Guangyu Zhu, Jaehyun Han, Sangjin Lee, Yongseok Son
{"title":"基于Intel Optane DC持久内存模块的nvm感知文件系统的实证评估","authors":"Guangyu Zhu, Jaehyun Han, Sangjin Lee, Yongseok Son","doi":"10.1109/ICOIN50884.2021.9333911","DOIUrl":null,"url":null,"abstract":"Cloud computing arouses the need for large-scale data processing which in turn promises vigorous developments on distributed file systems. The performance of the underlying storage nodes determines the performance of the overall system. Emerging byte-addressable Non-volatile memories (NVM) are promising techniques that can greatly improve storage performance. Researchers have already investigated NVM or NVMaware file systems to take advantage of the characteristics of NVM. However, previous researchers usually perform the studies based on simulations or emulations. In this paper, we provide an empirical evaluation of NVM-aware file systems on the first commercially available byte-addressable NVM (i.e., the Intel Optane DC Persistent Memory Module (DCPMM)) We first evaluate the performance of Ext4, XFS, F2FS, Ext4-DAX, XFSDAX, and NOVA file systems on DCPMM, Optane SSD, and NVMe SSD. Then we compare them in terms of throughput and different access patterns. Second, we observe how remote NUMA node access and device mapper striping affect the performance of DCPMM. We anticipate that the experimental results and performance analysis will provide the implications on various memory and storage systems.","PeriodicalId":6741,"journal":{"name":"2021 International Conference on Information Networking (ICOIN)","volume":"40 1","pages":"559-564"},"PeriodicalIF":0.0000,"publicationDate":"2021-01-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"6","resultStr":"{\"title\":\"An Empirical Evaluation of NVM-aware File Systems on Intel Optane DC Persistent Memory Modules\",\"authors\":\"Guangyu Zhu, Jaehyun Han, Sangjin Lee, Yongseok Son\",\"doi\":\"10.1109/ICOIN50884.2021.9333911\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Cloud computing arouses the need for large-scale data processing which in turn promises vigorous developments on distributed file systems. The performance of the underlying storage nodes determines the performance of the overall system. Emerging byte-addressable Non-volatile memories (NVM) are promising techniques that can greatly improve storage performance. Researchers have already investigated NVM or NVMaware file systems to take advantage of the characteristics of NVM. However, previous researchers usually perform the studies based on simulations or emulations. In this paper, we provide an empirical evaluation of NVM-aware file systems on the first commercially available byte-addressable NVM (i.e., the Intel Optane DC Persistent Memory Module (DCPMM)) We first evaluate the performance of Ext4, XFS, F2FS, Ext4-DAX, XFSDAX, and NOVA file systems on DCPMM, Optane SSD, and NVMe SSD. Then we compare them in terms of throughput and different access patterns. Second, we observe how remote NUMA node access and device mapper striping affect the performance of DCPMM. We anticipate that the experimental results and performance analysis will provide the implications on various memory and storage systems.\",\"PeriodicalId\":6741,\"journal\":{\"name\":\"2021 International Conference on Information Networking (ICOIN)\",\"volume\":\"40 1\",\"pages\":\"559-564\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2021-01-13\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"6\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2021 International Conference on Information Networking (ICOIN)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/ICOIN50884.2021.9333911\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2021 International Conference on Information Networking (ICOIN)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICOIN50884.2021.9333911","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 6
摘要
云计算引起了对大规模数据处理的需求,这反过来又保证了分布式文件系统的蓬勃发展。底层存储节点的性能决定着整个系统的性能。新兴的字节可寻址非易失性存储器(NVM)是一种很有前途的技术,可以极大地提高存储性能。研究人员已经研究了NVM或NVMaware文件系统,以利用NVM的特性。然而,以前的研究人员通常是基于模拟或仿真来进行研究的。在本文中,我们在第一个商业可用的字节可寻址NVM(即Intel Optane DC Persistent Memory Module (DCPMM))上对NVM感知文件系统进行了实证评估。我们首先在DCPMM、Optane SSD和NVMe SSD上评估了Ext4、XFS、F2FS、Ext4- dax、XFSDAX和NOVA文件系统的性能。然后比较它们的吞吐量和不同的访问模式。其次,我们观察了远程NUMA节点访问和设备映射器条带化如何影响DCPMM的性能。我们预计实验结果和性能分析将为各种内存和存储系统提供启示。
An Empirical Evaluation of NVM-aware File Systems on Intel Optane DC Persistent Memory Modules
Cloud computing arouses the need for large-scale data processing which in turn promises vigorous developments on distributed file systems. The performance of the underlying storage nodes determines the performance of the overall system. Emerging byte-addressable Non-volatile memories (NVM) are promising techniques that can greatly improve storage performance. Researchers have already investigated NVM or NVMaware file systems to take advantage of the characteristics of NVM. However, previous researchers usually perform the studies based on simulations or emulations. In this paper, we provide an empirical evaluation of NVM-aware file systems on the first commercially available byte-addressable NVM (i.e., the Intel Optane DC Persistent Memory Module (DCPMM)) We first evaluate the performance of Ext4, XFS, F2FS, Ext4-DAX, XFSDAX, and NOVA file systems on DCPMM, Optane SSD, and NVMe SSD. Then we compare them in terms of throughput and different access patterns. Second, we observe how remote NUMA node access and device mapper striping affect the performance of DCPMM. We anticipate that the experimental results and performance analysis will provide the implications on various memory and storage systems.