Abstract: Provided is a method of configuring a cluster, which is a method of assigning graphics processing unit (GPU) servers in a cloud in which a plurality of machine learning (ML) services are executed using an apparatus for configuring a cluster. The apparatus for configuring a cluster is configured to measure the power consumption characteristics of each of the GPU servers constituting the cloud for each of a plurality of different models processing the plurality of ML services and assign at least one GPU server to each of the plurality of models using power consumption characteristics of each of the GPU servers for each of the plurality of models to configure a GPU cluster for each of the plurality of models.

Abstract: Disclosed herein are an apparatus and method for accelerating file I/O offload for a unikernel. The method, performed by the apparatus and server for accelerating file I/O offload for the unikernel, includes; executing, by the apparatus, an application in the unikernal and calling, by the thread of the application, a file I/O function; generating, by the unikernal, a file I/O offload request using the file I/O function; transmitting, by the unikernal, the file I/O offload request to Linux of the server; receiving, by Linux, the file offload request from the thread of the unikernel and processing, by Linux, the file I/O offload request; transmitting, by Linux, a file FO offload result for the file I/O I/O offload request to the unikernel; and delivering the file I/O offload result to the thread of the application.

Abstract: Disclosed is an apparatus and method of processing input and output in a multi-kernel system. A method of processing input and output in a multi-kernel system according to the present disclosure includes: setting a shared memory between a first kernel on a main processor and a lightweight kernel on a parallel processor; setting a data transmission and reception channel between the first kernel on the main processor and the lightweight kernel on the parallel processor using the shared memory; providing, on the basis of the data transmission and reception channel, an input/output task that occurs in the lightweight kernel to the first kernel on the main processor; processing, by the first kernel on the main processor, an operation corresponding to the input/output task; and providing a result of the processing to the lightweight kernel.

Abstract: A method and an apparatus for processing to support scalability in a many-core environment are provided. The processing apparatus includes: a counter unit configured to include a global reference counter, at least one category reference counter configured to access the global reference counter, and at least one local reference counter configured to access the category reference counter; and a processor connected to the counter unit and configured to increase or decrease each reference counter. The at least one category reference counter has a hierarchical structure including at least one layer.

Abstract: Disclosed herein is an operating method of an apparatus for offloading file I/O of a unikernel based on Remote Direct Memory Access (RDMA). The operating method may include calling, by the application of the unikernel, a file I/O kernel function; generating, by the file I/O kernel function, file I/O information; transmitting, by the unikernel, the file I/O information to a Linux; configuring, by the Linux, a file I/O function using the file I/O information and calling the same; and transmitting, by the file I/O function, a file I/O request, corresponding to the file I/O information, to a file server.

Abstract: Scheduling threads in a system with many cores includes generating a thread map where a connection relationship between a plurality of threads is represented by a frequency of inter-process communication (IPC) between threads, generating a core map where a connection relationship between a plurality of cores is represented by a hop between cores, and respectively allocating the plurality of threads to the plurality of cores defined by the core map, based on a thread allocation policy defining a mapping rule between the thread map and the core map.

Abstract: Disclosed is an apparatus and method of processing input and output in a multi-kernel system. A method of processing input and output in a multi-kernel system according to the present disclosure includes: setting a shared memory between a first kernel on a main processor and a lightweight kernel on a parallel processor; setting a data transmission and reception channel between the first kernel on the main processor and the lightweight kernel on the parallel processor using the shared memory; providing, on the basis of the data transmission and reception channel, an input/output task that occurs in the lightweight kernel to the first kernel on the main processor; processing, by the first kernel on the main processor, an operation corresponding to the input/output task; and providing a result of the processing to the lightweight kernel.

Abstract: A nuclear fuel simulation method is provided that includes: step (a) for receiving an input of data on the order in which nuclear fuels are moved; step (b) for extracting, from the data, information on nuclear fuels, the coordinates of locations from which nuclear fuels are unloaded, and the coordinates of locations into which nuclear fuels are loaded; and step (c) for simulating the information extracted in step (b) according to a flowchart of the data. The present invention has an advantage in that it is possible to accurately and quickly verify all fuel movement works requiring the unloading and loading of nuclear fuels by receiving an input of a huge amount of data on the order in which nuclear fuels are moved and systematically verify an error that may occur during a simulation according to a flowchart, which enables the workload of about three man-days, required per cycle for each reactor, to be done in three man-hours, thereby achieving a significant reduction in working time.

Abstract: A method for managing the movements of nuclear fuels is provided that includes: (a) loading a storage status map in which storage racks of an SFPR where spent fuels are stored and an NFS where new fuels are stored are mapped; (b) assigning the storage locations of nuclear fuels and the colors thereof in the storage status map; (c) receiving a pattern type input of tasks for designating the order in which nuclear fuels are moved and the locations to which the nuclear fuels are moved, and generating a movement flowchart; and (d) updating the storage status map according to the level at which the movement flowchart has progressed. The present invention has an advantage in that it is possible to quickly create a nuclear fuel movement flowchart by automating all fuel movement works requiring the unloading and loading of nuclear fuels, which enables the workload of about 30 man-days, required per cycle for each reactor, to be done in three man-hours, thereby achieving a significant reduction in working time.

Abstract: Provided is a many-core system including a resource unit including a resource needed for execution of an operating system and a resource needed for execution of a lightweight kernel, a programing constructing unit configured to convert an input program into an application program and to load the application program into the resource unit, a run-time management unit configured to manage a running environment for executing the application program, and a self-organization management unit configured to monitor the application program and the resources in the resource unit, to dynamically adjust the running environment to prevent a risk factor from occurring during the execution of the application program, and to cure a risk factor which occurred.

Abstract: Provided is a method of scheduling threads in a many-cores system. The method includes generating a thread map where a connection relationship between a plurality of threads is represented by a frequency of inter-process communication (IPC) between threads, generating a core map where a connection relationship between a plurality of cores is represented by a hop between cores, and respectively allocating the plurality of threads to the plurality of cores defined by the core map, based on a thread allocation policy defining a mapping rule between the thread map and the core map.

Abstract: Disclosed herein is an apparatus for supporting automation of configuration management of a virtual machine applicable to a multi-cloud environment. In accordance with an embodiment, the apparatus includes an interface unit for receiving configuration management information or information of a virtual machine to which configuration management is to be applied. A configuration management verification unit verifies the received configuration management information, and combines the virtual machine information with configuration management information corresponding to the virtual machine. A configuration management distribution unit distributes the configuration management information combined with the virtual machine information to a cloud in which the virtual machine is created.

Abstract: Disclosed herein is a cloud service broker apparatus and method for providing a cloud service using the broker apparatus, which provide an optimal cloud service for a cloud service user through negotiations between the cloud service user and a cloud service provider. The cloud service broker apparatus includes an operation information management unit for examining a demand entered by a cloud service user. A broker intermediary unit detects a cloud service based on the demand for which verification has been completed by the operation information management unit, and sends a request for positioning of the cloud service for the cloud service user to a cloud service provider that provides the detected cloud service. A life cycle management unit monitors a cloud service positioned in and used by the cloud service user at a request of the broker intermediary unit.

Abstract: Disclosed is a software architecture supporting a large-capacity collective memory layer in a multi-node system by using a remote direct memory access technique and a software virtualization technique and a computing system performing computing processing by using the architecture. In particular, provided is a software architecture including: a memory region managing module collectively managing a predetermined memory region of a node, a memory service providing module providing a large-capacity collective memory service to a virtual address space in a user process, and a memory sharing support module supporting sharing of the large-capacity collective memory of the multi-node system.

Abstract: Disclosed is an apparatus for processing a remote page fault included in an optional local node within a cluster system configuring a large integration memory (CVM) by integrating individual memories of a plurality of nodes. The apparatus includes a memory including a CVM-map, a node memory information table, a virtual memory area, and a CVM page table, and a main controller mapping the large integration memory to an address space of a process when a user process requests memory allocation.

Abstract: Provided are a software reliability test method using selective fault activation, a test area restriction method, a workload generation method and a computing apparatus for testing software reliability using the same. The software reliability test method registers a test target module. The software reliability test method injects a fault into a fault injection target function when a caller of the fault injection target function is included in the registered module, in a case of calling the fault injection target function.

Abstract: A data transfer apparatus, system and method using the same are provided. A data transfer system according to an exemplary embodiment includes a user process space, a kernel space and a hardware space. A plurality of user processes are executed in the user process space. The kernel space includes a kernel thread. The hardware space performs an input/output according to the input/output request of the each user process. When input data based on the input request are received to the hardware space, the data transfer system checks whether the user process requesting the input is in an execution state, and allows the kernel thread to copy the input data from the kernel space to the user process space when the user process is in the execution state.

Abstract: Disclosed is an apparatus for processing a remote page fault included in an optional local node within a cluster system configuring a large integration memory (CVM) by integrating individual memories of a plurality of nodes. The apparatus includes a memory including a CVM-map, a node memory information table, a virtual memory area, and a CVM page table, and a main controller mapping the large integration memory to an address space of a process when a user process requests memory allocation.

Abstract: Disclosed is a software architecture supporting a large-capacity collective memory layer in a multi-node system by using a remote direct memory access technique and a software virtualization technique and a computing system performing computing processing by using the architecture. In particular, provided is a software architecture including: a memory region managing module collectively managing a predetermined memory region of a node, a memory service providing module providing a large-capacity collective memory service to a virtual address space in a user process, and a memory sharing support module supporting sharing of the large-capacity collective memory of the multi-node system.

Abstract: Disclosed are a memory apparatus for a collective volume memory and a method for managing metadata thereof. The memory apparatus for a collective volume memory includes a CVM (Collective Volume Memory) command tool configured to provide a command tool for CVM operation and translate a command input by a user to control the CVM operation; and a CVM engine configured to perform at least one of CVM configuration and initialization, and CVM allocation and access according to data transmitted from the CVM command tool.