A fine-grained memory allocation method

Abstract

The invention discloses a fine-grained memory allocation method, which is characterized by adopting virtual machine type detection, virtual machine page type detection, fine-grained differentiated page allocation strategy and memory access-aware dynamic memory allocation strategy. Due to the distinction between the types of virtual machines, the pages used by I / O-intensive and computing-intensive virtual machines are all small pages. Compared with the system's default strategy of allocating large pages to virtual machines, memory expansion is alleviated. Reduces memory allocation overhead and improves memory deduplication rate; at the same time, for memory-intensive virtual machines, allocating their anonymous pages as large pages can maintain high memory access performance, while the page of memory-intensive virtual machines Cache pages and kernel pages are allocated as small pages. Compared with the system's default policy of allocating large pages, memory expansion is alleviated, memory allocation overhead is reduced, memory deduplication rate is improved, and system performance loss is minimized.

Description

technical field [0001] The invention belongs to the technical field of computer memory management, and specifically relates to a fine-grained implementation of a performance-efficient memory allocation method through a detection mechanism of a virtual machine type and a page type. Background technique [0002] Linux version 2.6.38 introduces the kernel feature of transparent huge pages (Transparent Hugepages), which achieves better performance by improving the efficiency of the processor's memory-mapped hardware. In the Linux memory management system, the size of ordinary pages is 4KB, and these pages are called small pages, while the size of transparent huge pages is 2MB, and these pages are called huge pages. Although large pages have higher memory access performance than small pages, they bring problems such as memory expansion, high allocation overhead, and low memory deduplication rate to the system. During the running of a virtual machine, different virtual machines h...

AI Technical Summary

Problems solved by technology

During the running of the virtual machine, the memory management system does not consider the differences in page access characteristics, but uniformly allocates large pages to the virtual machine. Although this memory allocation mechanism brings performance advantages to the virtual machine, it also Problems such as memory expansion, increased memory allocation overhead, and reduced memory deduplication rate are introduced to the system

Images