136 results about "Computer cluster" patented technology

The invention relates to the technical field of chips, in particular to a GPU topology sensing scheduling method, electronic equipment and a medium, and the method comprises the following steps: S1, receiving a target task processing request and analyzing the target task processing request to obtain a target GPU number M; s2, selecting a target computer node from a computer cluster based on M; s3, obtaining a current available GPU list of the target computer node and a topological relation between the current available GPUs; s4, generating all candidate allocation schemes based on the current available GPU list and the target GPU number M; s5, obtaining a weight AXn corresponding to the An and a weight BXn corresponding to the Bn based on the topological relation between the currently available GPUs; and step S6, determining the maximum An of the corresponding (AXn + BXn) as a target GPU for processing the target task. According to the method, the utilization rate maximization of the GPUs in the computer cluster and the communication efficiency among the GPUs are improved.

The invention discloses a collaborative resource scheduling method and system based on a programmable network. The method comprises the following steps: performing performance modeling on heterogeneous computing resource data in all computing resource data to obtain a uniform resource performance index; defining target computing resource data required by multi-dimensional resource demand services, and determining matching relationship information between each multi-dimensional resource demand service and the current computer cluster; in the deployment process, according to all the matching relation information, the correspondingly selected multi-dimensional resource demand service is deployed to each computer cluster; determining a shortest deployment path between the routing nodes in each computer cluster based on a Floyd algorithm; and in the actual scheduling process, executing the computing power arrangement scheme generated in the deployment process. According to the collaborative resource scheduling method and system based on the programmable network provided by the embodiment of the invention, the utilization rate of computing power resources in a wide area network can be improved, and a computing power scheduling strategy is dynamically generated.

In some embodiments, a computer cluster system comprises a plurality of nodes and a software package comprising a user interface and a kernel for interpreting program code instructions. In certain embodiments, a cluster node module is configured to communicate with the kernel and other cluster node modules. The cluster node module can accept instructions from the user interface and can interpret at least some of the instructions such that several cluster node modules in communication with one another and with a kernel can act as a computer cluster.

An end-cloud cooperative data mining method, comprising: determining a target text and a task configuration file according to a business requirement by a cloud end; encoding the target text by a text encoder to obtain a text feature; placing the text feature in the task configuration file and issuing it to a vehicle end together; encoding image data by a first picture encoder to obtain an image feature by the vehicle end; calculating a value of a similarity of the text feature and the image feature; determining a target picture according to the value of the similarity and the task configuration file; and uploading the target picture to the cloud end; wherein the first picture encoder is obtained by compressing and optimizing a second picture encoder; and the text encoder and the second picture encoder are two modules of a picture-text multimodal large model. The application is applied to an automatic driving shadow mode, can complete data mining of any interesting target text by using one large model, and does not need to design and develop detection rules for each type of interesting target. The picture encoder of the vehicle end can be continuously updated and optimized, and the model iteration efficiency is improved.

The embodiment of the invention provides a data transmission method and a computing node. The method is applied to a first computing node of a computer cluster, the computer cluster comprises a plurality of computing nodes, a storage node and computing interconnection CXL equipment, the plurality of computing nodes are connected with the storage node through the CXL equipment, and the plurality of computing nodes are further connected through Internet links. The method comprises the steps that a first computing node stores target data in a target storage position of a target storage node through CXL equipment, the first computing node is any computing node in a plurality of computing nodes, and the storage nodes comprise the target storage node; the first computing node sends a notification message to the second computing node through the internet link, the notification message comprises the node information of the target storage node and the target storage position, and the second computing node is any computing node except the first computing node in the plurality of computing nodes. Through the method, the data transmission efficiency in the computer cluster can be improved.

The invention discloses a data processing method, a network card, computer equipment and a computer cluster, and relates to the technical field of communication. The method comprises the following steps: dividing a plurality of computing nodes of a computer cluster into a first group and a second group, wherein each of the first group and the second group comprises at least two computing nodes; and establishing a full-connection corresponding relationship between the processes on the intra-group computing nodes of the first group and a corresponding relationship between the processes belonging to the same application on the inter-group computing nodes of the first group and the second group. The corresponding relation between the first process and the second process can be determined based on the corresponding relation, and then the data is processed according to the corresponding relation. Compared with a full-connection corresponding relation between running processes on different computing nodes, the method can reduce the number of connections so as to enlarge the computer cluster scale. Based on the corresponding relationship, the first process and the second process can directly communicate, and compared with dynamically creating and destroying the connection relationship between the processes, the computing overhead and the communication delay can be reduced.

According to one aspect, a method includes: receiving, by a node of a computer cluster, a command for a health status report, the node having a plurality applications and being one of a plurality of nodes of the cluster, the command indicating a class of a health checks to be executed and a scope; identifying, from a configuration, one or more health checks to be executed based on the indicated class; executing the one or more health checks on one or more of the applications and on one or more of the nodes of the cluster according to the indicated scope, where different ones of the health checks are configured to determine state of different ones of the applications; aggregating, by the node, results of the executed health checks; and outputting the aggregated results.

The utility model provides a geographic information computer cluster device, which belongs to the technical field of computers, and comprises a cabinet main body, a main node assembly and a sub node assembly, the cabinet main body comprises a frame and an outer box arranged at the outer side, the frame is respectively provided with a connecting frame and a mainboard fixing plate, and the connecting frame is provided with a heat dissipation support; the main node assembly comprises an operation main node, the operation main node is arranged on the front side of the main board fixing plate, and three fans A are arranged at the top of the main board fixing plate; the child node assembly comprises a node installation circuit board and operation child nodes, the node installation circuit board is arranged on the heat dissipation support, a plurality of node coamings are arranged on the node installation circuit board, the operation child nodes penetrate through the node coamings and are inserted into the node installation circuit board, and an internal circulation heat dissipation device is arranged in the child node assembly. The problems that an existing cabinet type server is large in size, heavy in weight, large in occupied space and not flexible enough in use scene are solved.

In some embodiments, a computer cluster system comprises a plurality of nodes and a software package comprising a user interface and a kernel for interpreting program code instructions. In certain embodiments, a cluster node module is configured to communicate with the kernel and other cluster node modules. The cluster node module can accept instructions from the user interface and can interpret at least some of the instructions such that several cluster node modules in communication with one another and with a kernel can act as a computer cluster.

In some embodiments, a computer cluster system comprises a plurality of nodes and a software package comprising a user interface and a kernel for interpreting program code instructions. In certain embodiments, a cluster node module is configured to communicate with the kernel and other cluster node modules. The cluster node module can accept instructions from the user interface and can interpret at least some of the instructions such that several cluster node modules in communication with one another and with a kernel can act as a computer cluster.

The invention relates to the technical field of simulation, in particular to a simulation method and system for expanding a meta-model of a virtual integrated system, equipment and a medium, by adopting the method provided by the invention, a mechanism for expanding the meta-model is plug-in, simulation modeling and operation scheduling interfaces are separated, and the coupling degree between plug-in function modules is reduced. The modeling scheduler and the engine scheduler can be integrated into the simulation modeling platform and the scheduling platform respectively, and after modeling work is completed in the simulation modeling platform, local simulation can be carried out, and distributed simulation in a computer cluster can also be carried out. The modeling scheduler and the engine scheduler perform interface interaction with the resolver and the actuator in the expansion plug-in respectively, the modeling scheduler and the engine scheduler have respective function responsibilities, corresponding instance handles are created and destroyed through different derived functions, and interface expansion can be performed according to requirements on the premise of being compatible with old interfaces to realize function iteration.

The invention discloses a computer cluster security protection method and system. The method comprises the following steps: collecting multi-dimensional operation data in real time through an agent program deployed at each node; feature parameters are extracted from the data, and a safety evaluation model is adopted to calculate a safety score and map the safety score into a discrete safety level; gradient protection strategies including basic monitoring, enhanced protection and strict isolation strategies are automatically matched and loaded according to the security level; and monitoring a strategy execution effect, and performing closed-loop optimization on the model parameters based on feedback data. The system correspondingly comprises a data acquisition module, a safety evaluation module, a strategy management module, a strategy execution module and a feedback optimization module. According to the method, the problems of static strategy stiffness, inaccurate risk quantification, single response means and lack of self-learning are solved, dynamic self-adaptive protection is realized, the evaluation accuracy is improved, false alarms and missing alarms are reduced, and the cluster environment is ensured to be safe and reliable.

The invention discloses a cluster computing power configuration management and automatic deployment system, which belongs to the technical field of computer cluster management, and comprises a resource state tracking module used for monitoring resource state information of each computing node in a cluster in real time and establishing a resource pool, an exclusive resource pool and a shared resource pool according to the resource state information, the resource pool comprises a first idle node; the resource allocation arbitration module is used for generating a second idle node and a third idle node according to the first idle node; the task resource matching module is used for screening a matched third idle node from the shared resource pool; and the resource recovery coordination module is used for recovering the low-utilization-rate computing node and the first idle node, the second idle node and the third idle node which have completed the computing task into a resource pool. The multi-level resource pool is established through the resource state tracking module, the problem of rigidity of resource allocation of a traditional system is effectively solved, and the computing resource allocation flexibility and utilization rate are remarkably improved.

The embodiment of the invention provides a data transmission method and a computing node. The method is applied to a first computing node of a computer cluster, the computer cluster comprises a plurality of computing nodes, a storage node and computing interconnection CXL equipment, the plurality of computing nodes are connected with the storage node through the CXL equipment, and the storage node comprises a message space and a data space. The method comprises the steps that a first computing node stores target data in a first storage position through CXL equipment, the first computing node is any computing node in a plurality of computing nodes, and a data space comprises the first storage position; the first computing node stores a notification message corresponding to the target data in a second storage position through the CXL equipment; the notification message is used for notifying the second computing node that the target data is stored in the first storage position, the second computing node is any computing node except the first computing node in the plurality of computing nodes, and the message space comprises a second storage position. Through the method, the data transmission efficiency in the computer cluster can be improved.

A method, computer system, and a computer program for managing computer jobs in a queue is provided. This comprises extracting metadata from a new job received for processing and upon determining when a similar enriched metadata exists in a database. A job score and storage footprint may then be determined for the new job from the extracted metadata. It is then determined whether the new job can be grouped for processing with any other jobs already placed on a queue. The new job is then added to the queue based on the new job's score and footprint, and whether it can be grouped with other jobs. The queue is then updated and sent to a scheduler for further processing.

The invention provides a PKS-based resource integration optimization method and device and a storage medium, and relates to the technical field of resource integration, and the method comprises the steps: determining the number of containers needing to be deployed and a resource distribution strategy according to the scale and performance requirements of a computer cluster, and configuring a virtual network according to the network requirements of a training system, the method comprises the following steps: establishing a container cloud platform, establishing a storage resource pool according to a data type of an accessed target device, allocating corresponding computing resources to an application program based on the container cloud platform, configuring a load balancer and setting a load balancing algorithm on the container cloud platform, and when an operation request of a user on the target device is received, sending the operation request to the target device; the operation request is allocated to the matched computing node for request analysis, the working condition of each application program based on the container cloud platform and the use condition of each storage resource pool are counted in real time, and visual display is performed, so that efficient operation and stable function exertion of the training system are effectively ensured.

The embodiment of the invention discloses a computer cluster parallel processing method for a reverse time migration program, which comprises the following steps that: a main node carries out parallel processing according to a list file containing to-be-processed shot information in a global disk, a flag file in a first global file directory and a flag file in a second global file directory; determining a specified gun and distributing a gun data processing task corresponding to the specified gun to the computing node; the first global file directory is used for storing mark files of shots which are subjected to reverse time migration processing on each computing node, and the second global file directory is used for storing mark files of shots which are subjected to reverse time migration processing on each computing node; and the computing node responds to the received shot data processing task, calls a single-shot reverse time migration program to execute reverse time migration processing for the corresponding specified shot, and updates the flag file in the first global file directory and / or the second global file directory according to the processing state. In this way, the processing efficiency of the multi-shot reverse time migration task can be improved.

The invention relates to the technical field of computer cluster communication, in particular to a set communication method suitable for a core particle equipment cluster with unified bus interconnection, which comprises the following steps of: acquiring hierarchical topology and link information of the core particle equipment cluster, and constructing a global topology information base according to the acquired hierarchical topology and link information; optimizing the core size level primitive, the node level primitive and the super node level primitive; a link monitoring unit is deployed to monitor a link state index in real time, an alarm is triggered when a link fault is detected, an optimal backup link is screened based on a global topology information base, a communication path is updated, and a dynamic resource scheduling algorithm is adopted to allocate computing power and communication resources; and dynamically adjusting a communication path and bandwidth allocation according to a set communication task requirement and a link dynamic parameter. The method is suitable for a three-level link architecture of a core particle equipment cluster, can improve the communication efficiency and the bandwidth utilization rate, enhances the cluster reliability and the resource utilization rate, and effectively meets the requirements of high computing power and high communication intensity of large model training.

Provided is a processing system that executes a process flow in which computation results are aggregated after data is divided and processed. This processing system comprises: a computer cluster 10 that is composed of a plurality of computation units which each process divided data; a DPU cluster 20 that is composed of at least one DPU which aggregates computation results; and a cross-connect switch 30 that connects devices. A setting device enumerates process flows which can be established by the processing system while changing the number l of interfaces of a computation unit to be used and the number m of DPUs to be used, combines the process flows such that data communication time in the processing system is minimized, and sets the cross-connect switch 30 such that a combination of the process flows can be executed.

In some embodiments, a computer cluster system comprises a plurality of nodes and a software package comprising a user interface and a kernel for interpreting program code instructions. In certain embodiments, a cluster node module is configured to communicate with the kernel and other cluster node modules. The cluster node module can accept instructions from the user interface and can interpret at least some of the instructions such that several cluster node modules in communication with one another and with a kernel can act as a computer cluster.

Provided in the embodiments of the present application are a data transmission method and a computing node. The method is applied to a first computing node of a computer cluster, wherein the computer cluster comprises a plurality of computing nodes, a storage node and a compute express link (CXL) device, the plurality of computing nodes are connected to the storage node by means of the CXL device, and the storage node comprises a message space and a data space. The method comprises: a first computing node storing target data in a first storage location by means of a CXL device, wherein the first computing node is any computing node among a plurality of computing nodes, and a data space comprises the first storage location; and by means of the CXL device, the first computing node storing in a second storage location a notification message corresponding to the target data, wherein the notification message is used for notifying a second computing node that the target data is stored in the first storage location, the second computing node is any computing node among the plurality of computing nodes other than the first computing node, and a message space comprises the second storage location. The method can improve the data transmission efficiency in a computer cluster.

The application provides a resource fragment consolidation method, a controller, a control node and a computer cluster, and relates to the technical field of data processing. The method simulates resource scheduling according to a first resource capacity and a first task currently being executed to obtain a plurality of candidate resource scheduling paths; determines a target resource scheduling path from the plurality of candidate resource scheduling paths; and schedules the first task based on the target resource scheduling path to perform resource fragment consolidation. After each candidate resource scheduling path is used to perform resource fragment consolidation, the residual resource capacity of at least one node can reach the first resource capacity. That is, the method can actively eliminate resource fragments in the computer cluster by scheduling the first task to perform resource fragment consolidation. Compared with the method of slowing down the generation of new resource fragments from the source, the scheduling reliability of a task with high resource demand can be effectively improved.

The invention belongs to the field of gas turbine engines, and particularly relates to a simplified calculation method for aero-engine outer culvert flow resistance, when the engine outer culvert flow resistance is calculated, a complex outer culvert structure is split, a whole outer culvert simulation model does not need to be established, M * N outer culvert sectors are used for replacement, the size of the simulation model is reduced, and the calculation efficiency is improved. And the maximum grid quantity of simulation calculation is reduced, so that the simulation problem which can only be solved by a computer cluster or supercomputing originally can be solved by a single machine. After the structure of the outer culvert is split, flow resistance data is obtained through parts developed by accessories such as a duct heat exchanger, a borescope and a pipeline in outer culvert sectors, part-level conversion flow-relative resistance loss correlation correction is achieved, and the utilization rate of test data can be improved.

The application relates to the technical field of data processing, and discloses a computer cluster data processing method and device, the method comprises the following steps: acquiring cluster data, wherein the cluster data is any one of continuous storage data and dispersed storage data; performing chain layer storage and analysis on a computer cluster according to the cluster data to obtain a cluster data block; wherein the computer cluster refers to a computer executable program running on a cluster storage structure; performing vectorization coding according to the cluster data block to obtain a data block vector; performing mapping according to the data block vector to obtain mapping data which can be operated on a computer; and inputting the mapping data into a data calculation formula to perform calculation processing, so as to obtain a processed data linked list. The method can solve the problem of low data processing efficiency in the prior art due to the limited capacity of structured modeling for mass data processing.

The application discloses a dual-computer high-availability implementation method and device, equipment and a medium, relates to the technical field of computers, and is applied to a Keepalived service tool of any service node in a dual-computer cluster, and comprises the following steps: acquiring a current node state of a local service node; determining a current script corresponding to the current node state from a plurality of preset scripts; wherein the plurality of preset scripts comprise a first preset script corresponding to a node state being a master node state, a second preset script corresponding to the node state being a backup node state, and a third preset script corresponding to the node state being a Keepalived service hanging state; and executing the current script. When the node state is the master node state, the backup node state and the Keepalived service hanging state respectively, the corresponding current script is automatically executed, so that the operation and maintenance of the local service node is implemented, and the high availability and stability of the dual-computer cluster can be ensured.

A device-cloud collaborative data mining method includes: A cloud determines a target text and a task configuration file based on a service requirement; encodes the target text by using a text encoder to obtain a text feature; and places the text feature in the task configuration file and delivers the task configuration file to a vehicle. The vehicle encodes image data by using a first picture encoder to obtain an image feature; calculates a value of a similarity between the text feature and the image feature; determines a target picture based on the value of the similarity and the task configuration file; and uploads the target picture to the cloud. The first picture encoder is obtained by compressing and tuning a second picture encoder, and the text encoder and the second picture encoder are two modules of a large image-text multimodal model. This application is applied to an autonomous driving shadow mode, and data mining of any target text of interest can be completed by using a large model, without designing and developing a detection rule for each type of target of interest. In this way, a picture encoder on the vehicle may be continuously updated and optimized, to improve model iteration efficiency.

The present invention discloses a cluster monitoring and alarm method based on a DQN algorithm, and belongs to the field of computer technology. The method includes: collecting monitoring data in a computing cluster state monitoring system, and preprocessing the data; extracting features effective for monitoring and alarming from the processed data, annotating the data set, and constructing an experience pool; designing the structure of a deep Q network and initializing it; using a loss function based on a TD error to train the deep Q network, so that the prediction of the action value by the main Q network will be closer and closer to the actual situation, so that the intelligent agent can select the optimal monitoring and alarm action when facing a specific state; the present invention is mainly used for the establishment of a computer cluster monitoring and alarm system, and by introducing deep reinforcement learning technology, the accuracy of monitoring and alarming is improved, and at the same time, it also provides strong support for coping with complex and dynamic system environments.