TW201520744A - Fan controller and server system with the fan controller - Google Patents

Abstract

A server system comprises a fan backplate, a server array, a substrate, a first fan controller and a second fan controller. The fan backplate includes fans. The server array includes calculation nodes. The substrate has a multiplexer. The calculation nodes couple with the multiplexer. The first fan controller and the second fan controller couple with the calculation nodes through the multiplexer. A fan control signal is generated according to the real-time temperature of the calculation nodes to control the fans. The first fan controller and the second fan controller form a redundancy system.

Description

Fan controller and server system having the same

The present invention relates to a fan controller, and more particularly to a fan controller applied to a server.

For the server, heat control has always been an important part of it. In previous server applications, the main cooling method was implemented by each server's own internal fan. With the combined use of multiple servers, it is necessary to use a fan equivalent to the number of servers. Especially in the application of the Microserver array, in a 2U (Unit, a unit representing the external dimensions of the server), including 12 CPU boards, and each central The processor substrate in turn contains four independent System on Chips. Each system level chip is a separate server system.

Traditionally, all central processor substrates in each cabinet are controlled by a unified fan control system for fan cooling. Once this fan control system is damaged, it will cause the cabinet to not dissipate heat, even destroying the internal server. In view of the above, it is necessary to provide a fan control system that can solve the above problems applied to a server.

In view of the above, the present invention provides a fan control system for a server to provide reliable heat dissipation control of the server.

One aspect of the present invention provides a server system having a fan backplane, a server array, a substrate, and a first fan controller and a second fan controller. The fan backboard is provided with a plurality of fans. The server array contains a plurality of compute nodes. The substrate has a multiplexer, and the computing node is coupled to the multiplexer. The first fan controller and the second fan controller are coupled to the computing node through the multiplexer, and generate a fan control signal according to the instantaneous temperature of the computing node to control the fan speeds, where the first fan controller and the second fan controller Mutual redundancy.

In an embodiment, each of the first fan controller and the second fan controller further includes a control unit, a current monitoring unit, a power module, and a plurality of current sampling units and a plurality of switches. The current monitoring unit is coupled to the control unit. The power module transmits power to the fan through a plurality of power supply lines. The current sampling units and switches are respectively disposed on a plurality of power supply branch lines, and the current sampling units are coupled to the current monitoring unit. These switches are coupled to the control unit. The control unit is configured to control the opening and closing of the switches, and the current monitoring unit samples the current values flowing through the current sampling units. When the current value flowing through one of the current sampling units exceeds a set value, the current is The current monitoring unit sends a current overload signal to the control unit to control the corresponding switch to be turned off to stop the power module from transmitting power to the corresponding fan.

In one embodiment, the switch is a transistor, wherein the transistor is coupled to the branch through its source terminal and the drain terminal, and the control unit is coupled through the gate terminal.

In an embodiment, when a fan is damaged, the control unit generates a fan damage indication signal to illuminate a corresponding indicator light.

In one embodiment, the fans respectively correspond to a printed circuit board for mounting the corresponding indicator light.

In an embodiment, the control unit is further coupled to a heat generating unit, and generates a fan control signal according to the instantaneous temperature of the heat generating unit to control the fan speeds.

In an embodiment, the control unit stores a fan speed control table, wherein the fan speed control table is used to indicate a correspondence between the temperature value of the heat generating unit and the fan speed, and the control unit selects the fan according to the instantaneous temperature value of the heat generating unit. A set speed value is obtained in the speed control table, and the fan control signal is generated according to the set speed value.

In an embodiment, the fan further generates a feedback speed signal to the control unit, and the control unit determines the feedback speed of the fan according to the feedback speed signal, and determines that the fan appears when the feedback speed does not match a set speed value. malfunction.

In an embodiment, if the control unit controls the switch to turn off and receives the feedback speed signal, it is determined that the fan control unit is faulty.

In an embodiment, the control unit further performs a self-test procedure, and if it detects that the fan speed control table cannot be read and written normally, determining the fan control The device has failed.

In an embodiment, the plurality of fans receive the fan control signal through a first connector, and the second connector is coupled to the power module.

In an embodiment, the control unit of the first fan controller is coupled to the second fan controller via a universal serial input/output bus.

In an embodiment, the method further includes: when the second fan controller cannot obtain the information of the first fan controller through the universal serial input/output bus, determining that the first fan controller is damaged, by the second The fan controller controls the fan speeds.

In an embodiment, the second fan controller is further notified by the first fan controller through the universal serial input/output bus, and the fan speed is controlled by the second fan controller.

In summary, the present invention additionally provides a fan controller as a backup component, which avoids the danger of the server system being unable to dissipate heat when the only fan controller is damaged in the design of the conventional single fan controller. Or the inconvenience caused by the need to stop the fan controller server system.

100‧‧‧Server system

110‧‧‧Compute node

120‧‧‧Substrate

130‧‧‧Fan backplane

140‧‧‧First Fan Controller

150‧‧‧Second fan controller

160‧‧‧Power Module

170‧‧‧Common Serial Input/Output Busbar

180‧‧‧Power cord

190‧‧‧ signal line

200‧‧‧Control unit

201‧‧‧ storage unit

210‧‧‧ Current monitoring unit

1201‧‧‧Multiplexer

1202 and 1203‧‧‧I2C busbars

1301~1306‧‧‧Fan

1307 and 1308‧‧‧ connectors

1801,1802,...,1806‧‧‧Power spur

R1~R6‧‧‧current sampling resistor

Q1~Q6‧‧‧ switch

I_OC1~I_OC6‧‧‧ Current overload signal

FAIL_LED‧‧‧Fan damage indicator

PWM<1..6>‧‧‧Fan control signal

TACH<1..6>‧‧‧Speed signal

1 is a schematic diagram of a server system in accordance with an embodiment of the present invention.

2 is a schematic diagram of a first fan controller according to an embodiment of the invention.

The following description of the preferred embodiments of the invention is in the

1 is a schematic diagram of a server system in accordance with an embodiment of the present invention. The server system 100 includes a server array having a plurality of computing nodes 110, a substrate 120, a fan backplane 130, a first fan controller 140, a second fan controller 150, and a power module 160. The plurality of computing nodes 110 are respectively coupled to the substrate 120. The substrate 120 selects the communication of the computing node 110 through a multiplexer 1201 to obtain the information of the computing node 110, and transmits the obtained information to the first fan controller 140 and the second fan through the I2C bus bars 1202 and 1203. 150. A plurality of fans 1301 to 1306 are disposed on the fan back plate 130. The power module 160 transmits a voltage signal to the first fan controller 140 and the second fan controller 150 through the power line 180, and then the first fan controller 140 and the second fan controller 150 transmit the voltage signal through the power line 180 to The plurality of fans 1301~1306 on the fan backplane 130, for example, provide a voltage signal of 12V to the plurality of fans 1301~1306. In use, according to the computing node 110 information transmitted by the multiplexer 1201 on the substrate 120, the first fan controller 140 or the second fan controller 150 may generate a control signal to control the fan backplane 130 via the signal line 190. The rotational speed of the plurality of fans 1301~1306. The power line 180 and the signal line 190 are respectively associated with the fan backplane 130. The upper connectors 1307 and 1308 are coupled because the fan backplane 130 has no other electronic components and has high reliability.

In this embodiment, the first fan controller 140 and the second fan controller 150 do not operate at the same time, that is, when the first fan controller 140 performs the speed control of the fans 1301~1306, the second fan controls The device 150 is in a standby state. On the other hand, when the second fan controller 150 performs the rotational speed control of the fans 1301 to 1306, the first fan controller 140 is in a standby state. The first fan controller 140 and the second fan controller 150 communicate with each other via a serial general purpose input/output (SGPIO) 170. By additionally providing a fan controller as a backup, it is possible to avoid the danger of the server system being unable to dissipate heat when the only fan controller is damaged in the design of a conventional single fan controller. Moreover, the present invention can replace the fan controller by means of hot plugging, so there is no inconvenience caused by the necessity of stopping the fan controller server system.

2 is a schematic diagram of a first fan controller according to an embodiment of the invention. The first fan controller 140 and the second fan controller 150 have the same structure. In the present embodiment, the first fan controller 140 controls six fans 1301 to 1306 as an example to illustrate the application of the present invention. The first fan controller 140 includes a control unit 200, a current monitoring unit (I-monitor) 210, and a power control module 160. Since the first fan controller 140 can simultaneously control the six fans 1301~1306, the power line 180 coupled to the power module 160 is divided into six power supply lines 1801, 1802, ... , 1806 to separately transmit voltage signals to 6 units Fans 1301~1306. The power supply lines 1801, 1802, ..., 1806 are respectively provided with current sampling units R1 to R6 and switches Q1 to Q6. The current sampling units R1 to R6 are coupled to the current monitoring unit 210, and the switches Q1 to Q6 are coupled to the control unit 201. Accordingly, after the power of the power module 160 is input to the first fan controller 140, the power supplies of the six fans 1301 to 1306 are respectively supplied through the current sampling resistors R1 to R6 and the switches Q1 to Q6. The control unit 200 is configured to respectively control the opening and closing of the switches Q1 to Q6, and the current signals of the current sampling units R1 R R6 are respectively input to the current monitoring unit 210, and the current monitoring unit 210 respectively monitors the power supply lines 1801, 1802, ... 1806 pairs the power supply status of the corresponding fan 1301. When the current monitoring unit 210 finds that the current flowing through one of the current sampling units R1 R R6 exceeds the set value, the current overload signals I_OC1~I_OC6 are respectively sent to the control unit 200, and one of the corresponding switches Q1~Q6 is controlled to be disconnected. The power module 160 is stopped to transmit power to the corresponding fan to prevent the current overload from damaging the fans 1301~1306. Once the fans 1301~1306 are damaged due to the current overload, the control unit 200 generates a fan damage indication signal FAIL_LED, lights up the corresponding indicator light, and informs the user that the fan needs to be replaced. In this embodiment, the replacement of the fan can be performed by hot plugging. The current sampling unit is a resistor. The switches Q1~Q6 are P-type transistor switches, wherein the transistors are coupled to the control unit 200 through the gate terminals through the branch lines 1801, 1802, ..., 1806 whose source terminals are coupled to the 汲 terminal. A printed circuit board can be embedded in the outer casing of each fan to install the indicator light of the LED, and when the fan failure is detected, the indicator light is illuminated.

The control unit 200 passes through the 12C bus bar 1202 from the substrate 120. The multiplexer 1201 communicates with each computing node 110 to obtain instant temperature information of a heat generating unit on the computing nodes 110. According to the instant temperature information, the control unit 200 can obtain, from the fan speed control table stored in the storage unit 201, a fan control signal PWM<1..6> corresponding to the instantaneous temperature setting speed value of each of the fans 1301~1306. The speeds of the fans 1301~1306 are controlled accordingly. The fan speed control table of the storage unit 201 stores a temperature value and a corresponding fan speed control signal. At the same time, the control unit 200 can determine the rotational speed state of the corresponding fans 1301~1306 through the rotational speed signals TACH<1..6> fed back by the fans 1301~1306. If the feedback speed does not match the set speed value, the control unit 200 determines that the corresponding fan 1301~1306 is faulty, and generates a fan damage indication signal FAIL-LED, illuminates the corresponding indicator light, and informs the user that the fan needs to be replaced. . At the same time, the control switches Q1~Q6 are disconnected, and the power supply to the faulty fan is stopped.

Moreover, the control unit 200 controls the switch to be disconnected, and the first fan controller 140 can be normally operated by the speed signals TACH<1..6> fed back by the fans 1301~1306. For example, when the control unit 200 controls the switch Q1 to be turned off, but the speed signal TACH<1> fed back by the fan 1301, it is still determined that the fan 1301 is in a normal operation state. At this time, the representative control unit 200 cannot control the switch normally, or The switch has been damaged. In addition, the control unit 200 can further perform a self-test procedure. If it is detected that the control unit 200 cannot correctly read and write the fan speed control table of the storage unit 201, it is determined that the fan controller 140 is faulty. At this time, it is determined that the first fan controller 140 cannot operate normally, and the first fan control The controller 140 notifies the second fan controller 150 of this condition via the universal serial input/output bus bar 170, and the second fan controller 150 acquires the control of the fans 1301~1306. In other words, in this case, the first fan controller 140 actively informs the second fan controller 150 to exit the control of the fans 1301 to 1306 through the universal serial input/output bus bar 170. In another embodiment, the first fan controller 140 and the second fan controller 150 can implement data synchronization through the universal serial input/output bus bar 170. Therefore, when the second fan controller 150 is in the periodic inquiry, if the synchronization signal input by the first fan controller 140 through the universal serial input/output bus bar 170 is not received, the first fan controller 140 is determined to be damaged. Or being removed, the second fan controller 150 takes control of the fans 1301~1306.

In summary, the present invention additionally provides a fan controller as a backup component, which avoids the danger of the server system being unable to dissipate heat when the only fan controller is damaged in the design of the conventional single fan controller. Or the inconvenience caused by the need to stop the fan controller server system. Furthermore, each fan of the present invention has independent current monitoring, so that when the fan power supply is short-circuited and overloaded, the power supply to the corresponding fan can be cut off in time, and other fans can continue to work normally to avoid the expansion of the fault. In this case, regardless of whether the fan or the fan controller can be replaced by hot plugging, it is more convenient to use.

Although the present invention has been disclosed in the above embodiments, it is not intended to limit the present invention, and the present invention can be modified and modified without departing from the spirit and scope of the present invention. The scope is subject to the definition of the scope of the patent application attached.

110‧‧‧Compute node

120‧‧‧Substrate

130‧‧‧Fan backplane

140‧‧‧First Fan Controller

150‧‧‧Second fan controller

160‧‧‧Power Module

170‧‧‧Common Serial Input/Output Busbar

180‧‧‧Power cord

200‧‧‧Control unit

201‧‧‧ storage unit

210‧‧‧ Current monitoring unit

1201‧‧‧Multiplexer

1202 and 1203‧‧‧I2C busbars

1301~1306‧‧‧Fan

1801,1802,...,1806‧‧‧Power spur

R1~R6‧‧‧current sampling resistor

Q1~Q6‧‧‧ switch

I_OC1~I_OC6‧‧‧ Current overload signal

FAIL_LED‧‧‧Fan damage indicator

PWM<1..6>‧‧‧Fan control signal

TACH<1..6>‧‧‧Speed signal

Claims (14)

A server system comprising: a fan backplane, a plurality of fans; a server array comprising a plurality of computing nodes; a substrate having a multiplexer, the computing nodes coupled to the multiplexer; and a The first fan controller and the second fan controller are coupled to the computing nodes through the multiplexer, and generate a fan control signal according to the instantaneous temperature of the computing nodes to control the fan speeds, wherein the first fan control The second fan controller and the second fan controller are mutually redundant. The server system of claim 1, wherein each of the first fan controller and the second fan controller further comprises: a control unit; a current monitoring unit coupled to the control unit; The plurality of power sampling lines respectively transmit the power to the fans; and the plurality of current sampling units and the plurality of switches are respectively disposed on the plurality of power supply lines, and the plurality of current sampling units are coupled to the current monitoring unit, The plurality of switches are coupled to the control unit; wherein the control unit is configured to control the opening and closing of the plurality of switches, and the current monitoring unit samples current values flowing through the plurality of current sampling units, when flowing through the plurality of When the current value of one of the current sampling units exceeds a set value, the current monitoring unit sends a current overload signal to the current The control unit controls one of the corresponding plurality of switches to be disconnected to stop the power module from transmitting power to the corresponding fan. The server system of claim 2, wherein the switch is a transistor, wherein the transistor is coupled to the branch through its source terminal and the drain terminal, and the control unit is coupled through the gate terminal. The server system of claim 2, wherein when a fan is damaged, the control unit generates a fan damage indication signal to illuminate a corresponding indicator light. The server system of claim 4, wherein the plurality of fans respectively correspond to a printed circuit board for mounting the corresponding indicator light. The server system of claim 2, wherein the control unit is further coupled to a heat generating unit, and a fan control signal is generated according to an instantaneous temperature of the heat generating unit to control the fan speeds. The server system of claim 6, wherein the control unit stores a fan speed control table, wherein the fan speed control table is used to indicate a correspondence between a temperature value of the heat generating unit and a fan speed, and the control unit is configured according to The instantaneous temperature value of the heat generating unit obtains a set speed value from the fan speed control table, and generates the fan control signal according to the set speed value. The server system of claim 2, wherein the fans further generate a feedback speed signal to the control unit, and the control unit determines the feedback speed of the fan according to the feedback speed signal, when the feedback speed is When the set speed value does not match, it is judged that the fan is faulty. The server system of claim 8, wherein if the control unit controls the switch to turn off the signal after receiving the feedback, the control unit is determined to be faulty. The server system of claim 2, wherein the control unit further performs a self-test procedure, and if it detects that the fan speed control table cannot be read and written normally, it is determined that the control unit is faulty. The server system of claim 2, wherein the fan backplane further includes a first connector coupled to the fan control signal, and a second connector coupled to the plurality of power supply legs. The server system of claim 1, wherein the control unit of the first fan controller is coupled to the second fan controller via a universal serial input/output bus. The server system of claim 12, further comprising: when the second fan controller cannot obtain information of the first fan controller through the universal serial input/output bus, determining the first fan control The device is damaged, and the fan speed is controlled by the second fan controller. The server system of claim 12, further comprising: notifying, by the first fan controller, the second fan controller through the universal serial input/output bus, controlled by the second fan controller The fan speeds.