公开(公告)号：WO2015192812A1

公开(公告)日：2015-12-23

申请号：PCT/CN2015/081988

申请日：2015-06-19

Applicant: TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED

Inventor： JIN, Xing ,  LI, Yi ,  ZOU, Yongqiang ,  GUO, Zhimao ,  WANG, Eryu ,  XUE, Wei ,  CHEN, Bo ,  LI, Yong ,  BAO, Chunjian ,  XIAO, Lei

IPC: G06F9/38 ,  G06F9/50

CPC classification number: G06T1/20 ,  G06F9/5016 ,  G06F9/5027 ,  G06F9/52 ,  G06N3/0454 ,  G06N3/063 ,  G06N3/084 ,  G06T1/60

Abstract: A parallel data processing method based on multiple graphic processing units (GPUs) is provided, including: creating, in a central processing unit (CPU), aplurality ofworker threads for controlling a plurality ofworker groups respectively, the worker groups including one or more GPUs; binding each worker thread to a corresponding GPU; loading a plurality ofbatches of training data from a nonvolatile memory to GPU video memories in the plurality ofworker groups; andcontrolling the plurality of GPUs to perform data processing in parallel through the worker threads. The method can enhance efficiency ofmulti-GPU parallel data processing. In addition, a parallel data processing apparatus is further provided.

Abstract translation: 提供了一种基于多个图形处理单元（GPU）的并行数据处理方法，包括：在中央处理单元（CPU）中分别创建用于控制多个工作组的多个工作线程，所述工作人员组包括一个或多个GPU; 将每个工作线程绑定到相应的GPU; 将来自非易失性存储器的多个训练数据的批次加载到所述多个工作组中的GPU视频存储器; 并且控制多个GPU以通过工作线程并行地执行数据处理。 该方法可以提高多GPU并行数据处理的效率。 另外，还提供并行数据处理装置。

公开(公告)号：WO2015078370A1

公开(公告)日：2015-06-04

申请号：PCT/CN2014/092258

申请日：2014-11-26

Applicant: TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED

Inventor： ZHUANG, Qianyu ,  BAO, Chunjian ,  MAI, Yihua ,  ZHAI, Yantang

IPC: G06F17/30

CPC classification number: G06F17/30117 ,  G06F11/14 ,  G06F11/1435 ,  G06F17/30138

Abstract: A method, a device, a node and a system for managing file in distributed data warehouse are provided. The method includes: acquiring, by a data node, a deleting instruction carrying a data block identifier, wherein the deleting instruction is sent by a management node; suspending, by the data node, the deleting instruction; and deleting, by the data node, a data block corresponding to the data block identifier after a condition is met, thereby resolving the technical issue that an accidentally deleted file can not be recovered by setting a trash in the management node in some cases and ensuring the data security of the Hadoop system.

Abstract translation: 提供了分布式数据仓库中管理文件的方法，设备，节点和系统。 该方法包括：由数据节点获取携带数据块标识符的删除指令，其中删除指令由管理节点发送; 由数据节点暂停删除指令; 并且在满足条件之后，由数据节点删除与数据块标识符相对应的数据块，从而解决在某些情况下通过设置管理节点中的垃圾而无法恢复意外删除的文件的技术问题，并确保 Hadoop系统的数据安全性。

公开(公告)号：WO2015062520A1

公开(公告)日：2015-05-07

申请号：PCT/CN2014/089907

申请日：2014-10-30

Applicant: TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED

Inventor： ZHANG, Yunlong ,  LEI, Hailin ,  WANG, Tiebo ,  PAN, Anqun ,  LV, Mingyang ,  LI, Gang ,  BAO, Chunjian

IPC: H04L12/24 ,  H04L29/06

CPC classification number: H04L69/40 ,  G06F11/20 ,  G06F17/30 ,  H04L63/10 ,  H04L67/1002 ,  H04L67/1034 ,  H04L67/1095

Abstract: A method and advice for handing over between master and standby storage nodes and a storage medium are disclosed by the embodiments of the present invention, and the method includes: a master storage node sends service request data to a consistency control center for registration and synchronizes a standby storage node with the service request data if the registration is successful; when the consistency control center determines that the synchronization of the registered standby storage node with the service request data from the master storage node is failed, the consistency control center stores a key value corresponding to the registered service request data involved in the failed synchronization into a blacklist; the standby storage node obtains the blacklist from the consistency control center after the handing over; and when the standby storage node receives new service request data that corresponds to a key value the same as one stored in the blacklist, refusing, by the standby storage node, to provide a data writing service for the new service request data; otherwise, allowing, by the standby storage node, to provide the data writing service for the new service request data. In the embodiments of the present disclosure, the high consistency between the service request data on the master storage node and that on the standby storage node is ensured and the data writing service function is maintained on the standby storage node.

Abstract translation: 通过本发明的实施例公开了在主存储节点和备用存储节点与存储介质之间切换的方法和建议，并且该方法包括：主存储节点向一致性控制中心发送服务请求数据以进行注册并同步 如果注册成功，则具有服务请求数据的备用存储节点; 当一致性控制中心确定登记的备用存储节点与来自主存储节点的服务请求数据的同步失败时，一致性控制中心将与失败同步中涉及的注册服务请求数据相对应的密钥值存储到 黑名单; 备用存储节点在交接之后从一致性控制中心获取黑名单; 并且当所述待机存储节点接收到与存储在所述黑名单中的密钥值相同的密钥值的新的服务请求数据时，由所述备用存储节点拒绝为所述新的服务请求数据提供数据写入服务; 否则，由备用存储节点允许为新的服务请求数据提供数据写入服务。 在本公开的实施例中，确保主存储节点上的服务请求数据与备用存储节点上的服务请求数据之间的高度一致性，并且在备用存储节点上维护数据写入服务功能。