CN111858187A - An electronic device and service switching method and device - Google Patents

Abstract

Embodiments of the present invention provide an electronic device and a service switching method and device. The electronic device includes two controllers, one of which is a master controller and the other a backup controller. In the event of a fault, the primary controller processes services, and the standby controller detects whether the primary controller is faulty. If the primary controller fails, the secondary controller switches to the primary controller to process services, and opens its own intranet during service processing. The network port communicates with the multiple extended processing chips; it can be seen that even if one controller fails, another controller can take over the services of the faulty controller, and can communicate with multiple extended processing chips, and then the multiple extended processing chips Data processing is still possible, reducing business interruptions.

Description

An electronic device and service switching method and device

technical field

The present invention relates to the technical field of hardware devices, and in particular, to an electronic device and a service switching method and device.

Background technique

At present, in order to improve the data processing capability of an electronic device, it is usually necessary to expand multiple processing chips in the device, so that the controller in the device can send the service data to be processed to each processing chip for processing. For example, expand multiple GPUs (Graphics Processing Unit, graphics processing unit) in the intelligent analysis device, the motherboard controller in the intelligent analysis device obtains the image data to be processed, and sends the image data to the expanded multiple GPUs. The multiple GPUs are used to analyze and process the image data, and feed back the processing results to the mainboard controller.

In the existing equipment, multiple expansion processing chips and a controller are usually installed. If the controller fails, the controller cannot send service data to the multiple expansion processing chips. It is impossible to analyze and process business data, resulting in business interruption.

SUMMARY OF THE INVENTION

The purpose of the embodiments of the present invention is to provide an electronic device and a service switching method and apparatus, so as to reduce the situation of service interruption.

In order to achieve the above purpose, an embodiment of the present invention provides an electronic device, including: a main controller, a backup controller, and a plurality of extended processing chips; each controller includes an intranet network port, and each controller uses its own The intranet network port is connected with the plurality of expansion processing chips;

The main controller communicates with the plurality of extended processing chips through the intranet network port of the main controller during business processing;

The standby controller detects whether the main controller is faulty; if faulty, the standby controller switches to the main controller to process services; the switched main controller opens its own intranet network port in the process of processing services; The internal network port of the main controller communicates with the plurality of expansion processing chips.

Optionally, the device further includes a first network switch chip and a backplane;

The first network switch chip is arranged on the backplane; the internal network port of each controller is connected to the first network switch chip; the first network switch chip passes through the backplane. The first type of interface is connected with the plurality of expansion processing chips.

Optionally, the plurality of expansion processing chips are arranged on a tray, a second network switching chip is further arranged on the tray, the tray is inserted into the first type interface on the backplane, and the plurality of expansion The processing chip is connected with the second network switch chip, and the second network switch chip is connected with the first type interface on the backplane.

Optionally, the multiple extended processing chips are arranged on multiple trays, each tray is provided with a second network switch chip and one or more extended processing chips, and each tray is respectively inserted into one of the backplanes. In the first type of interface, each expansion processing chip is respectively connected with the second network switch chip on its own tray, and each second network switch chip is respectively connected with the first type of interface inserted in its own tray.

Optionally, the device further includes a hard disk, and the hard disk is connected to the second-type interface on the backplane; each controller includes an adapter, and each controller is connected to the backplane through its own adapter. The second type of interface is connected.

Optionally, the main controller writes the data processing status parameters of the multiple expansion processing chips into a designated area of the hard disk; if the main controller fails, the standby controller switches to the main controller. After processing the service, the server accesses the designated area, obtains the processing state parameter, and processes the service based on the processing state parameter.

Optionally, each controller includes a heartbeat network port, and whether the peer controller is faulty is detected between the main controller and the standby controller through the heartbeat network port.

Optionally, after the standby controller switches to the main controller to process services and opens its own intranet network port, it controls the plurality of extended processing chips to reset.

Optionally, the extended processing chip is a GPU, and both the first type interface and the second type interface are SFF-8639 interfaces.

In order to achieve the above object, an embodiment of the present invention further provides a service switching method, which is applied to a first controller in an electronic device, and the device further includes a second controller and a plurality of extended processing chips, each of which is Including an intranet network port, each controller is connected to the plurality of expansion processing chips through its own intranet network port; the method includes:

If the first controller is the main controller, during the business process of the first controller, communicate with the plurality of expansion processing chips through the intranet network port of the first controller;

If the first controller is the standby controller, it is detected whether the second controller, which is the main controller, is faulty; if the first controller is faulty, the first controller switches to the main controller to process services; in the process of processing services, it is turned on The intranet network port of the first controller; communicates with the plurality of expansion processing chips through the intranet network port of the first controller.

Optionally, the communicating with the multiple expansion processing chips through the intranet network port of the first controller includes:

Send the service data to be processed to the plurality of extended processing chips through the intranet network port of the first controller;

The data processing results sent by the multiple expansion processing chips are received through the intranet network port of the first controller.

Optionally, after communicating with the multiple expansion processing chips through the intranet network port of the first controller, the method further includes:

writing the data processing state parameters of the plurality of extended processing chips into a preset storage area;

In the case where the failure of the second controller serving as the main controller is detected, it also includes:

By accessing the preset storage area, the processing state parameter is acquired, and the service is processed based on the processing state parameter.

Optionally, in the case where the failure of the second controller serving as the main controller is detected, and after the opening of the intranet network port of the first controller, the method further includes:

One or more extended processing chips in the plurality of extended processing chips are controlled to be reset.

In order to achieve the above object, an embodiment of the present invention also provides a service switching device, which is applied to a first controller in an electronic device, and the device further includes a second controller and a plurality of extended processing chips, each of which is Including an intranet network port, each controller is connected to the plurality of expansion processing chips through its own intranet network port; the device includes:

The first processing module is configured to, when the first controller is the main controller, communicate with the multiplexer through the internal network port of the first controller during the business process of the first controller. an expansion processing chip for communication;

The second processing module is configured to detect whether the second controller serving as the main controller fails when the first controller is the standby controller; if the first controller fails, switch the first controller to the main controller Processing services; in the process of processing services, opening the intranet network port of the first controller; and communicating with the plurality of extended processing chips through the intranet network port of the first controller.

Optionally, the first processing module is also used for:

In the case that the first controller is the main controller, the service data to be processed is sent to the plurality of extended processing chips through the intranet network port of the first controller; through the first controller The internal network port of the device receives the data processing results sent by the multiple expansion processing chips;

The second processing module is also used for:

In the case that the first controller is the standby controller and the failure of the second controller, which is the main controller, is detected, the service data to be processed is sent to the intranet port of the first controller to the multiple extended processing chips; receive the data processing results sent by the multiple extended processing chips through the intranet network port of the first controller.

Optionally, the device further includes:

The writing module is configured to write the data processing state parameters of the multiple expansion processing chips into the pre-processor after the communication with the multiple expansion processing chips through the internal network port of the first controller. set storage area;

An obtaining module, configured to obtain the processing state parameter by accessing the preset storage area when the detection module detects a failure of the main controller, and process services based on the processing state parameter.

Optionally, the device further includes:

The reset module is used to control the multiple expansion processing chips in the case of detecting the failure of the second controller as the main controller and after the opening of the internal network port of the first controller. One or more extended processing chips are reset.

The electronic device provided by the embodiment of the present invention includes two controllers, wherein one controller is the main controller and the other controller is the standby controller. In the case that the two controllers are not faulty, the main controller processes the business , the standby controller detects whether the main controller is faulty. If the main controller fails, the standby controller switches to the main controller to process services. In the process of processing services, it opens its own intranet network port to communicate with the multiple expansion processing chips; It can be seen that, in the first aspect, even if one controller fails, another controller can take over the business of the faulty controller, and can communicate with multiple extended processing chips, and then the multiple extended processing chips can still perform data processing, reducing the need for business interruption.

In the second aspect, some related active-standby switching solutions usually include: a master node, a backup node, and a computing node. The computing node is the node that analyzes and processes business data. The computing node manages, and the master node communicates with the computing node. If the master node fails, the standby node replaces the master node to manage the computing node, and the standby node communicates with the computing node. The master node and the backup node perform the master-slave switchover through the software communication mechanism. In other words, the communication link between the master node and the computing node is connected in hardware, and the communication link between the backup node and the computing node is in the The hardware is also connected, and the master node and the standby node control whether they communicate with the computing node through their own software communication mechanism (their own business logic). In this solution, the master node and the backup node need to use different addresses to communicate with the computing node, otherwise there will be address conflict. In this way, when the master and backup switch is performed, the computing node needs to switch the destination address to communicate normally, which makes the computing node logic is more complicated.

In this solution, the controller controls the communication with the extended processing chip by opening its own internal network port on the hardware. In other words, only one communication link between the controller and the extended processing chip is connected on the hardware. , when one controller fails, the other controller connects the hardware communication link between itself and the expansion processing chip, so that the two controllers use the same address to communicate with the expansion processing chip, and there will be no address conflict. Therefore, the extended processing chip does not need to switch the destination address, which simplifies the logic of the extended processing chip, and realizes the active-standby switching without the extended processing chip sensing.

In the third aspect, in some of the above active-standby switching solutions, in order to reduce the address conflict between the master node and the standby node, the means used are: the computing node switches the destination address; In the case of address conflict between the node and the standby node, the method used is: when the master node is not faulty, the master node and the standby node use different addresses. If the master node fails, the standby node changes its own address to the address of the master node. , which is also called address drift, that is, the address of the master node drifts to the standby node; in this solution, the computing node does not need to switch the destination address, but the standby node needs to switch its own address, which makes the logic of the standby node more complicated.

In this solution, the controller controls the communication with the extended processing chip by opening its own internal network port on the hardware. In other words, only one communication link between the controller and the extended processing chip is connected on the hardware. , when one controller fails, the other controller connects the hardware communication link between itself and the expansion processing chip, so that the two controllers use the same address to communicate with the expansion processing chip, and there will be no address conflict. Therefore, the standby controller does not need to switch the destination address, which simplifies the logic of the standby controller.

Of course, it is not necessary for any product or method of the present invention to achieve all of the advantages described above at the same time.

Description of drawings

In order to explain the embodiments of the present invention or the technical solutions in the prior art more clearly, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained according to these drawings without creative efforts.

FIG. 1 is a first structural schematic diagram of an electronic device provided by an embodiment of the present invention;

2 is a schematic structural diagram of a controller according to an embodiment of the present invention;

3 is a schematic diagram of a second structure of an electronic device provided by an embodiment of the present invention;

4 is a third schematic structural diagram of an electronic device provided by an embodiment of the present invention;

5 is a schematic diagram of a fourth structure of an electronic device provided by an embodiment of the present invention;

6 is a fifth structural schematic diagram of an electronic device provided by an embodiment of the present invention;

FIG. 7 is a schematic flowchart of a service switching method according to an embodiment of the present invention;

FIG. 8 is a schematic structural diagram of a service switching apparatus according to an embodiment of the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

In order to solve the above technical problems, the embodiments of the present invention provide an electronic device and a service switching method and apparatus. The electronic device provided by the embodiments of the present invention is first introduced in detail below.

1 is a schematic diagram of a first structure of an electronic device provided by an embodiment of the present invention, including: a main controller 100, a backup controller 200, and a plurality of extended processing chips 300 (extended processing chip 1, extended processing chip 2 . . . extended processing chips Chip N, N is a positive integer greater than 1), each controller includes an intranet network port, and each controller is connected to the plurality of extended processing chips through its own intranet network port. For the convenience of description, the intranet network port of the main controller is denoted as 110, and the intranet network port of the standby controller is denoted as 210.

The main controller 100 communicates with the plurality of extended processing chips 300 through the intranet network port 110 of the main controller in the process of processing services; the standby controller 200 detects whether the main controller 100 is faulty, and if it fails, the standby controller 200 Switch to the main controller to process services; the switched main controller (original/standby controller) opens its own intranet network port 210 in the process of processing services, and communicates with the plurality of expansion processing chips 300 through the intranet network port 210 communication.

Generally speaking, the controller obtains the service data to be processed, and sends the service data to a plurality of extended processing chips. Each extended processing chip processes the service data received by itself, and feeds back the processing result to the controller. In the embodiment of the present invention, the communication process between the controller and the extended processing chip may include: the controller sends service data to the extended processing chip, and the extended processing chip sends data processing results to the controller; in addition, the controller may also control the extended processing chip. The controller can also monitor the data processing status of the extended processing chip during reset, startup, and so on, which is not specifically limited.

The electronic device provided in this embodiment includes two controllers: a main controller and a standby controller. In one case, the main controller and the standby controller may be the same component in hardware. For example, as shown in FIG. 2 , the main controller 100 may include a mainboard and a controller circuit disposed on the mainboard; the controller circuit includes: a CPU (Central Processing Unit, central processing unit), a memory, an HBA (Host Bus Adapter, host bus) adapter), the external network card that provides the external network port, the internal network card that provides the internal network port 110 and the heartbeat network card that provides the heartbeat network port; the controller 200 can include a mainboard and a controller circuit that is arranged on the mainboard; The controller circuit includes: CPU, memory, HBA, external network card providing external network port, internal network card providing internal network port 210 and heartbeat network card providing heartbeat network port. Among them, the internal network port is used for internal communication of electronic equipment. For example, the controller communicates with the expansion processing chip through the internal network port; the external network port is used for communication between electronic equipment and external equipment. The network port obtains the business data to be processed; the HBA can be connected to the hard disk and exchange data with the hard disk; the heartbeat network ports of the two controllers are connected, and the two controllers can detect whether the other party is faulty through the heartbeat network port.

In this embodiment, one controller in the electronic device is used to process services, and the other controller is used as a backup. The controller that processes the service is called the main controller, and the controller that serves as the backup is the backup controller, and the two can perform active-standby switchover. During the business process of the main controller 100, the main controller 100 communicates with the multiple expansion processing chips 300. In this case, the intranet port 210 of the standby controller is closed, and the standby controller 200 cannot communicate with the multiple expansion processing chips 300. The expansion processing chip 300 communicates; if the primary controller fails, the standby controller 200 switches to the primary controller to process services; in the process of processing services, the internal network port 210 of its own is activated to communicate with the multiple expansion processing chips 300 . In other words, if the primary controller fails, the backup controller 200 takes over the primary controller 100 to process services.

For example, when the electronic device starts up, the two controllers can negotiate which controller becomes the master controller through the heartbeat network port. For example, when an electronic device starts up, the two controllers usually have a difference in the speed of startup completion after power-on. The controller that starts up quickly after power-on (or the controller that completes startup first) can be used as the main controller. The other controller acts as the standby controller. Specifically, after the controller is powered on and started, the heartbeat network port can be used to detect whether the peer controller is powered on and started. If the power-on and start-up is not completed, the controller that has been powered on and started becomes the master controller.

Each controller is configured with two kinds of business logic, one is the business logic of the primary controller, and the other is the business logic of the standby controller. For each controller, the controller determines whether it is the main controller or the standby controller; if it does not exist, it determines that it is the main controller and enters the business logic of the main controller; if it exists, it determines that it is the standby controller. controller, enter the business logic of the standby controller. The business logic of the main controller is to process the business, and in the process of processing the business, it opens its own intranet network port, and communicates with the multiple expansion processing chips through its own intranet network port; the business logic of the standby controller is: Detect whether the main controller is faulty, and if it fails, switch to the main controller to process services; in the process of processing services, open its own internal network port, and communicate with the multiple expansion processing chips through its own internal network port.

In one embodiment, as shown in FIG. 1 and FIG. 2 , each controller includes a heartbeat network port, and whether the peer controller is faulty is detected between the main controller 100 and the standby controller 200 through the heartbeat network port.

The extended processing chip is different from the chip inside the controller, or a processing chip extended outside the controller. For example, the extended processing chip in this embodiment may be a GPU, a CPU, an APU (Accelerated Processing Unit, accelerated processor), or a TPU (Tensor Processing Unit, tensor processing unit), etc., which is not specifically limited. The number of expansion processing chips can be set according to the actual situation, such as 3-9, 10-20, etc. The specific number is not limited.

As shown in FIG. 1 and FIG. 2 , each controller may include an intranet port, and the controller may open or close its own intranet port.

In one case, the main controller and the standby controller can use the same IP (Internet Protocol, Internet Protocol) address; for example, in the process of the main controller working normally, the main controller The chip communicates; if the main controller fails, the standby controller detects the failure of the main controller through the heartbeat network port, and the standby controller switches to the main controller and starts its own network port. server) still use this IP address to communicate with the expansion processing chip. In this way, for the expansion processing chip, the IP address of the communication object does not change, and the expansion processing chip does not need to perform the operation of switching the IP address of the communication object.

In one embodiment, after the standby controller 200 switches to the main controller to process services and opens its own intranet network port 210, it controls the plurality of expansion processing chips 300 to reset.

For example, suppose there are three extended processing chips: extended processing chip 1, extended processing chip 2, and extended processing chip 3; assuming that the main controller obtains 15 images to be processed, the main controller will , 10, 13 images are sent to extended processing chip 1, 2, 5, 8, 11, 14 images are sent to extended processing chip 2, 3, 6, 9, 12, 15 images are sent to extended processing Chip 3; Suppose the main controller fails, at this time the extended processing chip 1 is processing the seventh image, the extended processing chip 2 is processing the eighth image, and the extended processing chip 3 is processing the ninth image After the standby controller 200 detects the failure of the main controller 100 through the heartbeat network port, the standby controller 200 switches to the main controller to process business and opens its own intranet network port 210, and then controls these three expansion processing chips to reset; After reset, the extended processing chip 1 reprocesses the seventh image, the extended processing chip 2 reprocesses the eighth image, and the extended processing chip 3 reprocesses the ninth image.

In the above example, if the controller is switched in the process of processing an image by the extended processing chip, it is likely to interfere with the processing of the image by the extended processing chip, resulting in an error in the processing result of the image; in this embodiment, , after switching the controller, the extended processing chip is reset, and the image is processed again, which reduces the error of the processing result and improves the accuracy of the processing result.

In Figure 2, each controller includes an independent mainboard, in this case, the mainboards of the two controllers can be connected through a backplane. Alternatively, in other embodiments, the two controller circuits may also be provided on the same motherboard, which is not specifically limited.

As an embodiment, referring to FIG. 3 , the electronic device may further include a first network switch chip 400 and a backplane 500 ; the first network switch chip 400 is disposed on the backplane 500 ; the internal network ports of each controller are It is connected with the first network switch chip 400 ; the first network switch chip 400 is connected with the plurality of expansion processing chips 300 through the first type interface on the backplane 500 .

There can be multiple interfaces on the backplane. For example, the interface can be an SFF-8639 interface, which can be directly connected to the hard disk, or can be connected to the expansion processing chip through a network switch chip. In the embodiment, the interface connected to the network switch chip in the backplane is referred to as the first type of interface.

In this embodiment, the controller is connected to the first network switch chip; the first network switch chip is arranged on the backplane, so the first network switch chip can perform data interaction with the first type interface on the backplane; The first type of interface of the controller is connected with the extended processing chip; in this way, the controller can start its own intranet network port, and can communicate with the extended processing chip.

In one case, the backplane may have an independent power supply function, or in other words, the backplane may be connected to an independent power supply to supply power to two controllers, an expansion processing chip, a first network switch chip, and other devices. In this way, if the main controller fails, the power supply function of the backplane is not affected, and devices such as the expansion processing chip, the backup controller, and the first network switch chip can continue to work.

Still referring to FIG. 3 , in this embodiment, it can be understood that two controllers become an integral controller through the backplane, and the first type of interface in the backplane can be understood as an interface provided by the integral controller.

As an embodiment, the plurality of expansion processing chips 300 included in the electronic device are arranged on the tray 600, the second network switching chip 700 is further arranged on the tray 600, and the tray 600 is inserted into the first type interface on the backplane 500, The plurality of expansion processing chips 300 are connected to the second network switch chip 700 , and the second network switch chip 700 is connected to the first-type interface on the backplane 500 .

To distinguish the descriptions, the network switch chip on the backplane is called the first network switch chip, and the network switch chip on the tray is called the second network switch chip. The number of trays can be one or more, and the specific number is not limited.

In this embodiment, the controller is connected to the first network switch chip; the first network switch chip is arranged on the backplane, so the first network switch chip can perform data interaction with the first type interface on the backplane; The first type of interface is connected with the second network switch chip in the tray; the second network switch chip in the tray is connected with the expansion processing chip; in this way, the controller starts its own internal network port and can communicate with the expansion processing chip. communication is possible.

In one embodiment, the plurality of extended processing chips 300 are disposed on a plurality of trays 600 , each of the trays 600 is provided with a second network switch chip 700 and one or more extended processing chips 300 , and each tray 600 is inserted into each tray 600 respectively. In a first-type interface on the backplane 500, each expansion processing chip 300 is respectively connected with the second network switch chip 700 on the tray 600 where it is located, and each second network switch chip 700 is respectively inserted into the tray 600 where it is located. connected to the first-class interface.

Referring to FIG. 4 , it is assumed that the electronic device includes multiple trays 600 , the expansion processing chip is a GPU, each tray is provided with two GPUs and a second network switch chip, and each tray is inserted into a first-type interface.

This embodiment can be applied to a scenario where there are many extended processing chips. In other words, if there are many extended processing chips, these extended processing chips can be divided into multiple groups, and each group of extended processing chips is arranged in the same tray. The tray is also provided with a second network switch chip, and the expansion processing chips in the tray are all connected to the first type interface on the backplane through the second network switch chip.

In one embodiment, the electronic device may further include a hard disk 800, and the hard disk 800 is connected to the second type interface on the backplane 500; each controller includes an adapter, and each controller is connected to the backplane 500 through its own adapter. The second type of interface is connected. For the convenience of description, the adapter of the main controller is denoted as 120, and the adapter of the standby controller is denoted as 220.

For example, the electronic device may be a storage device, and the electronic device may include multiple hard disks. As mentioned above, the backplane can be provided with multiple interfaces, and the interface can be an SFF-8639 interface, which can be directly connected to the hard disk, or can be connected to the expansion processing chip through a network switch chip. In order to distinguish the description, the The interface connected with the network switch chip in the backplane is called the first type interface, and the interface connected with the hard disk in the backplane is called the second type interface.

The above-mentioned adapter may be an HBA (Host Bus Adapter, host bus adapter). In one case, the HBA of the controller may be directly connected to the hard disk through the second type of interface. Referring to FIG. 5, the HBAs (120 and 220) of each controller perform data interaction with the second-type interface on the backplane; the second-type interface on the backplane is connected to the hard disk 800; in this way, the HBA of the controller communicates with the second-type interface on the backplane. Data can be exchanged between hard drives.

Or, in another case, if the number of hard disks is large, an expansion component 900 may be provided in the backplane, for example, it may be an expander chip used for expanding the number of hard disks. Referring to FIG. 6, the HBAs (120 and 220) of each controller are connected to the expansion assembly 900, and the expansion assembly 900 can perform data interaction with the second-type interface on the backplane; the second-type interface on the backplane It is connected to the hard disk 800; in this way, data interaction can be performed between the HBA of the controller and the hard disk.

In one embodiment, the main controller 100 writes the data processing status parameters of the multiple expansion processing chips 300 into a designated area of the hard disk 800; if the main controller 100 fails, the standby controller 200 switches to the main controller to process services. Afterwards, the designated area is accessed, the processing state parameter is acquired, and the service is processed based on the processing state parameter.

As described above, the controller can monitor the data processing status of the extended processing chip. In this embodiment, the main controller writes the monitored data processing status parameters of the extended processing chip into the hard disk. Continuing the above example (an example in which each extended processing chip processes 5 images), taking extended processing chip 1 as an example, after extended processing chip 1 has processed the Mth (positive integer between 1-15) image, The processing result of the M th image is fed back to the controller, and the controller can record the processing state parameter of the extended processing chip 1, and the processing state parameter carries the information that the extended processing chip 1 has completed the processing of the M th image.

The main controller can periodically or aperiodically write the recorded processing state parameters into a designated area of the hard disk, and the designated area can be understood as a shared area of the two controllers, and both controllers can access the designated area. For example, the controller can write the recorded processing status parameters into a designated area of the hard disk once every minute, so that when the main controller fails, after the standby controller is switched to the main controller, the data can be read from the designated area of the hard disk. Process state parameters. This process can be understood as the migration process of the business in the controller. Another controller takes over the running business based on the read processing status parameters, so that the business can continue to run, and the extended processing chip can continue to process data.

The electronic device provided by the embodiment of the present invention includes two controllers, wherein one controller is the main controller and the other controller is the standby controller. In the case that the two controllers are not faulty, the main controller processes the business , the standby controller detects whether the main controller is faulty. If the main controller fails, the standby controller switches to the main controller to process services. In the process of processing services, it opens its own intranet network port to communicate with the multiple expansion processing chips; It can be seen that in this solution, in the first aspect, even if one controller fails, another controller can take over the business of the faulty controller, and can communicate with multiple extended processing chips, and then the multiple extended processing chips can still process data. processing, reducing the situation of business interruption; secondly, the controller controls the communication with the expansion processing chip through its own internal network port, and does not need to allocate an external IP address for the expansion processing chip, which saves external IP resources and simplifies the configuration. In the third aspect, the controller controls the communication with the expansion processing chip through its own intranet network port, which isolates the expansion processing chip from the external network and improves the security; in the fourth aspect, generally speaking, the intranet communication address is fixed. , and the external network IP address is assigned by the user. In this way, for the device, the internal network communication address is more convenient to manage than the external network communication address.

Some related active-standby switching solutions usually include: a master node, a standby node, and a computing node. The computing node is the node that analyzes and processes business data. When the master node does not fail, the master node manages the computing node. , the master node communicates with the computing node. If the master node fails, the backup node replaces the master node to manage the computing node, and the backup node communicates with the computing node. The master node and the backup node perform the master-slave switchover through the software communication mechanism. In other words, the communication link between the master node and the computing node is connected in hardware, and the communication link between the backup node and the computing node is in the The hardware is also connected, and the master node and the standby node control whether they communicate with the computing node through their own software communication mechanism (their own business logic). In this solution, the master node and the backup node need to use different addresses to communicate with the computing node, otherwise there will be address conflict. In this way, when the master and backup switch is performed, the computing node needs to switch the destination address to communicate normally, which makes the computing node logic is more complicated.

In this solution, the controller controls the communication with the extended processing chip by opening its own internal network port on the hardware. In other words, only one communication link between the controller and the extended processing chip is connected on the hardware. , when one controller fails, the other controller connects the hardware communication link between itself and the expansion processing chip, so that the two controllers use the same address to communicate with the expansion processing chip, and there will be no address conflict. Therefore, the extended processing chip does not need to switch the destination address, which simplifies the logic of the extended processing chip, and realizes the active-standby switching without the extended processing chip sensing.

In some of the above active-standby switching schemes, in order to reduce the address conflict between the master node and the standby node, the means used are: the computing node switches the destination address; In the case of address conflict, the method used is: when the master node is not faulty, the master node and the backup node use different addresses. If the master node fails, the backup node changes its own address to the address of the master node, which is also called the address of the master node. In this solution, the computing node does not need to switch the destination address, but the standby node needs to switch its own address, which makes the logic of the standby node more complicated.

In this solution, the controller controls the communication with the extended processing chip by opening its own internal network port on the hardware. In other words, only one communication link between the controller and the extended processing chip is connected on the hardware. , when one controller fails, the other controller connects the hardware communication link between itself and the expansion processing chip, so that the two controllers use the same address to communicate with the expansion processing chip, and there will be no address conflict. Therefore, the standby controller does not need to switch the destination address, which simplifies the logic of the standby controller.

Embodiments of the present invention further provide a service switching method and apparatus. The method and apparatus can be applied to a first controller in an electronic device. The device further includes a second controller and a plurality of extended processing chips. Each controller All include intranet ports, and each controller is connected to the plurality of extended processing chips through its own intranet ports. The service switching method will be described below with reference to FIG. 7 .

FIG. 7 is a flowchart of a service switching method provided by an embodiment of the present invention, including:

S701: Determine whether the first controller is the main controller; if yes, go to S702, and if not, go to S703.

The first controller in S701 is the execution body of the method embodiment, and the first controller may be any controller in the electronic device.

For example, when the electronic device starts up, the two controllers can negotiate which controller becomes the master controller through the heartbeat network port. For example, when an electronic device starts up, the two controllers usually have a difference in the speed of startup completion after power-on. The controller that starts up quickly after power-on (or the controller that completes startup first) can be used as the main controller. The other controller acts as the standby controller. Specifically, after the controller is powered on and started, the heartbeat network port can be used to detect whether the peer controller is powered on and started. If the power-on and start-up is not completed, the controller that has been powered on and started becomes the master controller.

Each controller is configured with two kinds of business logic, one is the business logic of the primary controller, and the other is the business logic of the standby controller. For each controller, the controller determines whether it is the main controller or the standby controller. If it is the main controller, it enters the main controller business logic; if it is the standby controller, it enters the standby controller business logic. The business logic of the main controller is to execute S702, and the business logic of the standby controller is to execute S703-S704.

S702: During the service processing process of the first controller, communicate with the plurality of extended processing chips through the intranet network port of the first controller.

S702 and subsequent S704 may include: in the process of processing the service by the first controller, sending the service data to be processed to the plurality of extended processing chips through the intranet network port of the first controller; The intranet network port receives the data processing results sent by the multiple expansion processing chips.

Generally speaking, the controller obtains the service data to be processed, and sends the service data to a plurality of extended processing chips. Each extended processing chip processes the service data received by itself, and feeds back the processing result to the controller. In the embodiment of the present invention, the communication process between the controller and the extended processing chip may include: the controller sends service data to the extended processing chip, and the extended processing chip sends data processing results to the controller; in addition, the controller may also control the extended processing chip. The controller can also monitor the data processing status of the extended processing chip during reset, startup, and so on, which is not specifically limited.

S703: Detect whether the second controller serving as the main controller is faulty; if it is faulty, execute S704, and if not, continue to detect.

For example, each controller may include a heartbeat network port, and whether the peer controller is faulty is detected between the master controller and the standby controller through the heartbeat network port.

S704: Switch to the main controller to process services; in the process of processing the services, open the intranet network port of the first controller; communicate with the plurality of expansion processing chips through the intranet network port of the first controller.

The electronic device provided in this embodiment includes two controllers. In one case, the two controllers may be the same component in hardware. For example, as shown in FIG. 2 , the main controller 100 may include a main board and a controller circuit disposed on the main board; the controller circuit includes: a CPU (Central Processing Unit, central processing unit), a memory, an HBA (Host Bus Adapter, host bus adapter) ), the external network card that provides the external network network port, the internal network card that provides the internal network network port 110 and the heartbeat network card that provides the heartbeat network port; The main board and the controller circuit that is arranged on the main board can be included in the backup controller 200; Control The device circuit includes: CPU, memory, HBA, external network card providing external network port, internal network card providing internal network port 210 and heartbeat network card providing heartbeat network port. Among them, the internal network port is used for internal communication of electronic equipment. For example, the controller communicates with the expansion processing chip through the internal network port; the external network port is used for communication between electronic equipment and external equipment. The network port obtains the business data to be processed; the HBA can be connected to the hard disk and exchange data with the hard disk; the heartbeat network ports of the two controllers are connected, and the two controllers can detect whether the other party is faulty through the heartbeat network port.

In this embodiment, one controller in the electronic device is used to process services, and the other controller is used as a backup. The controller that processes the service is called the main controller, and the controller that serves as the backup is the backup controller, and the two can perform active-standby switchover. During the business process of the main controller, the main controller communicates with the multiple extended processing chips. In this case, the internal network port of the standby controller is closed, and the standby controller cannot communicate with the multiple extended processing chips. ; If the main controller fails, the standby controller switches to the main controller to process services; in the process of processing services, it starts its own intranet network port to communicate with the multiple expansion processing chips. In other words, if the primary controller fails, the backup controller takes over the primary controller to process services.

After S704, the first controller may control one or more extended processing chips of the plurality of extended processing chips to reset.

For example, suppose there are three extended processing chips: extended processing chip 1, extended processing chip 2, and extended processing chip 3; assuming that the main controller obtains 15 images to be processed, the main controller will , 10, 13 images are sent to extended processing chip 1, 2, 5, 8, 11, 14 images are sent to extended processing chip 2, 3, 6, 9, 12, 15 images are sent to extended processing Chip 3; Suppose the main controller fails, at this time the extended processing chip 1 is processing the seventh image, the extended processing chip 2 is processing the eighth image, and the extended processing chip 3 is processing the ninth image After the standby controller 200 detects the failure of the main controller 100 through the heartbeat network port, the standby controller 200 switches to the main controller to process business and opens its own intranet network port 210, and then controls these three expansion processing chips to reset; After reset, the extended processing chip 1 reprocesses the seventh image, the extended processing chip 2 reprocesses the eighth image, and the extended processing chip 3 reprocesses the ninth image.

In the above example, if the controller is switched in the process of processing an image by the extended processing chip, it is likely to interfere with the processing of the image by the extended processing chip, resulting in an error in the processing result of the image; in this embodiment, , after switching the controller, the extended processing chip is reset, and the image is processed again, which reduces the error of the processing result and improves the accuracy of the processing result.

In an embodiment, after S702 and/or S704, the first controller may write the data processing state parameters of the plurality of extended processing chips into a preset storage area; When the second controller of the controller fails, the processing state parameter may be acquired by accessing the preset storage area, and the service may be processed based on the processing state parameter.

As described above, the controller can monitor the data processing status of the extended processing chip. In this embodiment, the main controller writes the monitored data processing status parameters of the extended processing chip into the hard disk. Continuing the above example (an example in which each extended processing chip processes 5 images), taking extended processing chip 1 as an example, after extended processing chip 1 has processed the Mth (positive integer between 1-15) image, The processing result of the M th image is fed back to the controller, and the controller can record the processing state parameter of the extended processing chip 1, and the processing state parameter carries the information that the extended processing chip 1 has completed the processing of the M th image.

The main controller can periodically or aperiodically write the recorded processing state parameters into a designated area of the hard disk, and the designated area can be understood as a shared area of the two controllers, and both controllers can access the designated area. For example, the main controller can write the recorded processing status parameters into the designated area of the hard disk once every minute, so that when the main controller fails, after the standby controller switches to the main controller, it can read from the designated area of the hard disk. The processing state parameter. This process can be understood as the migration process of the business in the controller. After one controller fails, another controller takes over the running business based on the read processing status parameters, so that the business can continue to run, that is, expanding the processing chip Continue with data processing.

Continuing the above example, it is assumed that the second controller as the main controller writes the processing status parameter "Extended processing chip 1 completes the processing of the M-th image" into the designated area of the hard disk, and then the second controller fails, and the first controller detects After the second controller fails, the first controller switches itself to the main controller, the first controller reads the processing status parameter from the designated area, and continues to allocate the extended processing chip 1 based on the processing status parameter. business data. For example, assuming that M is 13, it means that the extended processing chip 1 is about to finish processing the thirteenth image, and the first controller can prepare to send the next image to the extended processing chip 1 .

Applying the embodiment shown in the present invention, the electronic setting includes two controllers, each controller determines whether it is the main controller or the standby controller; The expansion processing chip communicates; if it is a standby controller, it detects whether the main controller is faulty. If the main controller fails, it switches itself to the main controller to process services, and opens its own intranet network port and the multiple expansion processing chips. It can be seen that, in the first aspect, even if one controller fails, another controller can take over the business of the faulty controller, and can communicate with multiple extended processing chips, and then the multiple extended processing chips can still perform data processing. , reducing the situation of business interruption; secondly, the controller controls the communication with the expansion processing chip through its own internal network port, and does not need to allocate an external IP address for the expansion processing chip, which saves external IP resources and simplifies the configuration; In the third aspect, the controller controls the communication with the expansion processing chip through its own intranet network port, which isolates the expansion processing chip from the external network and improves security; in the fourth aspect, generally speaking, the intranet communication address is fixed , and the external network IP address is assigned by the user. In this way, for the device, the internal network communication address is more convenient to manage than the external network communication address.

Some related active-standby switching solutions usually include: a master node, a standby node, and a computing node. The computing node is the node that analyzes and processes business data. When the master node does not fail, the master node manages the computing node. , the master node communicates with the computing node. If the master node fails, the backup node replaces the master node to manage the computing node, and the backup node communicates with the computing node. The master node and the backup node perform the master-slave switchover through the software communication mechanism. In other words, the communication link between the master node and the computing node is connected in hardware, and the communication link between the backup node and the computing node is in the The hardware is also connected, and the master node and the standby node control whether they communicate with the computing node through their own software communication mechanism (their own business logic). In this solution, the master node and the backup node need to use different addresses to communicate with the computing node, otherwise there will be address conflict. In this way, when the master and backup switch is performed, the computing node needs to switch the destination address to communicate normally, which makes the computing node logic is more complicated.

In this solution, the controller controls the communication with the extended processing chip by opening its own internal network port on the hardware. In other words, only one communication link between the controller and the extended processing chip is connected on the hardware. , when one controller fails, the other controller connects the hardware communication link between itself and the expansion processing chip, so that the two controllers use the same address to communicate with the expansion processing chip, and there will be no address conflict. Therefore, the extended processing chip does not need to switch the destination address, which simplifies the logic of the extended processing chip, and realizes the active-standby switching without the extended processing chip sensing.

In some of the above active-standby switching schemes, in order to reduce the address conflict between the master node and the standby node, the means used are: the computing node switches the destination address; In the case of address conflict, the method used is: when the master node is not faulty, the master node and the backup node use different addresses. If the master node fails, the backup node changes its own address to the address of the master node, which is also called the address of the master node. In this solution, the computing node does not need to switch the destination address, but the standby node needs to switch its own address, which makes the logic of the standby node more complicated.

In this solution, the controller controls the communication with the extended processing chip by opening its own internal network port on the hardware. In other words, only one communication link between the controller and the extended processing chip is connected on the hardware. , when one controller fails, the other controller connects the hardware communication link between itself and the expansion processing chip, so that the two controllers use the same address to communicate with the expansion processing chip, and there will be no address conflict. Therefore, the standby controller does not need to switch the destination address, which simplifies the logic of the standby controller.

Corresponding to the foregoing method embodiments, an embodiment of the present invention further provides a communication device, as shown in FIG. 8 ; the device includes:

The judgment module 801 is used to judge whether the first controller is the main controller; if so, trigger the first processing module 802; if not, trigger the second processing module 803;

A first processing module 802, configured to communicate with the plurality of extended processing chips through the intranet network port of the first controller during the business process of the first controller;

The second processing module 803 is configured to detect whether the second controller serving as the main controller is faulty; if it fails, switch the first controller to the main controller to process services; in the process of processing services, enable the first controller The intranet network port of the controller; communicates with the plurality of expansion processing chips through the intranet network port of the first controller.

As an implementation manner, the first processing module 802 is further configured to:

In the case that the first controller is the main controller, the service data to be processed is sent to the plurality of extended processing chips through the intranet network port of the first controller; through the first controller The internal network port of the device receives the data processing results sent by the multiple expansion processing chips;

The second processing module 803 is also used for:

In the case that the first controller is the standby controller and the failure of the second controller, which is the main controller, is detected, the service data to be processed is sent to the intranet port of the first controller to the multiple extended processing chips; receive the data processing results sent by the multiple extended processing chips through the intranet network port of the first controller.

As an implementation manner, the apparatus further includes: a writing module and an acquiring module (not shown in the figure), wherein,

The writing module is configured to write the data processing state parameters of the multiple expansion processing chips into the pre-processor after the communication with the multiple expansion processing chips through the internal network port of the first controller. set storage area;

An obtaining module, configured to obtain the processing state parameter by accessing the preset storage area when the detection module detects a failure of the main controller, and process services based on the processing state parameter.

As an embodiment, the device further includes:

A reset module (not shown in the figure) is used for, in the case of detecting the failure of the second controller as the main controller, and after the opening of the internal network port of the first controller, to control the One or more extended processing chips among the multiple extended processing chips are reset.

Applying the embodiment shown in the present invention, the electronic setting includes two controllers, each controller determines whether it is the main controller or the standby controller; The expansion processing chip communicates; if it is a standby controller, it detects whether the main controller is faulty. If the main controller fails, it switches itself to the main controller to process services, and opens its own intranet network port and the multiple expansion processing chips. It can be seen that, in the first aspect, even if one controller fails, another controller can take over the business of the faulty controller, and can communicate with multiple extended processing chips, and then the multiple extended processing chips can still perform data processing. , reducing the situation of business interruption; secondly, the controller controls the communication with the expansion processing chip through its own internal network port, and does not need to allocate an external IP address for the expansion processing chip, which saves external IP resources and simplifies the configuration; In the third aspect, the controller controls the communication with the expansion processing chip through its own intranet network port, which isolates the expansion processing chip from the external network and improves security; in the fourth aspect, generally speaking, the intranet communication address is fixed , and the external network IP address is assigned by the user. In this way, for the device, the internal network communication address is more convenient to manage than the external network communication address.

Some related active-standby switching solutions usually include: a master node, a standby node, and a computing node. The computing node is the node that analyzes and processes business data. When the master node does not fail, the master node manages the computing node. , the master node communicates with the computing node. If the master node fails, the backup node replaces the master node to manage the computing node, and the backup node communicates with the computing node. The master node and the backup node perform the master-slave switchover through the software communication mechanism. In other words, the communication link between the master node and the computing node is connected in hardware, and the communication link between the backup node and the computing node is in the The hardware is also connected, and the master node and the standby node control whether they communicate with the computing node through their own software communication mechanism (their own business logic). In this solution, the master node and the backup node need to use different addresses to communicate with the computing node, otherwise there will be address conflict. In this way, when the master and backup switch is performed, the computing node needs to switch the destination address to communicate normally, which makes the computing node logic is more complicated.

In this solution, the controller controls the communication with the extended processing chip by opening its own internal network port on the hardware. In other words, only one communication link between the controller and the extended processing chip is connected on the hardware. , when one controller fails, the other controller connects the hardware communication link between itself and the expansion processing chip, so that the two controllers use the same address to communicate with the expansion processing chip, and there will be no address conflict. Therefore, the extended processing chip does not need to switch the destination address, which simplifies the logic of the extended processing chip, and realizes the active-standby switching without the extended processing chip sensing.

In some of the above active-standby switching schemes, in order to reduce the address conflict between the master node and the standby node, the means used are: the computing node switches the destination address; In the case of address conflict, the method used is: when the master node is not faulty, the master node and the backup node use different addresses. If the master node fails, the backup node changes its own address to the address of the master node, which is also called the address of the master node. In this solution, the computing node does not need to switch the destination address, but the standby node needs to switch its own address, which makes the logic of the standby node more complicated.

In this solution, the controller controls the communication with the extended processing chip by opening its own internal network port on the hardware. In other words, only one communication link between the controller and the extended processing chip is connected on the hardware. , when one controller fails, the other controller connects the hardware communication link between itself and the expansion processing chip, so that the two controllers use the same address to communicate with the expansion processing chip, and there will be no address conflict. Therefore, the standby controller does not need to switch the destination address, which simplifies the logic of the standby controller.

Embodiments of the present invention further provide a computer-readable storage medium, where a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, any one of the foregoing service switching methods is implemented.

It should be noted that, in this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any relationship between these entities or operations. any such actual relationship or sequence exists. Moreover, the terms "comprising", "comprising" or any other variation thereof are intended to encompass a non-exclusive inclusion such that a process, method, article or device that includes a list of elements includes not only those elements, but also includes not explicitly listed or other elements inherent to such a process, method, article or apparatus. Without further limitation, an element qualified by the phrase "comprising a..." does not preclude the presence of additional identical elements in a process, method, article or apparatus that includes the element.

Each embodiment in this specification is described in a related manner, and the same and similar parts between the various embodiments may be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the method embodiments, apparatus embodiments, and computer-readable storage medium embodiments, since they are basically similar to the electronic device embodiments, the description is relatively simple, and for related parts, please refer to the partial descriptions of the electronic device embodiments. .

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims (11)

1. An electronic device, comprising: the system comprises a main controller, a controller and a plurality of extended processing chips; each controller comprises an intranet net port, and each controller is connected with the plurality of extended processing chips through the intranet net port of the controller;

the main controller communicates with the plurality of extended processing chips through an intranet port of the main controller in the process of processing the service;

the backup controller detects whether the main controller fails; if the fault occurs, the backup controller switches to the main controller to process the service; the switched main controller starts an intranet port of the main controller in the process of processing the service; and communicating with the plurality of extended processing chips through the switched intranet net port of the main controller.

2. The apparatus of claim 1, further comprising a first network switch chip and a backplane;

the first network switching chip is arranged on the back plate; the intranet net port of each controller is connected with the first network exchange chip; the first network exchange chip is connected with the plurality of extended processing chips through a first type interface on the backboard.

3. The device of claim 2, wherein the plurality of extended processing chips are disposed on a tray, the tray further having a second network switch chip disposed thereon, the tray being inserted into the first type of interface on the backplane, the plurality of extended processing chips being connected to the second network switch chip, the second network switch chip being connected to the first type of interface on the backplane.

4. The device of claim 3, wherein the plurality of extended processing chips are disposed on a plurality of trays, each tray is disposed with a second network switching chip and one or more extended processing chips, each tray is inserted into one of the first type interfaces on the backplane, each extended processing chip is connected to the second network switching chip on the tray where the extended processing chip is located, and each second network switching chip is connected to the first type interface into which the tray where the extended processing chip is located is inserted.

5. The device of claim 2, further comprising a hard disk, the hard disk being connected to the second type of interface on the backplane; each controller comprises an adapter, and each controller is connected with the second type interface on the back panel through the adapter of the controller.

6. The apparatus according to claim 5, wherein the main controller writes the processing state parameters of the plurality of extended processing chips on the data into a designated area of the hard disk; and if the main controller fails, the backup controller accesses the specified area after switching to the main controller to process the service, acquires the processing state parameters, and processes the service based on the processing state parameters.

7. The device according to claim 1, wherein each controller comprises a heartbeat network port, and the main controller and the slave controller detect whether the opposite controller fails through the heartbeat network port.

8. The device according to claim 1, wherein the controller controls the plurality of extended processing chips to reset after switching to the main controller to process the service and starting an intranet port of the controller.

9. The device of claim 5, wherein the extended processing chip is a GPU, and the first type of interface and the second type of interface are both SFF-8639 interfaces.

10. A service switching method is characterized in that the service switching method is applied to a first controller in electronic equipment, the equipment further comprises a second controller and a plurality of extended processing chips, each controller comprises an intranet network port, and each controller is connected with the plurality of extended processing chips through the intranet network port of the controller; the method comprises the following steps:

If the first controller is a main controller, communicating with the plurality of extended processing chips through an intranet port of the first controller in the process that the first controller processes the service;

if the first controller is a controller, detecting whether a second controller serving as a main controller fails; if the fault occurs, the first controller is switched to a main controller to process the service; in the process of processing the service, an intranet port of the first controller is started; and the internal network port of the first controller is communicated with the plurality of extended processing chips.

11. A service switching device is characterized in that the service switching device is applied to a first controller in electronic equipment, the equipment further comprises a second controller and a plurality of extended processing chips, each controller comprises an intranet network port, and each controller is connected with the plurality of extended processing chips through the intranet network port of the controller; the device comprises:

the first processing module is used for communicating with the plurality of extended processing chips through the intranet net port of the first controller in the process that the first controller processes the business under the condition that the first controller is a main controller;

The second processing module is used for detecting whether a second controller serving as a main controller fails or not under the condition that the first controller is a slave controller; if the fault occurs, the first controller is switched to a main controller to process the service; in the process of processing the service, an intranet port of the first controller is started; and the internal network port of the first controller is communicated with the plurality of extended processing chips.

Images