M2ram:A Memory Compression System Based On Large Page For Cloud Platform

In today's cloud computing environment,in order to improve the utilization rate of physical resources,a single physical server integrates even hundreds of virtual machines to host cloud services.In order to ensure the quality of cloud service,physical servers use large-page mechanism without swap partition to manage memory.However,Hugetlbfs,the existing large-page mechanism of Linux memory management,can not meet the demand of business.Because it more difficult to obtain large pages of contiguous physical space and can't compress the large pages.At the same time,in the actual deployment environment,the physical server memory has much duplicate data.And existing memory deduplication technology(KSM,UKSM)can't support large-page mechanism well,leading to the valuable physical memory resources underutilized.In order to improve the performance of physical server system and the efficiency of physical memory(reuse rate)in the cloud computing environment,and modify the system kernel as slightly as possible,this paper will design and implement a large-page memory compression system M2ram(2M ram)based on a new framework of large-page memory management allocation PHPA(Pristine Huge Page Allocator).It compresses the large page and stores that in memory to improve the physical memory utilization rate.In order to realize the system,this paper has done the following work:(1)Base on a new large-page management framework PHPA.The framework is low coupling with Linux kernel.It can also greatly reduce the large page meta data,further savings available physical memory.(2)Analyze the business scene,and design a complete large-page compression system M2ram which includes hot and cold page tracking,large page recovery,memory compression management and page fault exception handling.Use a multi-stream compression technology,improving the M2ram concurrency capabilities in the NUMA architecture.(3)Creatively put forward a new compressed data management mechanism,achieving a very low space waste and no fragmentation.(4)The innovative parallel large page exception handling and no SwapCache mechanism design,guarantee the high response of the kernel.The simulation experiments and standardized tests show that this compression system achieve a memory reuse rate of more than 2 times.The system's high memory reuse rate,high response and high robustness make it have good industrial application prospects.