CN114217943A - Initialization method of communication support environment - Google Patents
Initialization method of communication support environment Download PDFInfo
- Publication number
- CN114217943A CN114217943A CN202110398330.0A CN202110398330A CN114217943A CN 114217943 A CN114217943 A CN 114217943A CN 202110398330 A CN202110398330 A CN 202110398330A CN 114217943 A CN114217943 A CN 114217943A
- Authority
- CN
- China
- Prior art keywords
- resource
- sid
- file
- mapping
- per
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000004891 communication Methods 0.000 title claims abstract description 62
- 238000011423 initialization method Methods 0.000 title description 6
- 238000000034 method Methods 0.000 claims abstract description 134
- 230000008569 process Effects 0.000 claims abstract description 119
- 230000015654 memory Effects 0.000 claims description 60
- 238000013507 mapping Methods 0.000 claims description 42
- 238000007726 management method Methods 0.000 claims description 18
- 230000006870 function Effects 0.000 claims description 8
- 238000012367 process mapping Methods 0.000 claims description 6
- FFBHFFJDDLITSX-UHFFFAOYSA-N benzyl N-[2-hydroxy-4-(3-oxomorpholin-4-yl)phenyl]carbamate Chemical compound OC1=C(NC(=O)OCC2=CC=CC=C2)C=CC(=C1)N1CCOCC1=O FFBHFFJDDLITSX-UHFFFAOYSA-N 0.000 claims description 4
- 239000000284 extract Substances 0.000 abstract 1
- 239000000872 buffer Substances 0.000 description 4
- 239000002699 waste material Substances 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012384 transportation and delivery Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5011—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resources being hardware resources other than CPUs, Servers and Terminals
- G06F9/5016—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resources being hardware resources other than CPUs, Servers and Terminals the resource being the memory
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer And Data Communications (AREA)
Abstract
本发明公开一种通信支撑环境的初始化方法，包括以下步骤：S1、调用作业管理接口，解析本道作业的资源总数tasksize、各进程在作业中的初始资源序号Sid、资源名列表nodeidlist、资源片内唯一标识mpeidlist、pid列表pidlist、网络设备标识列表guidlist，并按资源列表自然序排列；S2、根据各进程的初始资源序号Sid计算Sid%cg_per_process，从作业系统返回信息中提取满足模cg_per_process为0的进程的相关信息。本发明有助于支撑用户求解更大的问题规模、追求更高的性能，且实现简便，不影响消息性能，用户调用消息接口时无额外开销。
The invention discloses a method for initializing a communication support environment, comprising the following steps: S1. Invoke a job management interface to analyze the total number of resources taskssize of the job in the job, the initial resource serial number Sid of each process in the job, a resource name list nodeidlist, and a resource slice. Uniquely identifies mpeidlist, pid list pidlist, and network device identification list guidlist, and arranges them in the natural order of the resource list; S2, calculates Sid%cg_per_process according to the initial resource serial number Sid of each process, and extracts from the information returned by the operating system that the modulus cg_per_process is 0 Information about the process. The present invention helps support users to solve larger problem scales and pursue higher performance, and is easy to implement, does not affect message performance, and has no extra overhead when users call the message interface.
Description
Technical Field
The invention relates to an initialization method of a communication support environment, and belongs to the technical field of high-performance computing.
Background
In a large-scale cluster environment, scheduling, allocation and management of system resources are generally completed by using a job management system. The user calls the operation management interface according to the resource requirement of the parallel program, the operation system applies for resources according to the requirement, and submits the resource list and necessary information to the communication support environment in an agreed data structure, and the communication support environment completes initialization according to the information provided by the operation system and prepares for interprocess communication.
Interprocess communication needs to be based on a communication context consisting of a series of hardware resources and software resources. The hardware resources refer to communication queue resources, including physical resources such as a sending queue, a receiving queue and a completion queue, and each type of physical resources needs to occupy a corresponding memory space; meanwhile, a series of data structures need to be maintained in both the upper layer communication interface and the bottom layer message library so as to realize resource management service and achieve efficient management and utilization of the physical resources. In addition, some buffers are required to be applied for storing user message requests and realizing the pre-delivery of short messages, so that a certain amount of memory space is required for a communication channel between any two processes, and the space is increased along with the enlargement of the process size.
Many processors facing the high-performance computing field support a certain range of shared memories, and based on the shared memories, different computing resources in the processors can realize data interaction. The reasonable use of the shared memory can effectively improve the data multiplexing degree in the processor, reduce the communication requirement of the whole chip and further improve the performance. Based on the above considerations, in some practical applications, not every process needs to participate in message transmission, but several processes use a shared memory to realize cooperative operation based on the actual task division requirement, and then a small number of processes complete data interaction.
The communication support environment is used as a common support environment on a large cluster, and needs to meet the requirements of different users and support various communication requirements which may appear and meet the specification. Therefore, the communication support environment usually requires that each process of the parallel program must call an initialization interface to prepare software and hardware for possible message transmission, during the initialization process, the upper layer communication interface and the bottom layer message library need to create and maintain a plurality of data structures and buffer spaces to realize and meet the communication requirements of the parallel program, and the memory overhead of the part is considerable, and the number of the split pins is increased continuously with the increase of the number of the processes. However, not every process in the parallel program has a communication requirement, and especially for an architecture with a shared memory, each processor only needs to send and receive messages by a respective process, and initialization of the other processes only causes waste of the memory and enlargement of the process scale, and has no practical significance. This will cause the process without actual communication requirement in the application to be forced to apply for physical resources, occupy memory and manage data structure, enlarge the process scale, and cause the waste of memory space and physical resources. Some applications have a large demand for memory capacity, and a larger memory space can help the applications to realize the solution of a larger problem scale or the improvement of performance.
Disclosure of Invention
The invention aims to provide an initialization method of a communication support environment, which is beneficial to supporting a user to solve a larger problem scale and pursue higher performance, is simple and convenient to realize, does not influence the message performance, and has no extra overhead when the user calls a message interface.
In order to achieve the purpose, the invention adopts the technical scheme that: the method for initializing the communication support environment is provided, wherein cg _ per _ process or mapping _ file is set through an environment variable, and is used for a process list marked with communication requirements: the former indicates that the process with the (process number% cg _ per _ process) of 0 has a communication demand, and the other processes have no communication demand; the latter represents a process mapping file, and the content of the file is a resource serial number Sid with communication requirements;
the method comprises the following steps:
s1, calling a job management interface, analyzing the total number of resources tasksize of the job, the initial resource sequence number Sid of each process in the job, a resource name list nodeidlist, unique identifiers mpeldist in the resource slice, a pid list pidlist and a network equipment identifier list guidlist, and arranging according to the natural sequence of the resource list, wherein the ith element of each list is information of the ith resource;
s2, for the scenario with cg _ per _ process set, calculating Sid% cg _ per _ process according to the initial resource sequence Sid of each process, and extracting the relevant information of the process satisfying the modulus cg _ per _ process as 0 from the return information of the operating system, the specific operations are as follows:
s21, skipping initialization when the resource serial number Sid% cg _ per _ process is not 0, and directly returning an initialization completion identifier;
s22, reducing the total number tasksize of the resources to 1/cg _ per _ process, and setting Sid as Sid/cg _ per _ process;
s23, nodeidlist, mpeldlist, pidlist and guidllist structures adopt an in-place merging mode, an array space is multiplexed, an effective element is extracted every cg _ per _ process-1 elements, and the effective element is stored in the position where the tasksize elements are located in front of the original structures;
s3, for the scene with mapping _ file, calculating the process number after logical mapping according to the initial resource serial number Sid of each process, each process reading the mapping file, and storing the mapping file in the memory file system through the file system and the operating system interface, the specific operation is as follows:
s31, the resource serial number Sid is not in mapping _ file, skipping initialization, and directly returning to the initialization completion identifier;
s32, modifying the total number of the mapped processes, tasksize, according to the number of the mapping _ file items, and setting the natural sequence number of the item of the process with the initial resource sequence number Sid in the mapping _ file as the Sid value after logical mapping;
s33, extracting elements of nodeidlist, mpeldlist, pidlist and guidllist structures in sequence according to mapping _ file entry content;
s34, covering the front tasksize elements of the original information lists by the extracted information lists;
s4, setting an environment variable memory _ balance, setting a memory management function pointer according to a variable value, and being used for specifying that a memory space initialized by a communication support environment is mapped to a process private memory space or a processor shared memory space, and balancing memory overhead of each resource in the processor when the memory space is mapped to the shared memory;
and S5, initializing the subsequent communication support environment by using the appointed memory management function pointer, the updated tasksize and each information list, wherein the process logic number adopted in the communication is the serial number of each resource in the updated information list.
Due to the application of the technical scheme, compared with the prior art, the invention has the following advantages:
the invention can effectively reduce the process scale and memory waste, is beneficial to improving the expandability of a communication support environment, and is also beneficial to supporting users to solve larger problem scale and pursue higher performance; the method is simple and convenient to implement, the message performance is not influenced, and no extra overhead is caused when a user calls a message interface.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Detailed Description
Example (b): the invention provides an initialization method of a communication support environment, which sets cg _ per _ process or mapping _ file through an environment variable and is used for marking a process list with communication requirements: the former indicates that the process with the (process number% cg _ per _ process) of 0 has a communication demand, and the other processes have no communication demand; the latter represents a process mapping file, and the content of the file is a resource serial number Sid with communication requirements;
the method comprises the following steps:
s1, calling a job management interface, analyzing the total number of resources tasksize of the job, the initial resource sequence number Sid of each process in the job, a resource name list nodeidlist, unique identifiers mpeldist in the resource slice, a pid list pidlist and a network equipment identifier list guidlist, and arranging according to the natural sequence of the resource list, wherein the ith element of each list is information of the ith resource;
s2, for the scenario with cg _ per _ process set, calculating Sid% cg _ per _ process according to the initial resource sequence Sid of each process, and extracting the relevant information of the process satisfying the modulus cg _ per _ process as 0 from the return information of the operating system, the specific operations are as follows:
s21, skipping initialization when the resource serial number Sid% cg _ per _ process is not 0, and directly returning an initialization completion identifier;
s22, reducing the total number tasksize of the resources to 1/cg _ per _ process, and setting Sid as Sid/cg _ per _ process;
s23, nodeidlist, mpeldlist, pidlist and guidllist structures adopt an in-place merging mode, an array space is multiplexed, an effective element is extracted every cg _ per _ process-1 elements, and the effective element is stored in the position where the tasksize elements are located in front of the original structures;
s3, for the scene with mapping _ file, calculating the process number after logical mapping according to the initial resource serial number Sid of each process, each process reading the mapping file, and in order to reduce the pressure of large-scale concurrent reading on the global file system, storing the mapping file in the memory file system through the file system and the operating system interface, the specific operation is as follows:
s31, the resource serial number Sid is not in mapping _ file, skipping initialization, and directly returning to the initialization completion identifier;
s32, modifying the total number of the mapped processes, tasksize, according to the number of the mapping _ file items, and setting the natural sequence number of the item of the process with the initial resource sequence number Sid in the mapping _ file as the Sid value after logical mapping;
s33, extracting elements of nodeidlist, mpeldlist, pidlist and guidllist structures in sequence according to mapping _ file entry content;
s34, covering the front tasksize elements of the original information lists by the extracted information lists;
s4, setting an environment variable memory _ balance, setting a memory management function pointer according to a variable value, and being used for specifying that a memory space initialized by a communication support environment is mapped to a process private memory space or a processor shared memory space, and balancing memory overhead of each resource in the processor when the memory space is mapped to the shared memory;
and S5, initializing the subsequent communication support environment by using the appointed memory management function pointer, the updated tasksize and each information list, wherein the process logic number adopted in the communication is the serial number of each resource in the updated information list.
The above embodiments are further explained as follows:
the invention modifies the structure array submitted to the message support environment by the operating system, and continues the initialization process of the standard message support environment as much as possible, thereby realizing the balance of the on-demand initialization and the memory overhead by the change as little as possible.
Starting with the communication support environment initialization process modification, the resource information submitted to the communication support environment by the operating system is adjusted according to the environment variables and the configuration file, and the initialization of the communication support environment is completed as required, so that the communication requirement of a user program can be ensured, and the memory overhead can be effectively reduced; in the use aspect, a user only needs to specify environment variables and configuration files; when the number of the configuration file entries is consistent with the total number of the resources, the method and the system can realize the process mapping customized by the user.
The invention still realizes the initialization of each process under the default condition, and when the user confirms that some processes have no communication requirements in the whole process of program operation, the communication support environment can complete the message initialization customized by the user according to the requirements so as to reduce the process scale and the memory overhead; and the memory space occupied by initialization can be mapped to the private memory space or the shared memory space of the default designated process according to the designation of the user, so that the memory overhead of each process is balanced.
The method comprises the following specific steps:
1. the cg _ per _ process or mapping _ file is set by an environment variable, the process list marked with communication requirements: the former indicates that the process with the (process number% cg _ per _ process) of 0 has a communication demand, and the other processes have no communication demand; the latter represents a process mapping file, and the content of the file is a resource serial number Sid with communication requirements.
2. And calling a job management interface, analyzing the total resource number tasksize, the resource name list nodeidlist, the unique identifier mpeldist in the resource slice, the pid list pidlist, the network equipment identifier list guidlist and the like of the job, arranging the contents according to the natural sequence of the resource list, wherein the ith element in each list is the information of the ith resource.
3. For the scenario of setting cg _ per _ process, Sid% cg _ per _ process is calculated according to the initial resource sequence Sid of each process, and the relevant information of the process satisfying the module cg _ per _ process of 0 is extracted from the return information of the operating system, specifically the following operations are performed:
a) and if the initial resource sequence number Sid% cg _ per _ process is not 0, skipping initialization and directly returning to the initialization completion identifier.
b) The total number of resources tasksize is reduced to 1/cg _ per _ process.
c) The method comprises the steps of multiplexing an array space by adopting an in-place merging mode for nodeidlist, mpeeidlist, pidlist, guidllist and other structural bodies, extracting an effective element every cg _ per _ process-1 element, and storing the effective element in the position where the tasksize elements are located in front of the original structural body.
4. For the scene of setting mapping _ file, the process number after logical mapping is calculated according to the initial resource serial number Sid of each process, each process reads the mapping file, and the mapping file can be stored in the memory file system through the file system and the operating system interface in order to reduce the pressure of large-scale concurrent reading on the global file system. The specific operation is as follows:
a) and the initial resource sequence number Sid does not skip the initialization in the mapping _ file, and directly returns an initialization completion identifier.
b) And modifying the total number of the mapped processes, namely tasksize, according to the number of the mapping _ file entries, and setting the entry natural sequence number of the process with the initial resource sequence number of Sid in the mapping _ file as the logically mapped Sid value.
c) And extracting elements of nodeidlist, mpeeidlist, pidlist and guidllist structures in order according to the mapping _ file entry content.
d) And covering the front tasksize elements of the original information lists by the extracted information lists.
5. Setting an environment variable memory _ balance, setting a memory management function pointer according to a variable value, and being used for appointing the memory space initialized by the communication support environment to be mapped to a process private memory space or a processor shared memory space.
6. And (4) finishing the initialization of the subsequent communication support environment by using the appointed memory management function pointer, the updated tasksize and each information list, wherein the process logic number adopted in the communication is the serial number of each resource in the updated information list.
Through the initialization process, the invention can find that other steps except the steps of adjusting the total number of resources and the information list according to the requirements are not modified for initialization, the modification amount is less, and the implementation is easy; when cg _ per _ process =1 or mapping _ file is a resource natural sequence number table, the initialization result is the same as the original message support environment, and the standard initialization mode is compatible; the process which is not involved in communication and is appointed by the user does not participate in the specific initialization step, so that various hardware resources and memory buffers do not need to be applied; the process size is effectively compressed, and the buffer space overhead related to the process size is reduced. In addition, when mapping _ file is used, if the given mapping _ file content is any arrangement of the resource sequence number table Sid, the patent can implement process mapping, that is, any designated corresponding relationship between the process logic sequence number and the resource sequence number.
When the initialization method of the communication support environment is adopted, the process scale can be effectively reduced, the memory waste is reduced, the expandability of the communication support environment is improved, the support of users for solving larger problem scale and pursuing higher performance is facilitated; the method is simple and convenient to implement, the message performance is not influenced, and no extra overhead is caused when a user calls a message interface.
To facilitate a better understanding of the invention, the terms used herein will be briefly explained as follows:
an operation management system: the method is used for scheduling and allocating system resources in a large cluster environment, and specifically, enough resources are divided from a cluster system according to the requirements of programs to be operated, and the application programs are submitted to the divided resources to be operated.
The above embodiments are merely illustrative of the technical ideas and features of the present invention, and the purpose thereof is to enable those skilled in the art to understand the contents of the present invention and implement the present invention, and not to limit the protection scope of the present invention. All equivalent changes and modifications made according to the spirit of the present invention should be covered within the protection scope of the present invention.
Claims (1)
1. A method for initializing a communication support environment, characterized by being based on the following configuration:
the cg _ per _ process or mapping _ file is set by an environment variable for the process list marked with communication requirements: the former indicates that the process with the (process number% cg _ per _ process) of 0 has a communication demand, and the other processes have no communication demand; the latter represents a process mapping file, and the content of the file is a resource serial number Sid with communication requirements;
the method comprises the following steps:
s1, calling a job management interface, analyzing the total number of resources tasksize of the job, the initial resource sequence number Sid of each process in the job, a resource name list nodeidlist, unique identifiers mpeldist in the resource slice, a pid list pidlist and a network equipment identifier list guidlist, and arranging according to the natural sequence of the resource list, wherein the ith element of each list is information of the ith resource;
s2, for the scenario with cg _ per _ process set, calculating Sid% cg _ per _ process according to the initial resource sequence Sid of each process, and extracting the relevant information of the process satisfying the modulus cg _ per _ process as 0 from the return information of the operating system, the specific operations are as follows:
s21, skipping initialization when the resource serial number Sid% cg _ per _ process is not 0, and directly returning an initialization completion identifier;
s22, reducing the total number tasksize of the resources to 1/cg _ per _ process, and setting Sid as Sid/cg _ per _ process;
s23, nodeidlist, mpeldlist, pidlist and guidllist structures adopt an in-place merging mode, an array space is multiplexed, an effective element is extracted every cg _ per _ process-1 elements, and the effective element is stored in the position where the tasksize elements are located in front of the original structures;
s3, for the scene with mapping _ file, calculating the process number after logical mapping according to the initial resource serial number Sid of each process, each process reading the mapping file, and storing the mapping file in the memory file system through the file system and the operating system interface, the specific operation is as follows:
s31, the resource serial number Sid is not in mapping _ file, skipping initialization, and directly returning to the initialization completion identifier;
s32, modifying the total number of the mapped processes, tasksize, according to the number of the mapping _ file items, and setting the natural sequence number of the item of the process with the initial resource sequence number Sid in the mapping _ file as the Sid value after logical mapping;
s33, extracting elements of nodeidlist, mpeldlist, pidlist and guidllist structures in sequence according to mapping _ file entry content;
s34, covering the front tasksize elements of the original information lists by the extracted information lists;
s4, setting an environment variable memory _ balance, setting a memory management function pointer according to a variable value, and being used for specifying that a memory space initialized by a communication support environment is mapped to a process private memory space or a processor shared memory space, and balancing memory overhead of each resource in the processor when the memory space is mapped to the shared memory;
and S5, initializing the subsequent communication support environment by using the appointed memory management function pointer, the updated tasksize and each information list, wherein the process logic number adopted in the communication is the serial number of each resource in the updated information list.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110398330.0A CN114217943A (en) | 2021-04-14 | 2021-04-14 | Initialization method of communication support environment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110398330.0A CN114217943A (en) | 2021-04-14 | 2021-04-14 | Initialization method of communication support environment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114217943A true CN114217943A (en) | 2022-03-22 |
Family
ID=80695906
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110398330.0A Pending CN114217943A (en) | 2021-04-14 | 2021-04-14 | Initialization method of communication support environment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114217943A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103605577A (en) * | 2013-12-04 | 2014-02-26 | 广州博冠信息科技有限公司 | Cross-process resource sharing method and equipment |
| CN107069950A (en) * | 2016-12-30 | 2017-08-18 | 国网天津市电力公司 | Automatic identifying method and system of a kind of distribution main website to terminal device |
| CN107368379A (en) * | 2017-07-14 | 2017-11-21 | 中南大学 | Towards EVP across GuestOS inter-process communication methods and system |
| US20180109432A1 (en) * | 2013-04-12 | 2018-04-19 | Groupon, Inc. | Centralized, scalable, resource monitoring system |
| CN112068808A (en) * | 2020-08-13 | 2020-12-11 | 中国航空无线电电子研究所 | General processing system for multi-bus data conversion of avionics system |
-
2021
- 2021-04-14 CN CN202110398330.0A patent/CN114217943A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180109432A1 (en) * | 2013-04-12 | 2018-04-19 | Groupon, Inc. | Centralized, scalable, resource monitoring system |
| CN103605577A (en) * | 2013-12-04 | 2014-02-26 | 广州博冠信息科技有限公司 | Cross-process resource sharing method and equipment |
| CN107069950A (en) * | 2016-12-30 | 2017-08-18 | 国网天津市电力公司 | Automatic identifying method and system of a kind of distribution main website to terminal device |
| CN107368379A (en) * | 2017-07-14 | 2017-11-21 | 中南大学 | Towards EVP across GuestOS inter-process communication methods and system |
| CN112068808A (en) * | 2020-08-13 | 2020-12-11 | 中国航空无线电电子研究所 | General processing system for multi-bus data conversion of avionics system |
Non-Patent Citations (1)
| Title |
|---|
| 曹宏嘉;卢宇彤;谢旻;周恩强;: "并行作业启动及其可扩展性分析", 计算机研究与发展, no. 08, 15 August 2013 (2013-08-15) * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11928508B2 (en) | Responding to application demand in a system that uses programmable logic components | |
| US9858241B2 (en) | System and method for supporting optimized buffer utilization for packet processing in a networking device | |
| JP5510556B2 (en) | Method and system for managing virtual machine storage space and physical hosts | |
| WO2021098182A1 (en) | Resource management method and apparatus, electronic device and storage medium | |
| CN108021449B (en) | Coroutine implementation method, terminal equipment and storage medium | |
| CN111176829B (en) | Flexible resource allocation of physical and virtual functions in virtualized processing systems | |
| US9063918B2 (en) | Determining a virtual interrupt source number from a physical interrupt source number | |
| CN103067468A (en) | Cloud scheduling method and system thereof | |
| CN112000463B (en) | A CUDA-based GPU resource allocation method, system, terminal and storage medium | |
| CN113419846B (en) | Resource allocation method and device, electronic equipment and computer readable storage medium | |
| CN110187970A (en) | A Distributed Big Data Parallel Computing Method Based on Hadoop MapReduce | |
| CN107273339A (en) | A kind of task processing method and device | |
| CN104156663B (en) | A kind of hardware virtual port and processor system | |
| CN119271421A (en) | Task migration method, device and storage medium | |
| CN114217943A (en) | Initialization method of communication support environment | |
| JP5043166B2 (en) | Computer system, data search method, and database management computer | |
| CN115756756A (en) | Video memory resource allocation method, device and equipment based on GPU virtualization technology | |
| KR20150101644A (en) | Executing performance enhancement method, executing performance enhancement apparatus and executing performance enhancement system for map-reduce programming model considering different processing type | |
| CN110399206B (en) | IDC virtualization scheduling energy-saving system based on cloud computing environment | |
| US20140237149A1 (en) | Sending a next request to a resource before a completion interrupt for a previous request | |
| CN113760798B (en) | RDMA device allocation method, computing device and storage medium | |
| CN116450298A (en) | GPU task fine granularity scheduling method and related device | |
| CN116932140A (en) | Data processing method based on cross-host resource and related products | |
| CN118656209A (en) | A microservice load balancing method, device, computer equipment and storage medium | |
| CN118012603A (en) | Resource scheduling method and related equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |