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

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

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: The present disclosure provides a method and system for implementing an analytic function based on MapReduce. The method includes: a table scan operator acquiring a data row from a file block, and sending the data row to a reduce sink operator; upon receipt of the data row, the reduce sink operator determining a reduce key, a partition key, and a sort key of the analytic function, and sending the data row to an analysis operator by means of a MapReduce framework; and upon receipt of the data row, the analysis operator analyzing the data row to obtain an analytic result, and forwarding the data row and the analytic result to a subsequent operator. The present disclosure can implement an analytic function in a distributed data warehouse of the MapReduce framework, thereby solving a problem that the analytic function cannot be used in the distributed data warehouse based on the MapReduce framework to perform data analytical processing.