Method of rebinding computing unit in heterogeneous computing clouds and apparatus thereof

A dynamic rebinding device for a computing unit in heterogeneous computing clouds, comprising:

• a master node and a plurality of compute nodes,
• wherein each of the plurality of compute nodes comprises heterogeneous accelerators and a cache memory to temporarily store data,
• wherein the master node comprises:
• a transceiver transmitting or receiving data to or from a terminal and the plurality of compute nodes;
• a memory storing device binary codes and libraries for the heterogeneous accelerators;
• at least one processor connected to the memory;
• and
• an input/output unit,
• wherein the at least one processor is configured to:
• receives a first computing request from the terminal;
• generate a static FAT binary by compiling the device binary codes for each of all types of heterogeneous accelerators based on the first computing request;
• generate and store global data including the device binary codes and the libraries of all types of the heterogeneous accelerators;
• generate a structure capable of dynamic binding by separating the device binary codes for each type of the heterogeneous accelerators from the static FAT binary;
• allocate a first compute node and first heterogeneous accelerators based on the first computing request;
• dynamically bind the computing unit including the device binary codes and the libraries for the allocated first heterogeneous accelerators;
• transmit the computing unit dynamically bound to the first compute node;
• operate the allocated first heterogeneous accelerators based on the computing unit;
• generate local caching data including the device binary codes and the libraries of all heterogeneous accelerators of the assigned first computer node excepting the assigned first heterogeneous accelerators from the stored global data;
• transmit the local caching data to the assigned first compute node;
• receive a second computing request from the terminal;
• and
• determine an operation of the computing unit based on the second computing request, and
• wherein the at least one processor is further configured to:
• allocate a second compute node and second heterogeneous accelerators based on the second computing request, and
• determine the operation of the computing unit based on whether the first compute node and the second compute node are identical and whether the first heterogeneous accelerators and the second heterogeneous accelerators are identical.

A non-transitory recording medium readable by a digital processing device, in which a program of instructions executable by the digital processing device is tangibly implemented to dynamically rebind a computing unit in heterogeneous computing clouds, wherein the digital processing device comprising at least one processor is configured to execute the instructions of the program recorded in the recording medium, thereby configuring the at least one processor to:

• receive a first computing request from a terminal;
• generate a static FAT binary by compiling device binary codes for each of all types of heterogeneous accelerators based on the first computing request;
• generate and store global data including the device binary codes and libraries of all types of the heterogeneous accelerators;
• generate a structure capable of dynamic binding by separating the device binary codes for each type of the heterogeneous accelerators from the static FAT binary;
• allocate a first compute node and first heterogeneous accelerators based on the first computing request;
• dynamically bind the computing unit including the device binary codes and the libraries for the allocated first heterogeneous accelerators;
• transmit the computing unit dynamically bound to the first compute node;
• operate the allocated first heterogeneous accelerators based on the computing unit;
• generate local caching data including the device binary codes and the libraries of all heterogeneous accelerators of the assigned first computer node excepting the assigned first heterogeneous accelerators from the stored global data;
• transmit the local caching data to the assigned first compute node;
• receive a second computing request from the terminal;
• and
• determine an operation of the computing unit based on the second computing request, and
• wherein the at least one processor is further configured to:
• allocate a second compute node and second heterogeneous accelerators based on the second computing request, and
• determine the operation of the computing unit based on whether the first compute node and the second compute node are identical and whether the first heterogeneous accelerators and the second heterogeneous accelerators are identical.