TWI704463B - Server system and management method thereto - Google Patents

Abstract

The disclosure provides a server system, which comprises the following elements. A plurality of computing nodes and a plurality of storage nodes start to operate after the computing nodes and the storage nodes are actuated. A switch is electrically connected to the computing nodes through a plurality of first ports, and the switch is electrically connected to the storage nodes through a plurality of second ports. A rack management controller is electrically connected to the computing nodes, the storage nodes, and the switch. When the rack management controller receives a hardware requirement, the rack management controller controls the switch to connect to at least a part of the computing nodes and at least a part of the storage nodes according to the hardware requirement.

Description

Server system and management method

The invention relates to a server system and management method, in particular to a server system and management method based on a cabinet management controller.

With the advent of the era of big data, because servers have the characteristics of strong computing power and large storage space, and can provide services to intranet or extranet through the Internet, more and more industries rely on servers to process large amounts of data. .

Generally speaking, the physical characteristics of the computing node and storage node of the server (for example: the temperature, voltage and power supply of each component of the motherboard) are monitored by the baseboard management controller (BMC) and collected The data is transmitted to the rack management controller (rack management controller, RMC). In addition, part of the server architecture can also be directly monitored by the rack management controller through the switch to directly monitor the status of each computing node and storage node to simplify the server architecture and save the cost of maintaining the baseboard management controller. .

However, in the aforementioned server architecture, due to factors such as component configuration and switch specifications, there is often only a single port connection between the switch and the rack management controller and each node. Therefore, when a specific node (for example, a computing node) or port is damaged, the server cannot continue to use the corresponding node (for example, a storage node connected to this computing node), which affects the work being performed.

Therefore, there is still a need for a server system and management method to improve the above problems.

The present invention is to provide a server system and a management method. The switch of the server system can connect a rack management controller and each node with a plurality of connection ports. When a specific port or node is damaged, the rack management controller can control the nodes required for operation through other ports, providing a more effective server system management method to improve the problems mentioned in the prior art.

The present invention provides a server system including the following components. Multiple computing nodes and multiple storage nodes begin to operate after being activated. A switch is electrically connected to the computing nodes through a plurality of first ports, and is electrically connected to the storage nodes through a plurality of second ports. A rack management controller is electrically connected to the computing nodes, the storage nodes, and the switch, and when receiving a hardware demand, controls the switch to connect to at least one of the computing nodes according to the hardware demand Part to at least part of the storage nodes.

The present invention provides a server system management method, which includes: using a rack management controller to activate multiple computing nodes and multiple storage nodes, and when the rack management controller receives a hardware request, use the cabinet to manage The controller controls the switch to connect at least a part of the computing nodes to at least a part of the storage nodes according to the hardware requirements.

The present invention is to provide a server system and a management method. The switch of the server system can connect a rack management controller and each node with a plurality of connection ports. When a specific port or node is damaged, the rack management controller can control the nodes required for operation through other ports. Therefore, the server system described provides a more effective server management method and improves the problems mentioned in the prior art.

The above description of the content of the disclosure and the description of the following embodiments are used to demonstrate and explain the spirit and principle of the present invention, and to provide a further explanation of the patent application scope of the present invention.

The detailed features and advantages of the present invention are described in detail in the following embodiments, and the content is sufficient to enable anyone familiar with the relevant art to understand the technical content of the present invention and implement it accordingly, and according to the content disclosed in this specification, the scope of patent application and the drawings Anyone who is familiar with the relevant art can easily understand the related purpose and advantages of the present invention. The following examples further illustrate the viewpoints of the present invention in detail, but do not limit the scope of the present invention by any viewpoint.

Please refer to FIG. 1, which is a block diagram of a server system according to an embodiment of the present invention. The server system includes a plurality of computing nodes 11, a plurality of storage nodes 12, a switch 13 (switch) and a rack management controller 14 (rack management controller, RMC).

To illustrate the computing node 11 and the storage node 12, please continue to refer to Figure 1. Both the computing node 11 and the storage node 12 start to operate after being activated; the method of being activated can be that the system automatically sends instructions to the nodes, or the user can input instructions to the nodes. After receiving the instruction, the computing node 11 and the storage node 12 begin to perform corresponding operations (for example, the computing node 11 accesses the data of the storage node 12 and performs operations). Specifically, the computing node 11 can be configured as a central processing unit (CPU) or other components with computing functions, and the storage node 12 can be configured as an error-correcting code memory (ECC memory), Registered memory (REG memory) or other components with storage function; the invention is not limited to this.

To illustrate the switch 13, please continue to refer to Figure 1. The switch 13 is connected to the rack management controller 14 and each node through multiple connection ports. Specifically, the switch 13 is electrically connected to the aforementioned computing node 11 through a plurality of first connection ports 15, and electrically connected to the aforementioned storage node 12 through a plurality of second connection ports 16. In addition, the first connection ports 15 and the second connection ports 16 may be hardware specifications supporting inter-integrated circuit bus (I2C bus) ports. However, according to different server configurations, the first ports 15 and the second ports 16 can also be configured as ports whose hardware specifications support other communication buses, and this embodiment is not limited thereto. In this embodiment, the switch 13 can be implemented as a SAS switch chip. In addition, in an implementation aspect of this embodiment, the above-mentioned SAS switch chip model is PM8056. However, the specifications and models of the switch 13 can also be changed according to different server configurations, and the present invention is not limited thereto.

To illustrate the rack management controller 14, please continue to refer to Figure 1. The rack management controller 14 is electrically connected to the computing nodes 11, the storage nodes 12 and the switch 13. In addition, when the server is executing the boot process or receives an external command, the rack management controller 14 will receive a hardware requirement. When the rack management controller 14 receives the aforementioned hardware requirements, the rack management controller 14 controls the switch 13 to connect at least a part of the computing nodes 11 to at least a part of the storage nodes 12. For example, the hardware requirements may include the amount of data calculation and data required for this job. The rack management controller 14 may determine the number of computing nodes 11 required based on the amount of data calculation, and determine the storage that needs to be selected based on the above data. Node 12. When the rack management controller 14 selects the required computing nodes 11 and storage nodes 12, it further controls the switch 13 to connect the selected computing nodes 11 to the selected storage nodes 12, so that these computing nodes 11 can store Take the data of the storage nodes 12 and perform calculations.

In summary, in practice, the computing nodes 11 may include complex programmable logic device (CPLD), real-time clock (RTC), and temperature sensor. , Field replaceable unit (FRU) or other components that can work with the computing node 11. It is worth mentioning that in the server system disclosed in the present invention, the rack management controller 14 can directly collect information (for example, temperature, voltage, and CPLD firmware version, etc.) of each computing node 11 through the switch 13. No additional configuration of baseboard management controller (BMC) is required. Therefore, the server system disclosed in the present invention can not only simplify the structure of the server, but also reduce the cost of maintaining the server.

On the other hand, when the computing node 11 is configured with a complex programmable logic device, the operation status of the computing node 11 is monitored by the rack management controller 14. Generally speaking, in the server architecture of "monitoring each computing node with the baseboard management controller" mentioned in the prior art, the complex programmable logic device is controlled by the baseboard management controller, so the complex programmable logic device The firmware cannot support in-band upgrades under the conventional architecture. However, under the framework of the present invention, the firmware of a complex programmable logic device can not only support out-of-band upgrades, but also support in-band upgrades. In detail, the out-of-band upgrade system can transfer the complex programmable logic firmware to the switch 13 through a serial attached SCSI (serial attached SCSI, SAS) high-speed topology network, and then start the upgrade. On the other hand, the in-band upgrade is performed through the port of the cabinet management controller 14 to transfer the complex programmable logic firmware to the switch 13; the port can support I2C bus connection with hardware specifications Port realization. In addition, in an implementation aspect of this embodiment, the aforementioned serial SCSI is implemented in SAS 3.0. However, according to different transmission rate requirements, the serial SCSI described above can also be implemented in other versions, and this embodiment is not limited to this. It can be seen from the above description that the server system disclosed in the present invention also improves the convenience of firmware upgrades of complex programmable logic devices, allowing users to choose upgrade methods more flexibly.

It should be added that the aforementioned serial SCSI is a computer hub technology, and its main function is to transmit data to computer peripheral parts (such as hard disks, CD-ROMs, etc.). On the other hand, the aforementioned SAS serial SCSI is a specification that supports 2.5-inch hard drives and adopts a direct point-to-point serial transmission method. The SAS 3.0 mentioned in the previous paragraph is the third-generation SAS, and each drive can provide a transmission rate of 12.0 Gbps (12000 Mbps).

Please refer to FIG. 2, which is a flowchart of a server system management method according to an embodiment of the present invention. Please refer to step S0: the rack management controller activates multiple computing nodes and multiple storage nodes. In detail, when the power of the server is turned on, the rack management controller will activate the computing nodes and the storage nodes to make the computing nodes and the storage nodes enter the standby (stand-by) state to For subsequent operations. When the server completes the boot process and generates calculation-related instructions, please refer to step S1: The rack management controller controls the switch to connect at least part of the computing nodes to at least one of the storage nodes according to the hardware requirements. Part. In detail, when the server generates instructions related to operations, the rack management controller will receive the hardware requirements associated with the instructions, and select the computing nodes and computing nodes required for operation from all nodes based on the hardware requirements. Storage node. When the rack management controller completes the selection of computing nodes and storage nodes required for operation, it can further control the switch to connect the selected computing node to the selected storage node to perform related operations.

Please refer to FIG. 3, which is a detailed flowchart of step S1 of the server system management method according to an embodiment of the present invention. As mentioned above, when the rack management controller receives the hardware demand, please refer to step S11: the rack management controller controls the switch to connect one of the computing nodes to one of the storage nodes. Specifically, one of the computing nodes is the computing node selected by the rack management controller, and one of the storage nodes is the storage node selected by the rack management controller. When the rack management controller controls the switch to connect one of the computing nodes to one of the storage nodes, please refer to step S12: the rack management controller determines whether the connected computing node can provide the hardware requirements. The amount of data calculation; the connected computing node is the computing node selected by the cabinet management controller. When the rack management controller determines that the connected computing node can provide the required data calculation amount for the hardware, please refer to step S13: the computing node performs calculations based on the data of the storage node, and the operation of the computing node is controlled by the rack management Monitored by the device.

Continuing from the above, when the rack management controller determines that the connected computing node cannot provide the amount of data calculation required by the hardware requirements, please refer to step S14: the rack management controller controls the switch to connect the other one of these computing nodes. One of these storage nodes. In detail, when the computing node currently selected by the rack management controller cannot load the aforementioned data calculation amount, the rack management controller needs to select the required calculation node from other unselected computing nodes again according to the current data calculation amount , In order to provide sufficient computing performance to load the current data calculation. It can be seen that when the amount of data calculation on the server increases suddenly (for example: special festivals increase online shopping, or online games hold special events to increase network traffic, etc.), the cabinet management controller can calculate based on the data in real time The amount of change, the flexible deployment of computing nodes through the switch. On the other hand, when a computing node or port in use suddenly fails, the rack management controller can also instantly select other operational computing nodes or ports through the switch, so that the current computing work can continue.

In summary, the present invention is to provide a server system and a management method, wherein the switch of the server system can connect the rack management controller and each node with multiple ports. When a specific port or node is damaged, the rack management controller can control the nodes required for operation through other ports. In addition, the rack management controller can adjust the number of operation nodes and storage nodes in real time based on changes in the amount of data calculations. Therefore, the server system described provides a more flexible server management method, and effectively improves the problems mentioned in the prior art.

Although the present invention is disclosed in the foregoing embodiments, it is not intended to limit the present invention. All changes and modifications made without departing from the spirit and scope of the present invention fall within the scope of patent protection of the present invention. For the scope of protection defined by the present invention, please refer to the attached patent scope.

11: compute node 12: storage node 13: switch 14: Cabinet Management Controller 15: The first port 16: second port

FIG. 1 is a block diagram of a server system according to an embodiment of the invention. FIG. 2 is a flowchart of a method for managing a server system according to an embodiment of the invention. FIG. 3 is a detailed flowchart of a management method of a server system according to an embodiment of the present invention.

11: compute node

12: storage node

13: switch

14: Cabinet Management Controller

15: The first port

16: second port

Claims (10)

A server system including: Multiple computing nodes and multiple storage nodes begin to operate after being activated; A switch electrically connected to the computing nodes through a plurality of first ports, and electrically connected to the storage nodes through a plurality of second ports; and A rack management controller is electrically connected to the computing nodes, the storage nodes, and the switch, and when receiving a hardware demand, controls the switch to connect to at least one of the computing nodes according to the hardware demand Part to at least part of the storage nodes. The server system according to claim 1, wherein the rack management controller controls the switch to connect one of the computing nodes to one of the storage nodes according to the hardware requirements, and determines whether Whether the connected computing node can provide the required data calculation amount for the hardware; when the rack management controller determines that the connected computing node can provide the required data calculation amount for the hardware request, the computing node is based on The data of the storage node performs calculations; when the rack management controller determines that the connected computing node cannot provide the required data calculation amount required by the hardware, the rack management controller controls the switch connected to the computing nodes The other to one of the storage nodes. The server system according to claim 1, wherein the hardware specifications of the first ports and the second ports support an integrated circuit bus. The server system according to claim 1, wherein the computing nodes include a complex programmable logic device. The server system according to claim 1, wherein the computing nodes include a real-time clock. The server system according to claim 1, wherein the computing nodes include a temperature sensor. The server system according to claim 1, wherein the computing nodes include a field replaceable unit. A management method for a server system includes: Actuate multiple computing nodes and multiple storage nodes with a rack management controller; and When the rack management controller receives a hardware requirement, the rack management controller controls a switch to connect at least part of the computing nodes to at least part of the storage nodes according to the hardware requirement. The management method according to claim 8, wherein when the rack management controller receives the hardware demand, the rack management controller controls at least a part of the switch connected to the computing nodes according to the hardware demand To at least part of these storage nodes, including: The rack management controller controls the switch to connect one of the computing nodes to one of the storage nodes; Use the rack management controller to determine whether the connected computing node can provide the amount of data calculation required by the hardware requirement; When the rack management controller determines that the connected computing node can provide the amount of data calculation required by the hardware, the computing node performs calculations based on the data of the storage node; and When the rack management controller determines that the connected computing node cannot provide the amount of data required by the hardware requirements, the rack management controller controls the switch to connect another of the computing nodes to the storage nodes one of them. The management method according to claim 8, wherein the switch connects the computing nodes and the storage nodes through an integrated circuit bus.

Images