CN102346707B - Server system and operation method thereof - Google Patents

Abstract

A server system and an operation method thereof, the operation method comprises: under the control of the hardware abstraction layer, a plurality of node management units share a hardware resource; (B) when one of the node management units wants to use the hardware resource, the node management unit sends a command or data to the hardware abstraction layer, and the hardware abstraction layer replaces the node management unit to use the hardware resource; and (C) if an external instruction is received, the hardware abstraction layer recognizes that the external instruction is received by the transmission port of the hardware resource and is transmitted to a corresponding node management unit for execution, and after the external instruction is executed, the corresponding node management unit transmits a message back to the hardware abstraction layer, so that the hardware abstraction layer transmits the message back to an external system manager through the transmission port.

Description

Server system and its method of operation

technical field

The invention relates to a server system and its operating method. the

Background technique

Traditionally, blade servers have been widely used in various applications. Generally speaking, a large number of blade servers are assembled in a chassis system, thereby improving user's operation convenience. The blade server clusters together the core computing circuits of all the computer service systems in the computer service workstation. System administrators are responsible for maintaining and controlling the computer service systems and network configuration inside the computer service workstation. In this way, system administrators can maintain and control multiple computer service systems clustered together. the

At present, the management of the server to the node (node) mainly follows the specification of IPMI (Intelligent Platform Management Interface, intelligent platform management interface), and uses BMC (Baseboard Management Controller, baseboard management controller) to monitor, record and Error recovery and other functions. The so-called node here refers to a computing unit with independent computing capability, which at least includes a CPU (Central Processing Unit) and a memory. For the products currently on the market, a single BMC can only manage a single node, and cannot manage multiple nodes at the same time. In addition, in the known technology, there is a hardware CMM (Chassis Management Module, Chassis Management Module) in the rack system to manage the entire rack system. the

With the development of cloud technology, the demand for data centers is increasing, and how to place more nodes in the limited space of the computer room to increase computing power is the focus of development. the

This application proposes a server system and its operation method, which can effectively reduce the number of BMC chips, so as to increase the board space in the server, facilitate the placement of more nodes to improve computing capabilities, and reduce server costs. the

Contents of the invention

The invention relates to a server system and an operation method thereof, which enables multiple node management units (which are software, respectively used to manage a node) of the BMC to share the hardware resources of the BMC through a hardware abstraction layer.

According to an embodiment of the present invention, a server system is proposed, including: at least one system board, the system board includes a baseboard management controller and a plurality of nodes, the baseboard management controller includes a plurality of node management units, a hardware abstraction Layer and a hardware resource, these node management units respectively manage these nodes, under the control of the hardware abstraction layer, these node management units share the hardware resource; a connection port, used to connect to an external system manager; and an internal channel, connected to the system board and the connection port. the

According to another embodiment of the present invention, an operating method of a server system is proposed, the server system includes at least one system board, the system board includes a baseboard management controller and a plurality of nodes, the baseboard management controller includes a plurality of nodes A management unit, a hardware abstraction layer and a hardware resource, these node management units respectively manage these nodes. The method includes: (A) under the control of the hardware abstraction layer, these node management units share the hardware resource; (B) when one of the node management units intends to use the hardware resource, the node management unit sending an instruction or a data to the hardware abstraction layer, and the hardware abstraction layer replaces the node management unit to use the hardware resource; and (C) if an external instruction is received, the hardware abstraction layer identifies that the external instruction is Received by the transmission port of the hardware resource to be sent to a corresponding node management unit for execution, and when the external command is executed, the corresponding node management unit sends a message back to the hardware abstraction layer to The HAL sends the information back to an external system manager through the transport port. the

In order to have a better understanding of the above-mentioned and other aspects of the present invention, the preferred embodiments are specifically cited below, and in conjunction with the accompanying drawings, the detailed description is as follows:

Description of drawings

FIG. 1 shows a schematic diagram of a rack system according to an embodiment of the present invention. the

Figure 2 shows a schematic diagram of a BMC according to an embodiment of the present invention. the

FIG. 3 shows a schematic diagram of multiple NMUs sharing the hardware part of the BMC through the HAL. the

FIGS. 4A-4C show schematic diagrams of forwarding instructions/information through the HAL according to an embodiment of the present invention. the

【Description of main component symbols】

100: rack system 101: connection port

102: Area network 103: I ² C bus

110～130: System board 111, 121, 131: BMC

112-1～112-Y, 122-1～122-Y, 132-1～132-Y: nodes

211: HAL 212-1～212-Y: Node management unit

221: GPIO pin 222: Storage unit

223: Serial port 224: Sensing unit

225: System interface 226: LAN interface

227: I ² C interface 410: System manager

421~466: Steps

Detailed ways

In the embodiment of the present invention, a single BMC can manage multiple nodes. In the embodiment of the present invention, the BMC is expanded from single-node management to multi-node management through HAL (Hardware Abstraction Layer, hardware abstraction layer), and is still fully compatible with the IPMI specification. In this way, the number of BMC chips in the rack system can be effectively reduced, which can not only reduce the cost, but also save space, and can reduce the internal ambient temperature of the rack system. the

FIG. 1 shows a schematic diagram of a rack system according to an embodiment of the present invention. As shown in FIG. 1 , the rack system 100 according to the embodiment of the present invention at least includes: a connection port 101, a LAN (Local Area Network, local area network) 102, an I ² C (Inter-Integrated Circuit, internal integrated circuit) bus 103, and multiple system board. Although the rack system 100 including three system boards 110 - 130 is taken as an example in FIG. 1 , it is known that the embodiments of the present invention are not limited thereto. The system board 110 includes: BMC 111 and nodes 112-1~112-Y; the system board 120 includes: BMC 121 and nodes 122-1~122-Y. The system board 130 includes: a BMC 131 and nodes 132-1˜132-Y. Here, Y is a positive integer.

Commands and signals issued by the system administrator can be transmitted to corresponding system boards through the connection port 101 . Of course, the messages sent by the system board can be sent back to the system administrator through the connection port 101 . the

As shown in FIG. 1 , LAN 102 and I ² C bus 103 provide communication paths between BMCs of these system boards. In addition, in other embodiments of the present invention, the BMC may also optionally have a CMM function.

Figure 2 shows a schematic diagram of a BMC according to an embodiment of the present invention. As shown in Figure 2, BMC includes a hardware part and a software part. The software part of the BMC includes: HAL 211 and node management units (NMU, Node Management Unit) 212-1~212-Y. The hardware part of BMC includes: GPIO (General Purpose Input/Output, general purpose input/output) pin 221, storage unit 222, serial port 223, sensing unit 224, system interface (System Interface, referred to as SI) 225, LAN interface 226 and I ² C interface 227 .

For each node, the BMC will read the readings of the sensing unit 224 to monitor the physical parameters of the node (such as CPU temperature, memory temperature, voltage, etc.). For example, the BMC may have three CPU temperature sensors, respectively sensing the temperature of the internal CPUs of the three nodes it manages. Moreover, the BMC controls the power on and off of the system through the GPIO pin 221 . In addition, the system administrator can send IPMI commands to the BMC through interfaces such as the LAN interface 226 or the system interface 225, so as to require the BMC to execute the IPMI commands. the

NMU is a management software that implements the IPMI specification. That is, in terms of BMC 111, NMU1-NMU3 can be used to manage nodes 112-1-112-3 respectively. In the embodiment of the present invention, since a single BMC is used to manage the relationship of multiple nodes, multiple NMUs must share the hardware part of the BMC, so the hardware abstraction layer (HAL) 211 can be used to solve this problem. HAL 211 will create a set of logical (virtual) hardware devices for each NMU, and make a corresponding relationship with the physical hardware devices. the

FIG. 3 shows a schematic diagram of multiple NMUs sharing the hardware part of the BMC through the HAL. As shown in FIG. 3, when the NMU wants to access the SDR (Sensor Data Record, sensing data record), the NMU does not need to know the actual access address of the node's SDR in the storage unit 222. When the NMU wants to read the SDR data, the NMU only needs to tell the HAL 211 to read the SDR data of its corresponding node (such as CPU temperature, memory temperature, applied voltage, etc.), and the HAL 211 will send the NMU The SDR data of the corresponding node is sent back to the NMU. SDR1～SDR3 respectively represent the SDR data of nodes 1～3, which correspond to NMU 1～NMU 3 respectively. the

Similarly, when the NMU intends to store SDR data, the NMU does not need to know the actual storage address of the node's SDR in the storage unit 222 . When the NMU wants to store SDR data, the NMU only needs to transmit the SDR data to be stored to the HAL 211, and the HAL 211 will store the SDR data in the storage unit 222. That is to say, the HAL 211 will perform mapping to map the data to be stored/retrieved by the NMU to the storage unit 222 . the

SEL is a system event log (System Event Log), which is used to store system events of nodes (such as system exceptions, etc.). Similarly, when the NMU 1-NMU 3 intend to access the SEL 1-SEL 3, the HAL 211 is also responsible for storing/retrieving the storage unit 222, as mentioned above. FRU is a field replaceable unit (Field Replaceable Unit), which records system information such as the serial number and product name of the system board. Similarly, when the NMU 1-NMU 3 want to access the FRU 1-FRU 3, the HAL 211 is also responsible for accessing the storage unit 222, as mentioned above. What's more, the functions that HAL 211 can be responsible for data correspondence are not limited to SDR, SEL and FRU. Other functions mentioned in the IPMI specification, such as network connection sequence (SOL, Serial Over LAN), platform event filter (PEF, Platform Event Filter), sensor monitor (Sensor Monitor), rack control (Chassis Control), etc., NMU The function of correspondence or forwarding can be achieved through HAL. the

FIGS. 4A-4C show schematic diagrams of forwarding instructions/information through the HAL according to an embodiment of the present invention. As shown in FIG. 4A, the communication between the system manager 410 and the HAL 211 is bidirectional, and the communication between the HAL 211 and the NMU is also bidirectional. the

FIG. 4B shows a schematic diagram of the system manager 410 sending IPMI commands to the BMC through the HAL 211. As shown in FIG. 4B, the system manager 410 will send the IPMI command to the HAL 211. Then, the HAL 211 judges whether the IPMI command is transmitted via the system interface (SI) (as shown in step 421) or via the LAN interface (LAN) (as shown in step 422). If the IPMI command is transmitted via SI, HAL 211 then judges that the IPMI is the first transmission port SI 1 (corresponding to node 1) and the second transmission port SI 2 (corresponding to node 2) of the system interface. ) or the third transmission port S1 3 (which corresponds to node 3), as shown in steps 431-433. That is, in this embodiment, the system interface of the BMC has multiple SI transmission ports, among which there are 3 SI transmission ports for connecting the BMC to the system manager 410 . If the IPMI command is transmitted via the LAN interface, then HAL 211 judges that the IPMI is composed of the first transmission port LAN 1 (corresponding to node 1) and the second transmission port LAN 2 (corresponding to node 1) of the LAN interface. 2) or the third transmission port LAN 3 (which corresponds to node 3), as shown in steps 434-436. That is, in this embodiment, the LAN interface of the BMC has a plurality of LAN transmission ports, among which there are 3 LAN transmission ports for connecting the BMC to the system manager 410 . After HAL 211 judges through step 431～436, HAL can judge that this IPMI command sent by system administrator 410 is to give that one of NMU1～NMU 3, then, HAL 211 sends this IPMI command to destination NMU. the

FIG. 4C shows a schematic diagram of the BMC returning information to the system manager 410 through the HAL 211. When the NMU receives the IPMI command from the system manager 410, the NMU will perform corresponding operations, and then, the NMU will return the response information to the system manager 410 through the HAL 211. As shown in Figure 4C, the NMU will send a response message to the HAL 211. Then, the HAL 211 determines whether the response message is received via the system interface (SI) (as in step 441) or via the LAN interface (as in step 442). If the response information is received via the system interface, the HAL 211 analyzes the received response information, and the HAL 211 can determine which NMU the response information is sent from (steps 451-453 and steps 454-456). That is, in the present embodiment, the system interface of BMC has a plurality of SI transmission ports, wherein 3 SI transmission ports are used to make the system manager 410 connect to BMC; and the LAN interface of BMC has a plurality of LAN transmission ports, There are 3 LAN transmission ports for connecting the system manager 410 to the BMC. HAL 211 will judge whether the NMU transmits the response information via the system interface, and then determine which NMU the response information is sent from (steps 451 to 453). In this way, HAL 211 can send the response information through the original receiving interface (such as SI ) is sent back to the system manager 410 (steps 461-463). Similarly, the HAL 211 will judge whether the NMU transmits the response information via the LAN interface, and then, the HAL 211 judges whether the response information is sent by which NMU (steps 454-456), and then the response information can be sent via the original receiving interface ( LAN interface) back to the system administrator 410 (steps 464-466). the

That is to say, in the embodiment of the present invention, when the system manager 410 transmits the IPMI command to the BMC through the LAN interface or the system interface, the HAL 211 will identify which transmission port the IPMI command is received by and send the command to the corresponding The corresponding NMU to execute. When the NMU executes the command, the NMU will return the information to the HAL 211, and the HAL 211 will return the response information to the system manager 410 through the original transmission port. Certainly, the embodiment of the present invention is not limited to the fact that the HAL 211 can only transmit the IPMI command via the LAN interface or the system interface, and the HAL 211 can also transmit the IPMI command via the interface supported in the IPMI specification. the

In summary, the embodiments of the present invention at least have the following advantages: (1) the embodiments of the present invention can reduce the number of BMC chips needed in high-density servers (such as blade servers), so as to reduce costs; and (2) the present invention The embodiment can effectively use space, increase the number of nodes and computing power of the server, and effectively reduce the temperature of the system (because the number of BMC chips is reduced). the

In summary, although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention. Those skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention should be defined by the appended claims. the

Claims (8)

1. A server system comprising: At least one system board, the system board includes a baseboard management controller and a plurality of nodes, the baseboard management controller includes a plurality of node management units, a hardware abstraction layer and a hardware resource, these node management units respectively manage these nodes, in Under the control of the hardware abstraction layer, these node management units share the hardware resources; a connection port for connecting to an external system manager; and an internal channel connected to the system board and the connection port, Wherein, the system board further includes a plurality of transmission ports, and these transmission ports are used to connect the baseboard management controller to the external system manager; If an external command is transmitted to the baseboard management controller through the hardware resource, the hardware abstraction layer identifies which transmission port the external command is received from, so as to transmit the external command to a corresponding node management unit for execution; and After the corresponding node management unit executes the external command, the corresponding node management unit returns a message to the hardware abstraction layer, so as to send the message back to the external system manager through the transmission port. 2. The server system according to claim 1, wherein the hardware abstraction layer establishes a logical hardware device for each node management unit to correspond to the hardware resource. 3. The server system according to claim 2, wherein, when one of the node management units wants to use the hardware resource, the node management unit sends an instruction to the hardware abstraction layer, and the hardware abstraction layer The hardware resource is accessed according to the command and a result is returned to the node management unit. 4. The server system as claimed in claim 2, wherein, when one of the node management units intends to use the hardware resource, the node management unit transmits a data to the hardware abstraction layer, and the hardware abstraction layer The hardware resource is accessed according to the data. 5. A method for operating a server system, the server system comprising at least one system board, the system board comprising a baseboard management controller and a plurality of nodes, the baseboard management controller comprising a plurality of node management units, a hardware abstraction layer and A hardware resource, these node management units respectively manage these nodes, and the operation method includes: (A) Under the control of the hardware abstraction layer, these node management units share the hardware resources; (B) When one of the node management units intends to use the hardware resource, the node management unit sends an instruction or a data to the hardware abstraction layer, and the hardware abstraction layer replaces the node management unit to use the hardware resource; and (C) If an external command is received, the hardware abstraction layer identifies which transmission port of the hardware resource the external command is received to transmit to a corresponding node management unit for execution, and when the external command is executed Afterwards, the corresponding node management unit returns an information to the hardware abstraction layer, so that the hardware abstraction layer sends the information back to an external system manager through the transmission port. 6. The operating method as claimed in claim 5, wherein, the step (A) comprises: The hardware abstraction layer establishes a logical hardware device for each node management unit to correspond to the hardware resource. 7. The operation method as claimed in claim 6, wherein, the step (B) comprises: When one of the node management units intends to use the hardware resource, the node management unit sends the instruction to the hardware abstraction layer, and the hardware abstraction layer accesses the hardware resource according to the instruction and returns a result Passed to the node management unit. 8. The operating method as claimed in claim 6, wherein, the step (B) comprises: When one of the node management units wants to use the hardware resource, the node management unit sends the data to the hardware abstraction layer, and the hardware abstraction layer accesses the hardware resource according to the data.