Huawei Technologies Co., Limited

Xfusion Digital Technologies Co., Ltd.

Shengyu Shen

Yumin Wu

Weibin Lin

Liang Li

Lin Ma

Yuchun Lu

Yongyao Li

Jianrong Xu

Wenkai Ling

Beilei Sun

Jiangle Huang

Yaoguo Hu

A data transmission method is provided. The method includes: a network interface card of a source device obtains a first notification message and a second notification message, wherein the first notification message indicates that a first to-be-processed remote direct memory access (RDMA) request exists in a first queue of the source device, the first queue stores a request of a first service application in the source device, the second notification message indicates that a second to-be-processed RDMA request exists in a second queue of the source device, and the second queue stores a request of a second service application in the source device; and the network interface card determines a processing sequence of the first queue and the second queue based on service levels, and sends the first to-be-processed RDMA request and the second to-be-processed RDMA request to a destination device according to the processing sequence.

Data transmission method, chip, and device

An artificial intelligence training technology, applied to an artificial intelligence training network. Before graphics processing units located on different servers need to communicate with each other, an optical channel used for communication is established in advance. Once a graphics processing unit of a previous server completes calculation of the graphics processing unit, a calculation result can be immediately sent to a graphics processing unit on a next server without waiting or only in a short time period, to reduce duration of artificial intelligence training.

AI training network and method

A data transmission method, performed by a receive end, includes: receiving notification information that is about to-be-transmitted data and that is sent by a transmit end, where the notification information includes a size of the to-be-transmitted data of the transmit end; generating a plurality of RDMA read requests based on the size of the to-be-transmitted data; and sending the plurality of RDMA read requests to the transmit end, where a bandwidth occupied by data read by the plurality of sent RDMA read requests is less than or equal to an ingress bandwidth of the receive end.

Data transmission method and related device

Embodiments of this application disclose an encoding method and a related device. The method includes: receiving a to-be-encoded code block whose length is L, where L is a positive integer; and encoding the to-be-encoded code block to obtain a forward error correction FEC code, where a valid information length K of the FEC code is an integer multiple of a largest prime factor of L, and a total length N of the FEC code is a sum of K and a product of 2 and an error correction capability T of the FEC code. According to the embodiments of this application, it can be ensured that an FEC codeword satisfies a requirement for a low latency and a high gain.

Encoding method and related device

A lock allocation method and apparatus, and a computing device to resolve a problem of lock contention, where the method includes a first node controller (NC) receives a first migration queue from a second NC, allocates a first target lock to a first central processing unit (CPU) associated with the first NC, deletes the first lock request at the queue head in the first migration queue when receiving a request for releasing the first target lock from the first CPU, changes a lock flag bit of a second lock request to a lock flag bit indicating a locked state when the first migration queue is not empty, determines a third NC sending the second lock request, and sends the first migration queue to the third NC.

Computing device shared resource lock allocation

A memory allocation method and a server, wherein the method includes: identifying, by a server, a node topology table; generating fetch hop tables of the NUMA nodes based on the node topology table; calculating fetch priorities of the NUMA nodes based on the fetch hop tables of the NUMA nodes, and using an NC hop count as an important parameter for fetch priority calculation; and when a NUMA node applies for memory, allocating memory based on the fetch priority table, and for a higher priority, more preferentially allocating memory from a NUMA node corresponding to the priority.

Growing community of inventors

Hangzhou, China

Shengyu Shen