WO2023109880A1

WO2023109880A1 - Service recovery method, data processing unit and related device

Info

Publication number: WO2023109880A1
Application number: PCT/CN2022/139182
Authority: WO
Inventors: 冷超
Original assignee: 华为技术有限公司
Priority date: 2021-12-16
Filing date: 2022-12-15
Publication date: 2023-06-22
Also published as: CN116266150A

Abstract

The present application provides a service recovery method. The method is executed by a DPU interface card, and specifically, after software (e.g. an operating system, another application program, etc.) of the DPU interface card is restarted, the DPU interface card acquires service information stored in a memory of a host, wherein the service information is information generated when the DPU interface card processes, before the software is restarted, an IO sent by a processor of the host, such that the DPU interface card recovers a service according to the service information stored in the memory of the host. Therefore, a DPU interface card does not need to depend on a local memory to recover an interrupted service, such that even if service data of a service is lost due to a version upgrade of software (e.g. an operating system, etc.) of the DPU interface card, a memory fault of the DPU interface card or other reasons, the DPU interface card can also quickly recover the service by using a memory of a host, thereby reducing the impact on the service.

Description

A service recovery method, data processing unit and related equipment

This application is required to be submitted to the State Intellectual Property Office of China on December 16, 2021, the application number is 202111540861.5, and the application name is "a data processing unit processing method, data processing unit processing and system" and submitted to China on March 18, 2022. The State Intellectual Property Office has the priority of a Chinese patent application with application number 202210269274.5 and titled "A Service Restoration Method, Data Processing Unit and Related Equipment", the entire contents of which are incorporated by reference in this application.

technical field

The present application relates to the technical field of data processing, and in particular to a service restoration method, a data processing unit and related equipment.

Background technique

With the development of business processing requirements, the computing power of a central processing unit (central processing unit, CPU) is required to be higher and higher. At present, the data processing unit (data processing unit, DPU) interface card can be used as the offload engine of the CPU, and can achieve more efficient data processing capabilities by cooperating with the CPU to process services. Wherein, the DPU interface card may include an application specific integrated circuit (ASIC), a processor, and a memory. The ASIC and the processor may provide computing power for the CPU, and the memory may temporarily store business information of the DPU interface card.

In actual application scenarios, due to the upgrade of the operating system of the DPU interface card, or a memory failure in the DPU interface card, such as a row (row), column (column) or storage array (bank) failure in the memory, etc., the DPU interface card The operating system is restarted, which causes the loss of service information stored in the memory of the DPU interface card, resulting in service interruption.

Contents of the invention

In view of this, the embodiment of the present application provides a service recovery method, so as to enable the DPU interface card to recover services after the software in the DPU interface card is restarted. The present application also provides corresponding data processing units, computing devices, interface cards, computer-readable storage media, and computer program products.

In the first aspect, the embodiment of the present application provides a service recovery method, the method is executed by a DPU interface card, and the DPU interface card is coupled with the host, for example, can be coupled through a bus; wherein, when the DPU interface card resumes services, Specifically, after the software (such as the operating system, etc.) of the DPU interface card is restarted, the service information stored in the memory of the host is obtained. The generated information, so that the DPU interface card can resume and process the interrupted service according to the service information stored in the memory of the host.

Since the DPU interface card can use the service information stored in the memory of the host computer to restore the service in time after restarting the software, it does not need to rely on the local memory of the DPU interface card to restore the interrupted service. System, etc.) due to version upgrade or memory failure and other reasons cause the software of the DPU interface card to restart, and business information is lost. The DPU interface card can also use the host memory to quickly restore the business, reducing the impact on the business.

In a possible implementation manner, the DPU interface card is coupled to the host based on the PCIe bus, and the PCIe link between the DPU interface card and the host is not disconnected during the restart process of the software in the DPU interface card. In this way, in the process of restarting the software, the host may not perceive the state change of the DPU interface card, thereby reducing the impact on the host.

In a possible implementation manner, the software restart of the DPU interface card is triggered by a failure of the DPU interface card or is triggered by an upgrade of the previous version of the software of the DPU interface card, so as to realize the fault recovery of the DPU interface card Or a software upgrade.

In other implementation manners, the DPU interface card may also restart the software after receiving an instruction to restart the software sent by the host.

In a possible implementation manner, when the memory of the DPU interface card fails, the DPU interface card restarts the software, so as to realize the repair of the faulty memory of the DPU interface card.

In a possible implementation manner, when the memory of the DPU interface card fails and the failed memory does not satisfy a preset condition, the DPU interface card restarts the operating system of the DPU interface card. In this way, when a memory failure occurs, the DPU interface card can decide whether to restart the operating system of the DPU interface card according to the failure of the memory, so as to implement constraints on restarting the operating system of the DPU interface card.

In a possible implementation, when the memory of the DPU interface card fails and the failed memory meets the preset conditions, the DPU interface card can also restart the kernel of the operating system of the DPU interface card to use the failed memory area service component. In this way, the DPU interface card does not need to restart the entire operating system, thereby reducing the impact of the faulty memory on the DPU interface card as much as possible.

Exemplarily, the preset condition that the failed memory in the DPU interface card satisfies, for example, the size of the failed memory does not exceed the preset size, such as the number of failed rows (or columns) in the memory does not exceed the preset number of rows (or preset number of columns), etc. At this time, the impact of the faulty memory on the DPU interface card is relatively small, so that the DPU interface card can implement fault recovery without restarting the entire operating system.

Or, the preset condition satisfied by the faulty memory may specifically be that the system component using the faulty memory is a preset system component, so that when the faulty memory affects a specific system component, the DPU interface card can restart the Part of the system components to achieve failure recovery.

Alternatively, the preset condition that the faulty memory satisfies may specifically be that the number of system components using the faulty memory does not exceed a preset number. At this time, the faulty memory only affects a small number of service components, but does not affect the rest of the service components. Therefore, the DPU interface card can restart the part of the affected service components without restarting the entire operating system and all service components. for fault recovery.

In a possible implementation manner, when the DPU interface card obtains the service information stored in the memory of the host, it may specifically obtain the first address identifier from the memory area, where the first address identifier is used to identify the memory area in the host, The memory area is a storage area in the memory of the host for storing business information, where the memory area can be a storage area in a volatile memory or a storage area in a non-volatile memory, etc., and, when When the operating system in the DPU interface card is restarted, the data stored in the memory area will not be lost; in this way, after the operating system is restarted, the DPU interface card can access the memory area of the host according to the first address identifier to obtain business information .

Exemplarily, the memory area can be located inside the DPU interface card, for example, it can be a logic block in the CPLD included in the DPU interface card, or the memory area can be located outside the DPU interface card, for example, it can be connected to the DPU interface card Storage area in external memory, etc.

In a possible implementation manner, before the software is restarted, the DPU interface card may also apply to the host for a memory area for storing service information, and obtain the first address identifier of the memory area. The first address identifier may be, for example, is the first address of the memory area, etc., so that the DPU interface card can store service information in the memory area according to the first address identifier. In this way, after the software in the DPU interface card is restarted, the DPU interface card can use the service information stored in the memory area to implement service recovery.

In a possible implementation manner, the DPU interface card may also obtain configuration information from the memory area, and the configuration information is used to configure the DPU interface card, where the configuration information may specifically include the second The address identifier (such as the first address of the sending queue, etc.) and the third address identifier of the completion queue (such as the first address of the completion queue, etc.), the sending queue is used to store the IO sent by the processor of the host, and the completion queue is used to store the DPU The execution result of the interface card for this IO.

Optionally, the configuration information may also include a communication format, a communication protocol version, etc. during data exchange between the DPU interface card and the host, or the configuration information may also include other content.

In the second aspect, the embodiment of the present application further provides a data processing unit DPU device, configured to execute the service recovery method described in the first aspect or any implementation manner of the first aspect.

In a third aspect, the present application provides a computing device, the computing device includes a host and a DPU (data processing unit) interface card, wherein the host includes a memory and a processor, and the DPU interface card is used to perform the following operations: After the software of the DPU interface card is restarted, obtain the business information stored in the memory of the host, the business information is before the software restarts, the DPU interface card processes the input and output IO sent by the processor Generated information; resume services according to the service information. Exemplarily, the DPU interface card is used to execute the service restoration method described in the first aspect or any implementation manner of the first aspect.

In a fourth aspect, the present application provides a data processing unit DPU interface card, the DPU interface card includes a printed circuit board, an interface, and a data processing unit DPU chip, the interface card communicates with the host through the interface, and the interface communicates with the The DPU is installed on the printed circuit board, and the DPU chip is used to obtain the service information stored in the memory of the host after the software of the DPU interface card is restarted, and the service information is in the software Before restarting, the DPU interface card processes the information generated by the input and output IO sent by the processor of the host; resumes the service according to the service information.

Exemplarily, the DPU chip in the DPU interface card may be used to execute the service restoration method described in the first aspect or any implementation manner of the first aspect.

In a fifth aspect, the present application provides a data processing unit DPU chip, which is applied to a DPU interface card, the DPU interface card is coupled to a host, and the DPU chip includes an acquisition circuit and a processing circuit, wherein the acquisition circuit is used for After the software of the DPU interface card is restarted, the business information stored in the memory of the host is obtained, and the business information is before the software restarts, the processing circuit processes the input and output sent by the processor of the host Information generated by the IO; the processing circuit is used to restore services according to the service information.

Exemplarily, the obtaining circuit and the processing circuit cooperate with each other and may be used to execute the service recovery method as described in the first aspect or any implementation manner of the first aspect.

On the basis of the implementation manners provided in the foregoing aspects, the present application may further be combined to provide more implementation manners.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings that need to be used in the description of the embodiments. Obviously, the drawings in the following description are only some implementations recorded in the application. For example, those skilled in the art can also obtain other drawings based on these drawings.

Fig. 1 is the schematic diagram of the architecture of an exemplary DPU interface card that the embodiment of the present application provides;

FIG. 2 is a schematic flowchart of a service recovery method provided in an embodiment of the present application;

FIG. 3 is a schematic structural diagram of a data processing unit DPU device provided in an embodiment of the present application;

FIG. 4 is a schematic diagram of a hardware structure of a computing device provided by an embodiment of the present application;

FIG. 5 is a schematic structural diagram of a DPU interface card provided by an embodiment of the present application;

FIG. 6 is a schematic structural diagram of a DPU chip provided by an embodiment of the present application.

Detailed ways

The solutions in the embodiments provided in the present application will be described below with reference to the drawings in the present application.

The terms "first", "second", "third" and the like in the description and claims of the present application and the above drawings are used to distinguish similar objects, and not necessarily used to describe a specific sequence or sequence. It should be understood that the terms used in this way can be interchanged under appropriate circumstances, and this is merely a description of the manner in which objects with the same attribute are described in the embodiments of the present application.

The DPU interface card is used as the offload engine of the CPU. During the process of assisting the CPU in processing services, it usually temporarily stores service data in the local memory of the DPU interface card. If the memory of the DPU interface card fails or the software (such as the operating system) etc.) the version upgrade causes the corresponding software of the DPU interface card to restart, resulting in the loss of data in the memory of the DPU interface card. For example, in actual application, some memory areas of the DPU interface card will inevitably fail. In order to repair the memory failure, the DPU software needs to be restarted. It is difficult for Kaka to continue processing the business due to missing business data. In addition, when the operating system or other software in the DPU interface card needs to be upgraded, the service data stored in the memory of the DPU interface card will also be lost due to restarting the operating system of the DPU interface card, thus affecting the processing of the DPU interface card. the business.

Based on this, an embodiment of the present application provides a service recovery method to recover services interrupted by the DPU interface card. During specific implementation, before restarting the software of the DPU interface card, the business information generated by the DPU interface card processing business is pre-stored in the memory of the host computer, so that after the DPU interface card restarts the software (such as because of memory failure or software version) The upgrade triggers the DPU interface card to re-operate the operating system, etc.), and the service information can be obtained from the memory of the host, so that the DPU interface card can resume and process the interrupted service by using the service information. In this way, even if the software of the DPU interface card is restarted due to the reasons described above, resulting in the loss of business data stored in the memory of the DPU interface card, the DPU interface card can also use the host memory to quickly restore services without relying on the local memory of the DPU interface card. Reduce impact on business.

Exemplarily, the above service recovery method may be applied to the DPU interface card 100 shown in FIG. 1 . As shown in FIG. 1 , the DPU interface card 100 is coupled to the host 200 through a peripheral component interconnect express (PCIe) bus or other buses. Moreover, the DPU interface card 100 includes a printed circuit board 1011 , an interface 1012 , a DPU chip 101 and software 1013 , and the interface 1012 and the DPU chip 101 are installed on the printed circuit board 1011 . Specifically, the interface 1012 may be a PCIe interface. The software 1013 can be the operating system of the DPU interface card 100, and the operating system includes a system service component unit 102, a soft reset (soft reset) unit 103, a memory supervision unit 104, and a microkernel (micro kernel) unit 105. Further, the DPU interface card 100 may also include a micro reset (micro reset) unit 106 and the like. The specific implementation of the software shown in FIG. 1 can be located in the memory of the DPU interface card; the software shown in FIG. 1 can also be embedded, which is not limited in the embodiment of the present invention.

Among them, the DPU chip 101 is used to control the DPU interface card 100 to provide storage services for the host 200, that is, to process storage-type services, such as fast non-volatile storage (non-volatile memory express, NVMe), virtiofs (see https: //virtio-fs.gitlab.io/), virtio_scsi (for details, please refer to https://www.ovirt.org/develop/release-management/features/storage/virtio-scsi.html) business, etc. Alternatively, the DPU chip 101 may also control the DPU interface card 100 to provide computing services for the host 200, that is, to process computing-type services. Moreover, after the software such as the operating system of the DPU interface card 100 is restarted, the DPU chip 101 can control the DPU interface card 100 to resume the interrupted service (such as the above-mentioned storage type service or calculation type service, etc.).

The system service component unit 102 includes multiple service components in the kernel of the operating system, and the multiple service components can use the memory in the DPU interface card 100 to provide services for the operating system in the DPU interface card 100, as shown in Figure 1 Drivers components, file system components, memory management components, network protocol components, etc. are shown. Among them, the driver component is used to drive the DPU interface card 100 to perform data communication with the host 200, and may include a driver framework and a group of entity business drivers; the file system component is used to provide file system services, such as data storage in the form of files, Reading and management, etc.; memory management components, used to provide memory management services, such as allocation, recycling and isolation of memory areas, etc.; network protocol components, used to provide network protocol services, such as Hyper Text Transfer Protocol (Hyper Text Transfer Protocol) , HTTP) etc.

The soft reset unit 103 is used to reset the hardware units in the DPU interface card 100, and restart software such as the operating system of the DPU interface card 100.

The memory supervision unit 104 is used to perform fault monitoring, fault repair, fault isolation, redundant area replacement, etc. to the memory in the DPU interface card 100, so that the memory in the DPU interface card 100 has reliability, availability, and serviceability ( reliability, availability, serviceability, RAS).

The microkernel unit 105 is configured to manage the resources in the DPU interface card 100 and split the service components of the kernel so that the service components of the kernel can be restarted separately.

The micro-reset unit 106 is configured to individually restart the service components in the kernel of the control operating system through the micro-kernel architecture.

Normally, the DPU interface card 100 can assist the host 200 to process services. When the memory supervision unit 104 detects that the memory in the DPU interface card 100 fails, if the memory failure causes business interruption and the software of the DPU interface card 100 restarts, then the DPU interface card 100 can use the memory of the host computer 200 to store business information to restore the business. For example, the memory monitoring unit 104 can isolate the faulty memory location or replace the failed unit, and trigger the soft reset unit 103 to reset the hardware units in the DPU interface card 100 and restart the operating system in the DPU interface card 100 . Then, the microkernel unit 105 initializes the restarted hardware unit, and restarts each service component in the system service component unit 102 . After the soft reset unit 103 completes the restart of the operating system, the DPU chip 101 obtains the service information stored in the memory area 201 of the host 200, and uses the service information of the DPU interface card 100 to restore the service.

Further, when the scope of the memory failure in the DPU interface card 100 is relatively small, such as only involving network protocol components in the kernel of the operating system that use this part of the failed memory, at this time, the memory supervision unit 104 can correct the failed memory. location, and trigger the micro reset unit 106 to execute the micro reset process, specifically trigger the micro reset 106 to restart the service components that use the part of the failed memory, for example, the network protocol components can be restarted by the micro reset unit 106 (the rest are not Affected service components do not need to perform a restart process), so as to restore the normal network communication function of the DPU interface card 100 and the like.

It should be noted that, in FIG. 1 , the coupling between the DPU interface card 100 and the host 200 via the PCIe bus is taken as an example for illustration. In practical applications, the DPU interface card 100 can also be coupled with the host 200 in other ways. This embodiment does not limit it. Moreover, the architecture of the DPU interface card 100 shown in FIG. 1 is only used as an exemplary illustration. In actual application, the DPU interface card 100 can also adopt other architectures, such as the DPU interface card 100 can also include more other types of services components etc.

Next, various non-limiting specific implementations of the service restoration process are described in detail.

Referring to FIG. 2 , it is a schematic flowchart of a service restoration method in the embodiment of the present application. This method may be applied to the DPU interface card 100 shown in FIG. 1 above, or may also be applied to other applicable DPU interface cards. The following uses the DPU interface card 100 shown in FIG. 1 as an example for description. The service recovery method shown in Figure 2 may specifically include:

S201: The DPU interface card 100 applies for a memory area 201 from the host 200, and acquires a first address identifier of the applied memory area 201.

Exemplarily, the first address identifier may be, for example, the first address of the memory area 201 , or may be other identification information for indicating the memory area 201 such as the last address, which is not limited in this embodiment.

In this embodiment, the DPU interface card 100 may apply for a section of memory area from the host 200 in advance, so that the applied memory area can be used later to store service information related to the service processed by the DPU interface card 100 . As an implementation example, the DPU chip 101 in the DPU interface card 100 can send a request to the host 200 to apply for a memory area, so that the host 200 can respond to the request and determine the memory area 201 of a preset size from the available memory area, Allocate it to the DPU interface card 100, and return the first address identifier of the memory area 201 to the DPU chip 101. In practical applications, after the DPU interface card 100 establishes a communication connection with the host 200, the host 200 actively allocates the memory area 201 for the DPU interface card 100, and sends its corresponding first address identifier to the DPU interface card 100, etc. .

S202: The DPU interface card 100 stores the first address identifier in the target storage area.

As an implementation example, the DPU interface card 100 may be configured with a target storage area, and when the DPU interface card 100 restarts software such as an operating system, the data stored in the target storage area may not be lost. For example, a complex programmable logic device (complex programmable logic device, CPLD) may be configured in the DPU interface card 100, and the DPU interface card 100 may store the first address identifier into the logic block in the CPLD (ie, the above-mentioned target storage area ). In practical applications, the target storage area can be realized by a non-volatile memory, such as an electrically alterable read only memory (EAROM), an electrically erasable programmable read only memory (electrically erasable programmable read only memory) , EEPROM), at least one implementation of flash memory. Alternatively, the target storage area may also be implemented by a volatile memory, such as at least one of static random access memory (static random access memory, SRAM) and dynamic random access memory (dynamic random access memory, DRAM). accomplish.

In other possible implementation examples, the target storage area can also be deployed outside the DPU interface card 100. For example, the DPU interface card 100 can be externally connected with a non-volatile memory or a volatile memory, so that the DPU interface card 100 can store the acquired first The address identification is written into an external non-volatile memory or a volatile memory.

It should be noted that the execution sequence of step 202 and step S203 is not limited in this embodiment. For example, in other embodiments, the DPU interface card 100 may also execute step S203 first, and then execute step S202, etc., or both The steps are executed simultaneously.

S203: During the process of assisting the host 200 in processing services, the DPU interface card 100 stores the service information generated by processing the service in the memory area 201 of the host 200 according to the first address identifier, and the service information is processed by the DPU interface card 100 Information generated by the IO sent by the processor of the host 200 .

After the configuration of the DPU interface card 100 is completed, the DPU interface card 100 can start to assist the host 200 to process one or more services. Taking processing a business as an example, the processor in the host 200 can send the input and output (input output, IO) corresponding to the business to the sending queue for storage (the number of IOs sent by the processor can be one or more), so that The DPU interface card 100 can read the IO from the sending queue of the host 200, and parse and execute the read IO, and the data obtained by parsing and executing the IO can be temporarily stored in the memory of the DPU interface card 100 . At the same time, the DPU interface card 100 can also store the IO-related information generated during the execution of the IO into the memory area 201 of the host 200 . Wherein, the IO-related information is business information, including, for example, IO execution stages, key states of IO execution, and the like. In this way, even if the business data stored in the internal memory of the DPU interface card 100 is lost, the subsequent DPU interface card 100 can also read business information such as the IO execution stage and the key state of IO execution from the memory area 201, and use the business information Continue to execute the IO to realize service recovery, and in this way, the DPU interface card 100 can also avoid re-execution of the IO, reducing service recovery delay.

In a further possible implementation manner, the DPU interface card 100 may store at least part of service information generated during the execution of the IO in the memory area 201, so as to reduce resource consumption of the DPU interface card 100 for processing services. For example, in the initial stage of IO execution, the DPU interface card 100 may not store the current execution stage of the IO and key states of IO execution in the memory in the host 200 . Correspondingly, if the service needs to be resumed based on the IO later, the DPU interface card 100 can re-execute the IO to resume the service. Since the DPU interface card 100 has not started to execute the IO or has just started the IO before restarting software such as the operating system, even if the DPU interface card 100 subsequently re-executes the IO in the process of resuming business, the DPU interface card 100 recovers The delay impact of processing services is also small. And when the stage of IO execution by DPU interface card 100 reaches the middle or late stage, DPU interface card 100 can store information such as the key state and IO execution stage of the IO execution in the memory area 201 of the host computer 200, like this, if subsequent needs based on this When the IO resumes processing services, the DPU interface card 100 can continue to execute IOs according to the information saved by the host 200 without re-executing the IOs, thereby reducing the service recovery delay.

Alternatively, the DPU interface card 100 may determine whether to store the service information generated by executing the IO into the memory area 201 according to the size of the IO. For example, when the IO size read by the DPU interface card 100 from the sending queue does not exceed the preset threshold, the DPU interface card 100 may not need to send the business information generated by processing the IO to the host 200 during the execution of the IO. stored in memory. In this way, even if the DPU interface card 100 restores the service by re-executing the IO, the cost to be paid is relatively small. And when the IO size read by the DPU interface card 100 from the sending queue exceeds the preset threshold, the DPU interface card 100 can store business information such as the key status of the IO execution and the IO execution stage in the memory of the host 200, so as to avoid the DPU interface The card 100 re-executes the IO to reduce service recovery delay. In actual application, the DPU interface card 100 can also comprehensively determine whether to send the service information generated by executing the IO to the memory of the host 200 for storage in consideration of the IO size, IO execution progress and other aspects.

In actual application, when the DPU interface card 100 is executing the IO, it can also send the IO execution result obtained during the IO execution to the memory of the host 200 for storage. In this way, when the IO is interrupted because the operating system of the DPU interface card 100 is restarted, the DPU interface card 100 can continue to process the IO from the interrupted position according to the IO execution result and the above-mentioned business information stored in the memory of the host computer 200, Therefore, the service recovery delay can be further reduced.

S204: When the conditions for restarting the software are met, the DPU interface card 100 restarts the software.

In this embodiment, the software in the DPU will be restarted when the restart condition is met, and the restart of the software will affect the processing or service recovery of the DPU interface card 100 . Exemplarily, the software may be, for example, the operating system in the DPU interface card 100, or may be other software. For ease of understanding, the software is specifically an operating system as an example for illustrative description below.

In actual application, the DPU interface card 100 may restart the operating system in some scenarios. As some examples, conditions to qualify for an operating system reboot could include the following:

Example 1: It is detected that the memory in the DPU interface card 100 fails.

During specific implementation, the DPU interface card 100 can sense in real time (or periodically) whether the memory in the DPU interface card 100 fails, such as sensing at least one row, column or bank failure in the memory and causing an uncorrectable error in data access ( uncorrected errors, UCE) etc. (or may be other failures), and report the location information of the memory failure to the memory supervision unit 104. The memory monitoring unit 104 may isolate or replace the failed memory part according to the location information of the failure, and trigger the soft reset unit 103 to perform a soft reset process. Then, the soft reset unit 103 can reset the hardware unit (such as the DPU chip 101 ) in the DPU interface card 100 and restart the operating system of the DPU interface card 100 .

Wherein, the soft reset unit can reset all hardware units in the DPU interface card 100, and at this time, the PCIe link between the DPU interface card 100 and the host 200 is disconnected. In another implementation, the soft reset unit 103 can reset hardware units other than the PCIe core (core), so that the PCIe core can continue to be connected to the host 200 because it is not reset, thereby maintaining the connection between the DPU interface card 100 and the host computer 200. The PCIe link between the hosts 200 is not disconnected. The PCIe core is used to establish a PCIe link with the host.

After the soft reset unit 103 completes the reset of the hardware unit, the microkernel unit 105 can initialize the hardware unit, and restart each service component in the system service component unit 102 to start each system service of the kernel, such as the kernel driver services, file system services, memory management services, and network protocol services.

It should be noted that, in the embodiment shown in Example 1, the DPU interface card 100 is detected to trigger the restart of the operating system as an example. In actual application, when other failures (such as DPU interface When there is an application running error in the card 100) and the operating system needs to be restarted, the DPU interface card 100 can also be triggered to restart the operating system.

Example 2: The operating system of the previous version of the operating system of the DPU interface card 100 is upgraded.

Specifically, the host 200 can generate an upgrade command for the previous version of the operating system of the DPU interface card 100, and send it to the DPU interface card 100, so that the DPU interface card 100 can perform an upgrade of the DPU interface according to the received upgrade command. The flow of the operating system of the card 100. For example, the DPU interface card 100 can read the new version of the operating system from the host computer 200 according to the upgrade instruction, and replace the previous version of the operating system of the DPU interface card 100 with a new version, and then the DPU interface card 100 can be in the After confirming that the version upgrade is complete, the operating system running the new version can be started.

Wherein, the host 200 can periodically issue upgrade instructions to realize the periodic update of the DPU interface card 100 operating system; or, the host 200 can generate corresponding upgrade instructions according to the user's upgrade operation for the DPU interface card 100 operating system And send it to the DPU interface card 100 and so on.

Example 3: An instruction to restart the operating system is received.

Specifically, the host computer 200 can generate a corresponding restart command according to the user's restart operation for the DPU interface card 100 operating system, and send it to the DPU interface card 100, so that the DPU interface card 100 can restart after receiving the restart command. , reboot and run the OS.

The above implementation of triggering the DPU interface card 100 to restart the operating system is only for some exemplary illustrations. In other embodiments, the DPU interface card 100 may also restart the operating system when other possible conditions are met, for example, when the DPU interface The operating system of the card 100 can automatically trigger the restart of the operating system when an error occurs during operation; Restarting other software in the DPU interface card 100 is implemented, which is not limited in this embodiment.

After the DPU interface card 100 restarts the operating system, the service data temporarily stored in the internal memory of the DPU interface card 100 is lost, so the DPU interface card 100 may interrupt processing services due to the loss of service data. For this reason, in this embodiment, the DPU interface card 100 continues to execute the following steps to realize the recovery and processing of interrupted services.

S205: After restarting the software, the DPU interface card 100 acquires the service information stored in the memory area 201.

In a possible implementation manner, after restarting the operating system or other software, the DPU chip 101 in the DPU interface card 100 can obtain the first address identifier from the target storage area, and the first address identifier (for example, the memory area 201 first address) is used to indicate the memory area 201 that the DPU interface card 100 pre-applied to the host 200, so that the DPU chip 101 can access the memory area 201 of the host according to the first address identification, and read the business information stored in the memory area 201 .

S206: The DPU interface card 100 restores the service according to the acquired service information.

Exemplarily, specifically, the acquired business information may specifically be the data generated when the DPU interface card 100 executes the unfinished IO, so that the DPU chip 101101 can acquire the unfinished IO from the host 200, and according to the memory area 201 The current execution stage of the IO and the key state of the IO execution are stored in the IO, and the IO is continued to be executed from the current execution stage, so as to realize the restoration of the processing of the business. Further, the DPU chip 101 may continue to execute the IO from the interrupted position of the current execution stage according to the IO execution result stored in the memory area 201 . Wherein, for the IO that has not been executed by the DPU interface card 100 before the operating system is restarted or the IO that has just been executed, the memory area 201 may not record the relevant information of the IO. At this time, the DPU chip 101 can directly re-execute the IO.

In a further possible implementation, the business information stored in the memory area 201 may be the information of some unfinished IOs, such as the execution stage of the IO and the corresponding execution results in the execution stage, etc., while the DPU interface card 100 Another part of IO that has been executed but not completed may not store relevant information of this part of IO in the memory area 201 . Therefore, after the DPU interface card 100 acquires an unfinished IO from the sending queue of the host 200, it may check whether the service information stored in the memory area 201 includes information related to the IO. And, if the information related to this IO is found, then the DPU interface card 100 can continue to execute the IO from the interrupted position according to the information found; Execute the IO.

In this way, even if the software of the DPU interface card 100 is restarted due to reasons such as a fault in the memory of the DPU interface card 100 or a software upgrade, causing service interruption, the DPU interface card 100 can quickly restore the service through the service information stored in the memory of the host computer 200. Reduce impact on business.

Further, when resetting the DPU interface card 100 hardware unit, the DPU interface card 100 may not reset the PCIe core, so that the DPU interface card 100 can continue to maintain the connection of the PCIe link with the host computer 200 through the PCIe core, thereby realizing the DPU interface card The PCIe link between 100 and host 200 is not disconnected. In this way, when repairing the faulty memory of the DPU interface card 100 or upgrading the operating system of the DPU interface card 100, the host 200 may not perceive the fault state of the DPU interface card 100 and the change of the upgrade state, thereby reducing the impact on the host 200. Moreover, the service recovery process of the DPU interface card 100 has relatively low requirements on hardware and operating systems, and can be compatible with various types of computing devices and operating systems, thereby improving the universality of solution implementation.

In the above embodiment, the DPU interface card 100 can directly trigger the soft reset unit 103 to reset the hardware unit and restart software such as the operating system when a memory failure occurs. Fault recovery is realized by starting some service components in the kernel of the operating system. In a possible implementation manner, when the DPU interface card 100 detects that there is a memory failure, it can further determine whether the failed memory meets the preset condition, and when the failed memory meets the preset condition, the DPU interface card 100 Restart the service component using the faulty memory in the kernel of the operating system, and determine the IO corresponding to the data stored in the faulty memory, so as to resume business operation by re-executing the IO, or the DPU interface card 100 can be based on the data stored in the memory area 201. The relevant information of the IO continues to execute the IO to resume business operation, etc. In this way, the DPU interface card 100 can implement fault recovery without restarting the entire operating system and reconfiguring the DPU interface card 100 , thereby reducing the cost of fault recovery. And when the faulty memory does not meet the preset condition, the DPU interface card 100 can resume the interrupted service through the method of the embodiment shown in FIG. 2 . In this way, different processing methods can be used to repair the fault according to the fault condition of the memory of the DPU interface card 100 , and the flexibility of the DPU interface card 100 to repair the faulty memory can be improved.

As some implementation examples, the preset condition that the faulty memory satisfies may specifically be that the size of the faulty memory does not exceed the preset size, such as the number of faulty rows (or columns) in the memory does not exceed the preset number of rows (or preset Set the number of columns), etc. At this time, the faulty memory portion has relatively little impact on the DPU interface card 100 , therefore, the DPU interface card 100 can implement fault recovery without restarting the entire operating system.

Or, the preset condition that the faulty memory satisfies may specifically be that the system component using the faulty memory is a preset system component, so that when the faulty memory affects a specific system component, therefore, the DPU interface card 100 can isolate or replace a faulty portion of memory and restart that portion of the system components to effectuate fault recovery.

Alternatively, the preset condition that the faulty memory satisfies may specifically be that the number of system components using the faulty memory does not exceed a preset number. At this point, the faulty memory only affects a small number of service components, but does not affect the rest of the service components. Therefore, the DPU interface card 100 can restart the affected service components without restarting the entire operating system (or other software) and all service components for fault recovery. In actual application, the preset condition satisfied by the faulty memory may also be other conditions, which are not limited in this embodiment.

In addition, in an actual application scenario, the DPU interface card 100 may complete corresponding configuration in advance, so as to realize normal communication between the DPU interface card 100 and the host 200 . For example, the DPU interface card 100 and the host 200 can be pre-configured to have a unified data communication format, communication protocol version, command parsing rules, etc., and the DPU interface card 100 can be configured with a sending queue (SQ) and a completion queue (CQ) in the host memory , where the sending queue is used to store at least one IO sent by the processor in the host 200 to the DPU interface card 100 for processing services, and the receiving queue is used to store the execution result of the IO fed back by the DPU interface card 100 . Since the DPU interface card 100 may lose the original configuration of the DPU interface card 100 after restarting the software, therefore, in a further possible implementation, the DPU interface card 100 may also use the DPU interface card 100 after obtaining the first address identifier. The interface card 100 stores configuration information for configuring the DPU interface card 100 in the memory area 201 according to the first address identifier. Wherein, the DPU interface card 100 may be manually configured by a technician in advance, so that the DPU interface card 100 may generate a corresponding configuration file based on the configuration operation of the technician and send it to the memory area 201 . Or, after the DPU interface card 100 establishes a communication connection with the host computer 200, the configuration file is generated by the host computer 200, and the configuration file is automatically used to configure the DPU interface card 100, etc., and the configuration file is written by the host computer 200. In the memory area 201, this embodiment does not limit it.

In this way, when the software is restarted (as the DPU interface card 100 restarts the software due to a memory failure or a software version upgrade, etc.) Obtain the configuration information, and use the configuration information to reconfigure the DPU interface card 100, such as reconfiguring the format of the communication data of the DPU interface card 100, the instruction analysis rules, etc., to ensure normal communication between the DPU interface card 100 and the host computer 200. Moreover, the configuration information also includes the second address identifier of the send queue and the third address identifier of the completion queue in the host 200, so that after the DPU interface card 100 is configured, the DPU interface card 100 can access the host 200 according to the second address identifier. The sending queue, and obtain the IO that has not been executed by the DPU interface card 100 from the sending queue, and the IO is delivered to the sending queue by the processor in the DPU interface card 100. Correspondingly, the final result of the IO performed by the DPU interface card 100 can be sent to the completion queue of the host 200 according to the third address identifier, so as to resume and process the interrupted service.

The service recovery method provided by the embodiment of the present application is introduced above with reference to FIG. 1 and FIG. 2 . Next, the functions of the data processing unit DPU device provided by the embodiment of the present application and the computing equipment for implementing the data processing unit are introduced in conjunction with the accompanying drawings.

Referring to FIG. 3 , it shows a schematic structural diagram of a data processing unit DPU device. Wherein, the DPU device 300 shown in FIG. 3 is coupled with a host (not shown in FIG. 3 ), and the DPU device 300 includes:

The acquiring module 301 is configured to acquire the business information stored in the memory of the host after the software of the DPU device 300 is restarted, the business information is that the DPU device 300 processes the The information generated by the input and output IO sent by the processor of the host;

A recovery module 302, configured to recover services according to the service information.

Optionally, the restarted software may be, for example, an operating system, a component in the kernel of the operating system, or other software except the operating system.

In a possible implementation manner, the DPU device 300 is coupled with the host based on the peripheral component interconnection PCIe bus, and, during the software restart process, the DPU device 300 is connected to the host The PCIe link between them is not disconnected.

In a possible implementation manner, the restart of the software of the DPU device 300 is triggered by a failure of the DPU device 300 or is triggered by a software upgrade of a previous version of the software of the DPU device 300 .

In a possible implementation manner, the DPU device 300 further includes the startup module 303, configured to restart the software when the memory of the DPU device 300 fails.

In a possible implementation manner, the startup module 303 is configured to restart the operating system of the DPU 300 when the memory of the DPU device 300 fails and the failed memory does not meet a preset condition.

In a possible implementation manner, the startup module 303 is further configured to restart the operation of the DPU device 300 when the memory of the DPU device 300 fails and the failed memory satisfies the preset condition A service component in the system's kernel that uses the failed memory.

In a possible implementation manner, the obtaining module 301 is configured to:

Obtaining a first address identifier, where the first address identifier is used to identify a memory area in the host, where the memory area stores the service information;

Accessing the memory area according to the first address identifier to obtain the service information.

In a possible implementation manner, the DPU device 300 further includes:

An application module 304, configured to apply for the memory area from the host and obtain the first address identifier of the memory area before the software is restarted;

The storage module 305 is configured to store the service information in the memory area according to the first address identifier.

In a possible implementation manner, the DPU device 300 may also obtain configuration information from the memory area, and the configuration information is used to configure the DPU device 300, where the configuration information includes the send queue in the memory of the host The second address identifier and the third address identifier of the completion queue, the sending queue is used to store the IO, and the completion queue is used to store the execution result of the IO by the DPU device 300 .

Since the DPU device 300 shown in FIG. 3 can implement the method shown in FIG. 2, the specific implementation of the DPU device 300 shown in FIG. 3 and its technical effects can be referred to the relevant descriptions in the foregoing embodiments. , which will not be described here.

The DPU device 300 shown in FIG. 3 may be implemented by an ASIC, or by a general-purpose CPU and an ASIC, or by software, or by a combination of software and hardware, which is not limited in this embodiment of the present invention.

Figure 4 provides a computing device. Wherein, as shown in FIG. 4 , the computing device 400 includes a host 401 and a DPU interface card 402 , and the host 401 and the DPU interface card 402 are coupled through a bus 403 . The bus 403 may be a peripheral component interconnect standard (peripheral component interconnect, PCI) bus or an extended industry standard architecture (extended industry standard architecture, EISA) bus, etc. The bus can be divided into address bus, data bus, control bus and so on. For ease of representation, only one thick line is used in FIG. 4 , but it does not mean that there is only one bus or one type of bus.

Wherein, the host 401 includes a memory 4011 and a processor 4012 , and the memory 4011 and the processor 4012 may be coupled through a bus 4013 .

The bus 4013 can be a PCI bus or an EISA bus, etc. The bus can be divided into address bus, data bus, control bus and so on. For ease of representation, only one thick line is used in FIG. 4 , but it does not mean that there is only one bus or one type of bus.

The processor 4012 can be a central processing unit (central processing unit, CPU), a graphics processing unit (graphics processing unit, GPU), a microprocessor (micro processor, MP) or a digital signal processor (digital signal processor, DSP) etc. Any one or more of them.

The memory 4011 may be implemented by a memory, and the memory may include a volatile memory (volatile memory), such as a random access memory (random access memory, RAM). And, the memory can also include non-volatile memory (non-volatile memory), such as read-only memory (read-only memory, ROM), flash memory, mechanical hard disk (hard drive drive, HDD) or solid state disk (solid state drive, SSD).

The DPU interface card 402 may specifically be used to implement the method executed by the DPU interface card 100 in the embodiment shown in FIG. 2 above.

Computing device 400 may be a server, a storage array, or a distributed storage system.

In addition, the embodiment of the present application also provides a DPU interface card. Referring to FIG. 5, FIG. 5 shows a schematic structural diagram of a DPU interface card. As shown in Figure 5, the DPU interface card 500 includes a printed circuit board 501, an interface 502 and a DPU chip 503, and the DPU interface card 500 communicates with the host through the interface 502, and the interface 502 and the DPU chip 503 are installed on the printed circuit board. On the circuit board 501 ; the interface 502 and the DPU chip 503 can communicate through lines on the printed circuit board, or through cable communication, or bus communication, or the interface 502 and the DPU chip 503 are integrated together. For example, an implementation in which the interface 502 and the DPU chip 503 are integrated is packaged in one chip. Wherein, the DPU interface card 500 is used to implement the service recovery method performed by the DPU interface card 100 in the above embodiment shown in FIG. 2 . Correspondingly, for the specific implementation of the printed circuit board 501 , the interface 502 and the DPU chip 503 , reference may be made to the printed circuit board 1011 , the interface 1012 and the DPU chip 101 in the foregoing embodiments, and details are not repeated here.

The embodiment of the present application also provides a DPU chip. Referring to FIG. 6, FIG. 6 shows a schematic structural diagram of a DPU chip. As shown in Figure 6, the DPU chip 600 is applied to a DPU interface card (not shown in Figure 6) coupled with the host, such as the DPU interface card 100 in the foregoing embodiment; the DPU chip 600 includes an acquisition circuit 601 and a processing Circuit 602, wherein the acquisition circuit 601 is used to realize the function of the DPU chip 600 to acquire data, such as acquiring the business information stored in the memory of the host, the business information is before the software on the DPU interface card restarts, the processing circuit 602 Process the information generated by the input and output IO sent by the processor of the host; the processing circuit 602 is used to realize the data processing function of the DPU chip 600, such as recovering services according to the service information obtained by the obtaining circuit 601. During specific implementation, the DPU chip 600 may be an application specific integrated circuit (ASIC).

Wherein, the acquisition circuit 601 and the processing circuit 602 cooperate with each other, and can be used to implement the service recovery method executed by the DPU chip 101 in the DPU interface card 100 in the embodiment shown in FIG. 2 above.

The embodiment of the present application also provides a computer-readable storage medium. The computer-readable storage medium may be any available medium that a computing device can store, or a data storage device such as a data center that includes one or more available media. The available media may be magnetic media (eg, floppy disk, hard disk, magnetic tape), optical media (eg, DVD), or semiconductor media (eg, solid state hard disk), etc. The computer-readable storage medium includes instructions, and the instructions instruct a computing device to execute the above service recovery method.

The embodiment of the present application also provides a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on the computing device, the processes or functions according to the embodiments of the present application will be generated in whole or in part.

The computer instructions may be stored in or transmitted from one computer-readable storage medium to another computer-readable storage medium, e.g. (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wirelessly (such as infrared, wireless, microwave, etc.) to another website site, computer or data center.

The computer program product may be a software installation package, and if any of the aforementioned service recovery methods needs to be used, the computer program product may be downloaded and executed on the computing device.

The description of the process or structure corresponding to each of the above drawings has its own emphasis. For the part that is not described in detail in a certain process or structure, you can refer to the relevant description of other processes or structures.

Claims

A service recovery method, characterized in that the method comprises:

After the software of the data processing unit DPU interface card restarts, the DPU interface card obtains the service information stored in the internal memory of the host, and the service information is before the software restarts, the DPU interface card processes the host information generated by requests sent by the processor;

The DPU interface card restores services according to the service information.
The method according to claim 1, wherein the DPU interface card is coupled with the host based on the shortcut peripheral component interconnection PCIe bus, and, in the process of restarting the software, the DPU interface card The PCIe link with the host is not disconnected.
The method according to claim 1 or 2, wherein the restart of the software of the DPU interface card is triggered by a failure of the DPU interface card or by the software of the previous version of the software of the DPU interface card Upgrade trigger.
The method according to claim 3, wherein the DPU interface card restarts the software, comprising:

When the memory of the DPU interface card fails, the DPU interface card restarts the software.
The method according to claim 4, wherein when the internal memory of the DPU interface card fails, the DPU interface card restarts the software, including:

When the memory of the DPU interface card fails and the failed memory does not meet the preset condition, the DPU interface card restarts the operating system of the DPU interface card.
The method according to claim 4, characterized in that the method further comprises:

When the memory of the DPU interface card fails and the failed memory satisfies a preset condition, the DPU interface card restarts the service component using the failed memory in the kernel of the operating system of the DPU interface card.
The method according to any one of claims 1 to 6, wherein the DPU interface card acquires service information stored in the memory of the host, including:

The DPU interface card acquires a first address identifier, the first address identifier is used to identify a memory area in the host, and the memory area stores the service information;

The DPU interface card accesses the memory area according to the first address identifier to obtain the service information.
The method according to claim 7, wherein before the software is restarted, the method further comprises:

The DPU interface card applies for the memory area from the host, and obtains the first address identifier of the memory area;

The DPU interface card stores the service information in the memory area according to the first address identifier.
A kind of data processing unit DPU device is characterized in that, described DPU device comprises:

An acquisition module, configured to acquire the service information stored in the memory of the host after the software of the DPU device is restarted;

A recovery module, configured to recover services according to the service information; the service information is information generated by the DPU device processing the request sent by the processor of the host before the software is restarted.
The DPU device according to claim 10, characterized in that, the DPU device is coupled with the host based on the express peripheral component interconnection PCIe bus, and, in the process of restarting the software, the DPU device is connected with the The PCIe link between the hosts is not disconnected.
The DPU device according to claim 9 or 10, wherein the restart of the software of the DPU device is triggered by a failure of the DPU device or is triggered by a software upgrade of a previous version of the software of the DPU device .
The DPU device according to claim 11, characterized in that, the DPU device further comprises a startup module, configured to restart the software when the memory of the DPU device fails.
The DPU device according to claim 12, wherein the startup module is configured to restart the operation of the DPU device when the memory of the DPU device fails and the failed memory does not meet the preset conditions system.
The DPU device according to claim 12, wherein the startup module is also used to restart the operation of the DPU device when the memory of the DPU device fails and the failed memory meets a preset condition The service component in the system's kernel that uses the failed memory.
The DPU device according to any one of claims 9 to 14, wherein the acquisition module is configured to:

Obtaining a first address identifier, where the first address identifier is used to identify a memory area in the host, where the memory area stores the service information;

Accessing the memory area according to the first address identifier to obtain the service information.
The DPU device according to claim 15, wherein the DPU device further comprises:

An application module, configured to apply to the host for the memory area and obtain the first address identifier of the memory area before the software is restarted;

A storage module, configured to store the service information in the memory area according to the first address identifier.
A computing device, characterized in that the computing device includes a host and a data processing unit DPU interface card, and the host includes a memory and a processor;

The DPU interface card is used to obtain the service information stored in the memory of the host after the software of the DPU interface card is restarted, and restore the service according to the service information; the service information is after the software restarts Before, the DPU interface card processes the information generated by the request sent by the processor of the host.
A data processing unit DPU interface card is characterized in that it includes a printed circuit board, an interface and a data processing unit DPU chip, the DPU interface card communicates with the host through the interface, and the interface and the DPU chip are installed on the On the printed circuit board, the DPU chip is used to obtain the service information stored in the memory of the host after the software of the DPU interface card is restarted, and restore the service according to the service information; the service information is in the Before the software restarts, the DPU chip processes the information generated by the request sent by the processor of the host.
A data processing unit DPU chip, characterized in that, the DPU chip is applied to a DPU interface card, and the DPU interface card is coupled with a host; the DPU chip includes an acquisition circuit and a processing circuit;

Wherein, the obtaining circuit is used to obtain the service information stored in the memory of the host after the software of the DPU interface card is restarted;

The processing circuit is configured to restore the service according to the service information; the service information is information generated by the processing circuit processing the request sent by the processor of the host before the software is restarted.