hereinafter, calculation of the allowable processing amount when there is temperature prediction will be described using equations (1) to (4).

The heat input and exhaust heat of the node computer is
Heat input = CPU load (%) × heat generation coefficient (W /%) + fixed heat generation amount (W) (1)
Waste heat = (ambient temperature-internal temperature) x thermal conductivity coefficient (W / K) (2)
It is represented by

The internal temperature is in an equilibrium state when heat input = exhaust heat. Therefore, the maximum CPU load under the condition of internal temperature ≦ limit temperature is obtained by solving equations (1) and (2) with respect to the CPU load. Maximum CPU load = {(ambient temperature−limit temperature) × heat conduction coefficient−fixed heat generation amount} / heat generation coefficient (3)
It becomes.

Therefore, the allowable processing amount is: allowable processing amount = maximum CPU load−current left CPU load = {(ambient temperature−limit temperature) × thermal conductivity coefficient−fixed heat generation amount} / heat generation coefficient−current CPU load (4)
It is represented by Here, the limit temperature, the heat conduction coefficient, the fixed heat generation amount, and the heat generation coefficient are unique numerical values for each node computer.

  In this embodiment, the processing distribution calculation device 2, the processing distribution device 3, and the processing result summarizing device 4 are configured as separate devices, but one device including all or two devices including any combination. You may comprise as an apparatus of. These devices can also be implemented as hardware or software on a node computer.

  FIG. 4 shows a second embodiment of grid computing according to the present invention, here an embodiment corresponding to claim 2 (12). In FIG. 4, the same components as those of the first embodiment are shown. It represents with the same code. That is, 1b is a plurality of node computers (# 1, # 2,...), 3 is a processing distribution device, 4 is a processing result collecting device, 5 is each node computer 1b, processing distribution device 3, and processing result collecting device 4. The network to connect to.

  Each node computer 1b is based on the ambient temperature acquisition unit 11, the internal temperature acquisition unit 12, the ambient temperature, the internal temperature, and the CPU load in addition to the function as a normal computer. An allowable processing amount calculation unit 13 that calculates an allowable processing amount that does not exceed a specific limit temperature by an amount calculation algorithm, and transmits the calculated allowable processing amount to the process distribution device 3 via the network 5.

  Other configurations and operations are the same as those in the first embodiment.

  FIG. 5 shows a third embodiment of grid computing according to the present invention, here, an embodiment corresponding to claims 5 to 7 (15 to 17). In the figure, the first and second embodiments are shown. Constituent parts identical to those in FIG. That is, 1c is a plurality of node computers (# 1, # 2,...), 3 is a processing distribution device, 4 is a processing result collecting device, 5 is each node computer 1c, processing distribution device 3, and processing result collecting device 4. The network to connect to.

  Each node computer 1c has an ambient temperature acquisition unit 11, an internal temperature acquisition unit 12, a battery remaining amount acquisition unit 14 that acquires a remaining battery level, an ambient temperature, an internal temperature, And an allowable processing amount calculation unit 15 that calculates an allowable processing amount that does not exceed a specific limit temperature based on the remaining battery level and the CPU load. Send to.

  There are two types of allowable processing amount calculation algorithms used in the allowable processing amount calculation unit 15 depending on the presence or absence of temperature prediction, which are shown in flowcharts in FIGS. 6 and 7, respectively. The allowable processing amount calculation and the work distribution are performed at an arbitrary cycle within a range allowed by the calculation capability of the allowable processing amount calculation unit 15 (node computer 1c) and the bandwidth of the network 5.

  FIG. 6 is a flowchart showing an allowable processing amount calculation algorithm without temperature prediction.

  First, it is determined whether the internal temperature is higher or lower than the limit temperature. If the internal temperature is high, the allowable processing amount = 0 and the process ends. The limit temperature is a unique numerical value determined for each node computer. If the internal temperature is low, it is determined whether or not the remaining amount of the battery is less than the limit battery remaining amount. The limit battery remaining capacity is a unique numerical value determined for each node computer. If it is equal to or greater than the limit battery remaining amount, the CPU load (%) is determined. Furthermore, if the CPU load is less than 100%, the processing is terminated as allowable processing amount = (100% −current CPU load) × CPU capacity. The CPU capacity is a unique numerical value determined for each node computer.

  FIG. 7 is a flowchart showing an allowable processing amount calculation algorithm with temperature prediction.

  First, it is determined whether the internal temperature is higher or lower than the limit temperature. If the internal temperature is high, the allowable processing amount = 0 and the process ends. The limit temperature is a unique numerical value determined for each node computer. If the internal temperature is low, it is determined whether or not the remaining amount of the battery is less than the limit battery remaining amount. The limit battery remaining capacity is a unique numerical value determined for each node computer. If it is equal to or greater than the limit battery remaining amount, the CPU load (%) is determined. Furthermore, if the CPU load is less than 100%, the allowable processing amount is calculated by the above-described equation (4), and the process is terminated. The CPU capacity is a unique numerical value determined for each node computer.

  Other configurations and operations are the same as those in the first and second embodiments.

  In the second and third embodiments, the processing distribution device 3 and the processing result collecting device 4 are configured as separate devices, but may be configured as a single device including all of them. These devices can also be implemented as hardware or software on a node computer.

  FIG. 8 shows an embodiment for performing grid computing with node computers distributed throughout the world. In this example, there are a node computer Z on which a processing distribution calculation device, a processing distribution device, and a processing result collecting device are mounted, and node computers A, B, C, D, and E. By giving a large work load to the node computers A, B, and E having a low ambient temperature, the cooling energy of the entire system is lowered.

  FIG. 9 shows an embodiment in which grid computing is performed by node computers arranged in a rack. In this example, there are a node computer Z on which a processing distribution calculation device, a processing distribution device, and a processing result collecting device are mounted, and node computers A, B, C, D, and E. By giving a large work load to the node computers C, D, and E having a low ambient temperature, the cooling energy of the entire rack is reduced.

  FIG. 10 shows an embodiment in which grid computing is used when reading a digital watermark by a mobile phone (node computer). Since reading of a digital watermark requires a lot of calculation processing, it is difficult to process with one mobile phone, but it can be read out by a plurality of devices working together. In this case, the processing distribution device and the calculation result collecting device are mounted on a terminal that reads out the digital watermark. Although the cooperation terminal shares the reading process, since the process is not distributed to the terminal just before the battery runs out or the terminal having a high CPU load, the user of the cooperation terminal can participate in the sharing with peace of mind.

  FIG. 11 shows an embodiment in which grid computing is used in a network game where a user of a mobile phone (node computer) forms a team and cooperates in the team while competing with other teams. The node computer Z of the game provider takes charge of the processing distribution device and the calculation result collecting device. Game processing will not be distributed to mobile phones that run out of battery power during the game, or that want to allocate CPU load to other processes. You can prevent problems that occur.

  FIG. 12 shows an embodiment in which the exhaust heat in the node computer is not simply discarded but actively used. Here, A, B, C, D, and E are node computers, but the node computer E has a function of generating power using exhaust heat. In this case, by giving more work load to the node computer E, the cooling energy other than the node computer E can be saved and at the same time the power generation amount can be increased.

  FIG. 13 shows another embodiment in which exhaust heat in the node computer is positively used. Here, A, B, C, D, and E are node computers, but the node computer E has a hot water pool or a public bath that directly uses exhaust heat. In this case, by giving more work load to the node computer E, the cooling energy other than the node computer E can be saved, and at the same time, the temperature of the hot water can be increased or the amount of available hot water can be increased.

The block diagram which shows 1st Embodiment of the grid computing of this invention Flowchart showing allowable processing amount calculation algorithm without temperature prediction Flow chart showing algorithm for calculating allowable processing amount with temperature prediction The block diagram which shows 2nd Embodiment of the grid computing of this invention The block diagram which shows 3rd Embodiment of the grid computing of this invention Flowchart showing allowable processing amount calculation algorithm without temperature prediction Flow chart showing algorithm for calculating allowable processing amount with temperature prediction The block diagram which shows the 1st Example of the grid computing of this invention The block diagram which shows the 2nd Example of the grid computing of this invention The block diagram which shows the 3rd Example of the grid computing of this invention The block diagram which shows the 4th Example of the grid computing of this invention The block diagram which shows the 5th Example of the grid computing of this invention The block diagram which shows the 6th Example of the grid computing of this invention

Explanation of symbols

  1, 1b, 1c: node computers (# 1, # 2,...), 2: processing distribution calculation device, 3: processing distribution device, 4: processing result collecting device, 5: network, 11: ambient temperature acquisition unit, 12: Internal temperature acquisition unit, 13, 15: Allowable processing amount calculation unit, 14: Battery remaining amount acquisition unit.

Claims (17)

In grid computing that constructs a huge virtual computer using the surplus CPU capability of a plurality of node computers connected to a network,
Each node computer is provided with an ambient temperature acquisition unit that acquires the ambient temperature and an internal temperature acquisition unit that acquires the internal temperature,
On the network,
A processing distribution calculation device that obtains an ambient temperature and an internal temperature from each node computer together with a CPU load, and calculates an allowable processing amount for each node computer that does not exceed a limit temperature specific to each node computer based on these,
A processing distribution device that acquires an allowable processing amount for each node computer from the processing distribution calculation device, and distributes work within the allowable processing amount to each node computer;
Grid computing characterized by comprising a processing result collecting device for acquiring the work result from each node computer. In grid computing that constructs a huge virtual computer using the surplus CPU capability of a plurality of node computers connected to a network,
For each node computer, an ambient temperature acquisition unit that acquires the ambient temperature, an internal temperature acquisition unit that acquires the internal temperature, and an allowable processing amount that does not exceed a specific limit temperature based on the ambient temperature, internal temperature, and CPU load And an allowable processing amount calculation unit for
On the network,
A process distribution device that acquires an allowable processing amount from each node computer and distributes work within the allowable processing amount to each node computer;
Grid computing characterized by comprising a processing result collecting device for acquiring the work result from each node computer. The grid computing according to claim 1 or 2,
The allowable processing amount is 0 when the internal temperature is higher than the limit temperature or the CPU load is 100% or more, and when the internal temperature is the limit temperature or less and the CPU load is less than 100%, the current surplus CPU load is set to each node computer. Grid computing, characterized by a value multiplied by a unique CPU capability. The grid computing according to claim 1 or 2,
The allowable processing amount is 0 when the internal temperature is higher than the limit temperature or the CPU load is 100% or more, and when the internal temperature is lower than the limit temperature and the CPU load is less than 100%, the thermal conductivity coefficient specific to each node computer, Using fixed calorific value and heat generation coefficient,
{(Ambient temperature-Limit temperature) x Thermal conductivity coefficient-Fixed calorific value} / A value obtained by subtracting the current CPU load from the maximum CPU load obtained from the heat generation coefficient. In grid computing that constructs a huge virtual computer using the surplus CPU capability of a plurality of node computers connected to a network,
Each node computer has an ambient temperature acquisition unit that acquires the ambient temperature, an internal temperature acquisition unit that acquires the internal temperature, a battery remaining level acquisition unit that acquires the remaining battery level, the ambient temperature, the internal temperature, and the remaining battery level. An allowable processing amount calculation unit that calculates an allowable processing amount that does not exceed the inherent limit temperature based on the amount and the CPU load, and
On the network,
A process distribution device that acquires an allowable processing amount from each node computer and distributes work within the allowable processing amount to each node computer;
Grid computing characterized by comprising a processing result collecting device for acquiring the work result from each node computer. The grid computing according to claim 5.
The allowable processing amount is 0 when the internal temperature is higher than the limit temperature, or the remaining battery level is lower than the limit battery level specific to each node computer, or the CPU load is 100% or more. Grid computing, characterized in that if the remaining amount of battery exceeds the limit battery remaining amount and the CPU load is less than 100%, the current surplus CPU load is multiplied by the CPU capacity unique to each node computer. The grid computing according to claim 5.
The allowable processing amount is 0 when the internal temperature is higher than the limit temperature, or the remaining battery level is lower than the limit battery level specific to each node computer, or the CPU load is 100% or more. If the remaining battery capacity is more than the limit battery remaining capacity and the CPU load is smaller than 100%, the heat conduction coefficient, fixed heat generation amount and heat generation coefficient specific to each node computer are used.
{(Ambient temperature-Limit temperature) x Thermal conductivity coefficient-Fixed calorific value} / A value obtained by subtracting the current CPU load from the maximum CPU load obtained from the heat generation coefficient. The grid computing of claim 1.
Grid computing characterized in that the processing distribution calculation device, the processing distribution device, and the processing result collecting device are mounted on one node computer. The grid computing according to claim 2 or 5,
Grid computing characterized in that the processing distribution device and the processing result collecting device are mounted on one node computer. The grid computing according to any one of claims 1 to 9,
Grid computing, characterized in that at least one of the node computers is equipped with a power generation facility that uses exhaust heat, a hot water pool, or a public bath. A load balancing method in grid computing for constructing a huge virtual computer using surplus CPU capability of a plurality of node computers connected to a network,
Acquire ambient temperature and internal temperature from each node computer along with CPU load,
Based on these, the allowable processing amount not exceeding the limit temperature specific to each node computer is calculated for each node computer,
Distributing the work within the allowable processing amount for each calculated node computer to each node computer,
A load balancing method in grid computing, wherein the work result is obtained from each node computer. A load balancing method in grid computing for constructing a huge virtual computer using surplus CPU capability of a plurality of node computers connected to a network,
From each node computer, based on its ambient temperature, internal temperature and CPU load, obtain the allowable processing amount calculated as a value not exceeding the specific limit temperature,
Allocating work within the allowable processing amount to each node computer,
A load balancing method in grid computing, wherein the work result is obtained from each node computer. A load balancing method in grid computing according to claim 11 or 12,
The allowable processing amount is 0 when the internal temperature is higher than the limit temperature or the CPU load is 100% or more, and when the internal temperature is the limit temperature or less and the CPU load is less than 100%, the current surplus CPU load is set to each node computer. A load balancing method in grid computing, which is a value obtained by multiplying a unique CPU capability. A load balancing method in grid computing according to claim 11 or 12,
The allowable processing amount is 0 when the internal temperature is higher than the limit temperature or the CPU load is 100% or more, and when the internal temperature is lower than the limit temperature and the CPU load is less than 100%, the thermal conductivity coefficient specific to each node computer, Using fixed calorific value and heat generation coefficient,
{(Ambient temperature-Limit temperature) x Thermal conductivity coefficient-Fixed heat value} / A value obtained by subtracting the current CPU load from the maximum CPU load obtained from the heat generation coefficient. A load balancing method in grid computing. In grid computing that constructs a huge virtual computer using the surplus CPU capability of a plurality of node computers connected to a network,
From each node computer, based on the ambient temperature, internal temperature, battery remaining amount and CPU load, obtain the allowable processing amount calculated as a value not exceeding the specific limit temperature,
Allocating work within the allowable processing amount to each node computer,
A load balancing method in grid computing, wherein the work result is obtained from each node computer. The load balancing method in grid computing according to claim 15,
The allowable processing amount is 0 when the internal temperature is higher than the limit temperature, or the remaining battery level is lower than the limit battery level specific to each node computer, or the CPU load is 100% or more. Load distribution method in grid computing, wherein the remaining surplus CPU load is equal to or more than the limit battery remaining amount and the CPU load is less than 100%, and the current surplus CPU load is multiplied by the CPU capability specific to each node computer . The load balancing method in grid computing according to claim 15,
The allowable processing amount is 0 when the internal temperature is higher than the limit temperature, or the remaining battery level is lower than the limit battery level specific to each node computer, or the CPU load is 100% or more. If the remaining battery capacity is more than the limit battery remaining capacity and the CPU load is smaller than 100%, the heat conduction coefficient, fixed heat generation amount and heat generation coefficient specific to each node computer are used.
{(Ambient temperature-Limit temperature) x Thermal conductivity coefficient-Fixed heat value} / A value obtained by subtracting the current CPU load from the maximum CPU load obtained from the heat generation coefficient. A load balancing method in grid computing.

Images