173 results about "Remote direct memory access" patented technology

The invention relates to the technical field of remote direct memory access, in particular to an RDMA shared receiving queue optimization method and device, equipment, a medium and a product. A shared receiving queue is divided into multiple containers, and the multiple containers are connected in series through a logic linked list; when the RDMA data of the queue pair arrives, if the container is not bound, distributing a container for the queue pair; monitoring the release state of the shared receiving queue entry; and when the receiving side software monitors that the target container is in the completely idle state, re-linking the target container into the logic linked list for distribution independent of the states of other containers in the plurality of containers. According to the method disclosed by the invention, dynamic reuse of resources can be realized through resource monitoring taking the container as granularity and an out-of-order recovery mechanism independent of other container states. After all the SRQEs in any container are released, the SRQEs can be immediately put into redistribution, so that the problem of waste of tail idle resources caused by occupation of head resources in a traditional linear queue can be thoroughly avoided.

The invention provides a DPU-based remote direct memory access queue pair state machine implementation method and device, and the method comprises the steps: creating a flag bit, and setting the value of the flag bit according to each state of a queue pair state machine, so as to achieve state switching; creating a queue pair, and entering a reset state; configuring attributes of a queue pair, and switching to an initialization state; setting parameters of a receiving end, and switching to a ready receiving state; establishing communication connection, setting parameters of a sending end, switching to a ready-to-send state, and executing remote direct memory access operation so as to perform data transmission; if all the data in the sending queue is sent, switching to a sending completion state; if the sending queue has an error, switching from a sending preparation state or a sending completion state to a sending error state; if the queue pair has an error, switching from the current state to the queue pair error; and switching to a reset state after processing an error. According to the method, the flexibility and expansibility of the system are improved, and the processing efficiency of error and abnormal conditions is improved.

The invention provides an RDMA (Remote Direct Memory Access) out-of-order receiving and selective retransmission method, which comprises the following steps of: recording an arrival state of a data packet by adopting a Bitmap algorithm and / or a segmentation marking algorithm of a sliding window, dynamically generating ACK (Acknowledgement Character) or NAK (Negative Attached Keying) feedback information according to the algorithm, and only retransmitting a lost data packet according to the received NAK feedback information. The RDMA out-of-order receiving and selective retransmission method supports Bitmap statistics of a sliding window or a segment marking method of saving a memory, is compatible with an existing protocol and realizes an ACK mechanism of selective retransmission, and avoids the problem of redundant transmission in a traditional mechanism. The invention further discloses computer equipment, a data transmission system and a computer readable storage medium.

The invention discloses a data stream processing method and device, electronic equipment and a storage medium, and relates to the technical field of distributed storage. According to the data stream processing method and device, the data stream transmission mode is pre-judged to be intra-node transmission or cross-node transmission based on a hardware topology mapping table, and read-write caches are dynamically allocated according to the pre-judged data stream transmission mode; data received by the front-end network card is directly written into a controller memory buffer area of the target equipment through point-to-point direct memory access for data streams transmitted in the nodes, and is read and forwarded from the target equipment by the rear-end network card; the sending cache and the receiving cache of the data stream transmitted across the nodes are locked to the non-uniform memory access node to which the target network card belongs, and the data are sent to the target node through the remote direct memory access register memory, so that the problem that the data stream processing is not optimized in combination with hardware topology characteristics can be solved; the technical effects of reducing data replication overhead, avoiding redundant access across hardware nodes, improving data stream transmission efficiency and realizing end-to-end zero replication transmission are achieved.

The application discloses a network interface card, a storage device, a message receiving method and a message sending method, relates to the storage field, and is used for improving the reliability of a storage device using an RDMA network and improving the storage access performance of a single QP connection. The network interface card comprises a parser and at least one shared connector. The parser is used for distributing a received first remote direct memory access (RDMA) message to a first shared connector in the at least one shared connector. The first shared connector is used for establishing a QP connection with a peer network device, and determining a first queue pair (QP) corresponding to a first controller according to the first RDMA message to realize communication with the first controller. The first shared connector is connected with a plurality of controllers including the first controller, and one QP corresponding to each controller in the plurality of controllers corresponds to the first shared connector.

The invention discloses a selective retransmission function test method, device and system for remote direct memory access, a medium and a product, and relates to the technical field of network tests.The method comprises the steps that when message features of a received test message are successfully matched with a packet loss rule, redirection processing is carried out on the test message so as to achieve controllable discarding, and the message features of the test message are successfully matched with the packet loss rule; according to the method and the device, the selective retransmission function is triggered, and the release rule corresponding to the test message takes effect to ensure that the subsequent corresponding retransmission message can be normally forwarded, so that only the first transmission of the original message is intercepted, the retransmission message is released, and the purpose of controllable packet loss is achieved; and finally, the selective retransmission function is evaluated according to the redirected message information in the function test process and the message receiving information of the message receiving end, so that the technical problems that the test packet loss is uncontrollable and the accurate test requirement is difficult to meet in the related technology can be solved, and the test efficiency is improved. The technical effects of accurate and controllable verification and performance evaluation of the selective retransmission function are achieved.

Devices, systems, methods, and processes for incast congestion management are described herein. Typically, in a Packet Sequence Number (PSN) based Remote Direct Memory Access (RDMA) network, Priority Flow Control (PFC) is asserted upstream when an incast congestion event occurs, which can victimize unrelated flows. Thus, instead of asserting PFC, a switch in the PSN based RDMA network detects an incast congestion event and directly notifies one or more Reliable Connection (RC) Queue Pairs (QPs) of various sending devices, associated with the incast congestion event using Receiver Not Ready (RNR) negative acknowledgements (NACKs). These RNR NACKs are associated with unique pause time-periods. The associated RC QPs receive the RNR NACKs and pause packet transmission. The associated RC QPs resume packet transmission upon expiration of corresponding pause time-periods. Thus, the packet transmission from the contributing RC QPs is spaced out, avoiding all packets reaching a switch output port at the same time.

The invention relates to the technical field of supercomputing system operation and maintenance, and discloses a method for resetting an intelligent operating system of a supercomputing system node, which comprises the following steps of: 101, acquiring data of each computing node of a supercomputing system in real time, transmitting the data to a monitoring center through an RDMA (Remote Direct Memory Access) high-speed network for retention to form historical operating data, the historical operation data comprises historical index time sequence data and historical fault label data; 102, a three-dimensional array is generated in the historical operation data through a sliding window method, the three-dimensional array comprises sample information, time information and feature information, and the information in the three-dimensional array is spliced in sequence; and step 103, constructing a fault prediction model, inputting the three-dimensional array into the fault prediction model, and outputting a fault probability by the fault prediction model. According to the method, the fault prediction model is constructed, the node data of the supercomputing system is collected and processed in real time, the fault occurrence probability can be predicted, and an intelligent decision basis is provided for resetting of the supercomputing node operating system.

The invention provides an RDMA-based big data transmission and storage system and method and a semiconductor detection device.The big data transmission and storage system comprises a hardware layer and a software layer, the hardware layer comprises an NVMe SSD storage array, an NVMe RAID acceleration card based on a GPU architecture and an RDMA network card, and the software layer comprises an NVMe SSD storage array, an NVMe RAID acceleration card based on the GPU architecture and an RDMA network card; the software layer comprises a multi-thread transmission module, a self-adaptive memory pool module, a transmission control module, a data verification and retransmission module, a scheduling and optimizing module and a disk falling and storage management module. According to the method, multi-channel transmission parallelization is achieved through data direct connection, zero copy transmission, RAID acceleration and multi-thread parallel RDMA transmission in a hardware link, data is directly connected and transmitted from a memory to the NVMe SSD storage array, data receiving and disk falling thread decoupling directly access the NVMe SSD storage array through asynchronous writing, network overload or storage blockage is avoided, and the data transmission efficiency is improved. The data transmission efficiency and reliability are high, and popularization and application in scenes needing large-scale image transmission processing such as semiconductor detection are facilitated.

The invention discloses a memory pool device fusing an extended network and a management method, and the device comprises a super network card chip which is in communication connection with an intelligent computing cluster network through a network port, and provides a shared memory service for the intelligent computing cluster network based on a remote direct memory access protocol; the intelligent computing cluster network comprises a transverse expansion network and a longitudinal expansion network; the central controller is connected with the super network card chip; the CXL memory unit is connected with the super network card chip, and the FPGA expansion unit and / or the XPU expansion unit are / is connected with the super network card chip. According to the invention, through efficient communication between the super network card chip and the intelligent computing cluster networking, flexible scheduling and sharing of memory resources in the cluster are realized. Wherein the super network card chip is used as a core component, not only is responsible for communication of a transverse expansion network, but also realizes bridging access of a high-bandwidth domain of a longitudinal expansion network, so that the data transmission efficiency and the memory utilization rate can be improved, and high-bandwidth and low-delay data transmission and sharing are realized.

The invention relates to a method for accelerating file acquisition by using RDMA (Remote Direct Memory Access), and belongs to the technical field of AI (Artificial Intelligence) training, reasoning and file transmission in a large-scale CPU (Central Processing Unit) calculation-intensive scene. Aiming at the problems of IO bottleneck, client memory limitation and high special hardware deployment cost existing in traditional file reading, a client-server architecture is adopted, a control plane and a data plane are separated, a server loads a remote shared file through a large memory and registers as an RDMA memory area, and the remote shared file is stored in the RDMA memory area. And establishing a quick index of the file and the memory area by means of Hash mapping. The client directly accesses the memory of the server based on RDMAreadverbs, data does not need to be forwarded and copied by a kernel, oversized files can be read in blocks, and the system is compatible with a common NFS storage system and an ROCE network card. The method effectively breaks through the memory limitation of the client, reduces the deployment and hardware cost, guarantees the data transmission accuracy, is suitable for a multi-client concurrent access scene, and provides a low-delay and high-bandwidth efficient solution for high-frequency data set access.

An application identifies a metadata field within a Remote Direct Memory Access (RDMA) message structure that can be used for timestamp injection. The application generates a high-resolution timestamp that captures a time when the RDMA message is sent, where the high-resolution timestamp has a first size that exceeds a size of the metadata field. The application compresses the high-resolution timestamp into a compressed timestamp having a second size that fits within the size of the metadata field, and injects the compressed timestamp into the metadata field of the RDMA message, the RDMA message sent from a sender host to a receiver host. The application calculates a one-way delay between the sender host and the receiver host using the compressed timestamp, and outputs a control signal based on the one-way delay.

A method for remote direct memory access (RDMA) communication includes transmitting, from a first device to a second device via a network, a first RDMA message, storing, by the first device, a transmit byte count (tx_byte_count) of a total number of bytes transmitted in the first RDMA message, receiving, by the first device from the second device, a second RDMA message associated with the first RDMA message, the second RDMA message comprising a receive byte count (rx_byte_count) of a total number of bytes of the first RDMA message received by the second device, and determining, by the first device, a size of a congestion window on the network based on the tx_byte_count and the rx_byte_count.

The invention provides a congestion control method and device, electronic equipment, a storage medium and a program product. The method comprises the following steps: under the condition that remote direct memory access (RDMA) connection is established with receiving end equipment, acquiring real-time loads of a plurality of equivalent paths corresponding to the RDMA connection; and based on the real-time load of the plurality of equivalent paths, adjusting the transmission path of the message in the RDMA connection by changing the transport stream identification information of the message, so as to realize the load balancing of the plurality of equivalent paths.

A method processes data requests from a requester to a responder over a network by using remote direct memory access. The method includes the requester issuing two or more requests into a work queue of a network interface controller. The method further includes the network interface controller executing the queued requests by sending them to the responder, and receiving responses to the sent request from the responder. One or more given requests each includes dependency information related to the response from the responder to one or more previous requests in the work queue, the method further includes, before executing the given request by the network interface controller, evaluating the dependency information and executing the given request depending on the result of the evaluation.

Computing systems and methods for performing remote direct memory access in a computing system are provided. In some embodiments, the system includes a storage program processing circuit, a network interface circuit, a cache coherence switch, and a first memory module. In some embodiments, the first memory module is connected to the cache coherence switch, the cache coherence switch is connected to the network interface circuit, and the storage program processing circuit is connected to the cache coherence switch.

The application provides a remote direct memory access device and method. The remote direct memory access device comprises a sending module and a receiving module. The sending module acquires a current work queue element, and determines that the current work queue element is an abnormal work queue element after the current work queue element appears abnormal. The sending module determines whether all work queue elements before the abnormal work queue element have received corresponding response responses. If all work queue elements before the abnormal work queue element have received corresponding response responses, and the receiving module no longer receives response responses, the sending module generates an abnormal completion queue element for the abnormal work queue element, and writes the abnormal completion queue element into a completion queue. The application can still write corresponding CQEs into the completion queue for the WQE that has not been sent out after the work queue appears abnormal, and can also ensure that the order of the CQEs is consistent with the order of the corresponding WQE.

The embodiment of the invention provides a communication method and a related device, the method is applied to a terminal, and the method comprises the steps that first connection information is acquired, and the first connection information is used for establishing remote direct memory access (RDMA) connection; establishing the RDMA connection with a first functional entity through the first connection information, wherein the first functional entity is deployed in network equipment; and calling a memory resource of the first functional entity through the RDMA connection. By adopting the embodiment of the invention, the transmission delay during memory calling can be reduced.

The embodiment discloses a response control method and device, electronic equipment, medium and computer program product, which are applied to a request end device, the request end device is used for network communication with a response end device by using a remote direct memory access (RDMA) method, the response control method comprises the following steps: acquiring a current network state of a network between the request end device and the response end device; determining a congestion control stage currently entered by the request end device; determining a response acquisition frequency based on the current network state and the congestion control stage currently entered; wherein the response acquisition frequency represents a frequency of acquiring response information, and the response information represents a response of the response end device to a data packet sent by the request end device.

The embodiment of the invention relates to the technical field of computers, in particular to a message processing method and a remote direct memory access system. According to the embodiment of the invention, a first first element of a receiving work queue is obtained through a prefetching analysis module to serve as a reference element, the reference element is analyzed to obtain reference information of a reference receiving cache, the reference information is sent to an aggregation processing module, and the aggregation processing module determines target aggregation information of a target operation message according to the reference information; and obtaining the target data according to the target aggregation information, and sending the target data to the reference receiving cache, thereby dynamically matching one or more request operation messages processed by one first element, performing aggregation processing on the data carried by the target operation message, fully utilizing the storage space of the receiving cache, improving the utilization rate of the storage space of the receiving cache, and improving the user experience. The number of times of acquiring the first element by the remote direct memory access network card is reduced, so that PCIe operation is reduced, processing delay is reduced, and the PCIe bandwidth utilization rate is improved.

A vehicle network communications system including a sending processor for generating a data in response to a first vehicle control algorithm, a sending network interface configured to receive the data from the sending processor, to generate a message authentication code in response to the data, to generate a communications packet for transmission according to a remote direct memory access network protocol, wherein the message authentication code is appended to the communications packet, and to transmit the communications packet via a communications network, a receiving network interface configured to receive the communications packet from the sending network interface via the communications network, to decode the message authentication code to extract the data, and to couple the data the memory, and a memory configured to store the data for use by a receiving processor performing a second vehicle control algorithm.

An example network interface controller (NIC) includes: first resources configured to supply first traffic; second resources configured to supply second traffic; a load balancer, coupled to the first and second resources, configured to balance the first traffic and the second traffic between first and second port circuits of a link aggregation group (LAG) using a hash function; and remote direct memory access (RDMA) logic configured to, using the hash function, divide work requests into a first set of work requests for first packets that hash to the first port circuit and a second set of work requests for second packets that hash to the second port circuit, the RDMA logic configured to supply the first set of work requests to the first resources and the second set of work requests to the second resources.

The present disclosure provides for a converged compute platform architecture, including a first infrastructure processing unit (IPU)-only configuration and a second configuration wherein the IPU is coupled to a central processing unit, such as an x86 processor. Connectivity between the two configurations may be accomplished with a PCIe switch, or the two configurations may communicate through remote direct memory access (RDMA) techniques. Both configurations may use ML acceleration through a single converged architecture.

An application identifies a metadata field within a Remote Direct Memory Access (RDMA) message structure that can be used for timestamp injection. The application generates a high-resolution timestamp that captures a time when the RDMA message is sent, where the high-resolution timestamp has a first size that exceeds a size of the metadata field. The application compresses the high-resolution timestamp into a compressed timestamp having a second size that fits within the size of the metadata field, and injects the compressed timestamp into the metadata field of the RDMA message, the RDMA message sent from a sender host to a receiver host. The application calculates a one-way delay between the sender host and the receiver host using the compressed timestamp, and outputs a control signal based on the one-way delay.

This application discloses a file read operation method, apparatus, and medium for a distributed storage system, relating to the field of communications. To improve the efficiency of read operations in a distributed storage system, this application involves receiving a read operation request; allocating memory space based on the read operation request; packaging the read operation request and the memory space address information into a read operation message; and sending the read operation message to an object storage device using remote direct memory access (RDA) communication, so that the object storage device writes the requested data into the memory space according to the read operation message. The client sends a read operation message containing control data for the read operation request and the memory space address information to the object storage device using RDA communication. The object writes the requested data into the specified memory space, achieving separation of control and data transfer, reducing the number of data copies between memory locations, and improving network transmission efficiency through the application of RDA technology.

The invention belongs to the field of remote direct memory access, and discloses a bitmap-based RDMA (Remote Direct Memory Access) data packet out-of-order receiving circuit and method, a bitmap checking module verifies the validity of a data packet and outputs a bitmap switching identifier, and a bitmap management module maintains pointer information through three states of order, out-of-order and bitmap updating; the bitmap information storage module is used for storing bitmap information, calculating the offset of a bitmap writing pointer in an out-of-order state and starting a fragmentation timeout mechanism, generating a window sliding enable signal when the bitmap state is updated, the bitmap information storage module responds to the enable signal to execute bitmap writing and window sliding updating, and the bitmap pointer storage module manages a multi-queue bitmap state identifier and a pointer in an RAM (random access memory) structure. By adopting the circuit, the bandwidth waste and congestion risk caused by retransmission can be obviously reduced; the tail delay jitter is reduced; and the adaptive capacity to multi-path network disorder is enhanced, so that the bandwidth utilization rate and the transmission reliability are comprehensively improved.

The invention discloses an RDMA (Remote Direct Memory Access) protocol transparent adaptation system supporting a high RTT (Room-to-Tilt) environment, which is characterized in that a segmented logic transmission channel is formed by introducing logic transmission nodes and inserting two logic transmission nodes between two physical nodes, and a multi-stage assembly line transmission channel consisting of a near-end RDMA transmission channel and a far-end RDMA transmission channel is formed; and logic isolation between the transmission state of the near-end transmission channel and the transmission state of the far-end transmission channel is realized. Meanwhile, link measurement, RTT perception retransmission, RTT perception sorting, a near-end pickup mechanism facing near-end physical nodes and a rapid congestion feedback mechanism facing far-end logic transmission nodes are realized in the logic transmission nodes; the problems of unstable service flow caused by flow control delay, low link utilization rate caused by data retransmission and the like in traditional RDMA transmission in a high RTT environment are solved.

This application provides a data transmission method, a network interface card (NIC) for a data receiving end, an electronic device, and a storage medium. The data transmission method is applied to the NIC of the data receiving end. The NIC writes data to the host memory of the data receiving end via a bus. The data transmission method includes: if the bus bandwidth resources required for the NIC to write the target data to the host memory exceed the bus bandwidth resources available to the NIC, then writing a portion of the target data to the NIC's NIC buffer; the NIC buffer includes a first memory buffer configured for the NIC based on its on-chip memory and a second memory buffer configured for the NIC based on its onboard memory; the data sending end sends the target data to the data receiving end via remote direct memory access; and a portion of the data is written from the NIC buffer to the host memory. The solution of this application can better avoid data packet loss problems when facing sudden traffic surges.

The requesting device of the present invention is capable of transmitting multiple prompt messages simultaneously, and enables the responding device to optimize various aspects through memory prefetching, cache warming, and multiple atomic operations. These operations are used to enable the responding device to begin prefetching memory regions in advance (if necessary), or to prepare to receive incoming data packets in an efficient manner.

An entity for RDMA is configured to maintain a time-based queue pair (QP). The time-based QP comprises a first area associated with a time-based indication and is configured to hold one or more first WQEs. The time-based indication indicates that the one or more WQEs in the first area are to be periodically processed. The entity is further configured to periodically process the one or more first WQEs in the first area according to the time-based indication.