Abstract:
Disclosed herein are a method and apparatus for accelerating network transmission in a memory-disaggregated environment. The method for accelerating network transmission in a memory-disaggregated environment includes copying transmission data to a transmission buffer of the computing node, when a page fault occurs during copy of the transmission data, identifying a location at which the transmission data is stored, setting a node in which a largest amount of transmission data is stored as a transmission node, and sending a transmission command to the transmission node.
Abstract:
Disclosed herein is a method and apparatus for supporting virtualization. In the method, conversion of source code of a loadable module is initiated. A virtualization-sensitive instruction is searched for during the conversion of the source code. If the virtualization-sensitive instruction has been found, a virtualization-sensitive instruction table is generated based on the found virtualization-sensitive instruction. The virtualization-sensitive instruction is substituted with an instruction recognizable in a privileged mode, based on the generated virtualization-sensitive instruction table. The loadable module is loaded and executed in a kernel. Accordingly, the present invention supports virtualization, thus minimizing overhead occurring in full virtualization, and guaranteeing the high performance provided by para-virtualization without modifying a source.
Abstract:
A memory access method and device are provided. A memory access method may include: identifying, when an access to a page of a remote memory occurs, a type of the access; allocating a sparse buffer when the access is a sparse write; storing data for the sparse write in the sparse buffer; storing an address for the sparse write as a key and the sparse buffer as a value in a buffer table; and updating an instruction pointer to point to a next instruction.
Abstract:
Apparatus, method and systems for managing reference data, which can prevent duplicated data loading of reference data and eliminate redundancy of I/O operations for loading of the same reference data required by different virtual machines present in the same physical node to reduce use memory and I/O through sharing virtual machine leveled memories, are provided.
Abstract:
A method for memory management, include allocating an empty page of a physical memory for reference data according to execution of an application program, and mapping the empty page to a virtual memory; checking a physical address of the physical memory to which the reference data has been loaded; mapping the checked physical address to the virtual memory to which the empty page has been mapped, and mapping the reference data; and releasing allocation of the allocated physical memory when the reference data is mapped to the virtual memory.