NOVA: NVM驱动的LFS文件系统应对高并发挑战
需积分: 22 78 浏览量
更新于2024-07-18
收藏 2.26MB PDF 举报
本文档是关于"NOVA:一种面向混合挥发性/非挥发性主内存的逻辑结构文件系统"的学术论文，发表于第14届USENIX文件和存储技术会议(FAST'16)。该会议于2016年2月22日至25日在加利福尼亚州圣克拉拉举行，由USENIX协会主办。论文的ISBN为978-1-931971-28-7，全文开放获取。
NOVA系统的设计目标是充分利用即将出现的非易失性内存(NVM)与动态随机存取内存(DRAM)之间的协同工作，为软件提供亚微秒级的高带宽持久数据访问。然而，由于NVM的数据管理、访问和一致性维护带来了诸多挑战，传统的文件系统并不完全适应这种新型内存环境。
作者 Jian Xu 和 Steven Swanson 来自加州大学圣地亚哥分校，他们提出了一种创新的文件系统设计，它采用日志结构，旨在解决在混合内存系统中存储和处理数据时遇到的问题。日志结构的优势在于能够实现高效的写入操作，同时保持数据的一致性和可靠性，这对于支持NVM的高并发且原子性的数据操作至关重要。
论文详细探讨了NOVA如何通过利用NVM的特性，如低延迟、非易失性以及大容量，来优化数据布局、缓存策略以及事务处理机制。此外，NOVA还可能涉及一致性协议、数据迁移策略以及如何在内存失效或故障恢复时保持系统的可用性。
作者可能分析了现有的文件系统在面对NVM时的局限性，并提出了NOVA在这些方面的改进措施，例如通过细粒度的持久化控制、元数据管理以及数据分片，来提升系统的性能和鲁棒性。论文可能还会展示NOVA在真实环境中的实验结果，以证明其在实际应用场景中的有效性。
总结来说，这篇论文不仅提供了关于NOVA文件系统的技术细节，还为理解如何设计和优化未来的内存管理系统，特别是在混合内存架构下，提供了有价值的洞见。对于IT专业人士和研究者来说，这篇论文是深入理解NVM和高性能文件系统交互的重要参考资料。
USENIX Association 14th USENIX Conference on File and Storage Technologies (FAST ’16) 325
systems [21, 71, 73] bypass the DRAM page cache and access
NVMM directly using a technique calle d Direct Access (DAX)
or eXecute In Place (XIP), avoiding extra copies between
NVMM and DRAM in the storage stack. NOVA is a DAX
file system and we expect that all NVMM file systems will
provide these (or similar) features. We describe currently
available DAX file systems in Section 2.4.
Write reordering Modern processors and their caching
hierarchies may reorder store operations to improve perfor-
mance. The CPU’s memory consistency protocol makes guar-
antees about the ordering of memory updates, but existing
models (with the exception of research proposals [20, 46]) do
not provide guarantees on when updates will reach NVMMs.
As a result, a power failure may leave the data in an inconsis-
tent state.
NVMM-aware software can avoid this by explicitly flush-
ing caches and issuing memory barriers to enforce write
ordering. The x86 architecture provides the clflush in-
struction to flush a CPU cacheline, but clflush is strictly
ordered and needlessly invalidates the cacheline, incurring a
significant performance penalty [6, 76]. Also, clflush only
sends data to the memory controller; it does not guarantee
the data will reach memory. Memory barriers such as Intel’s
mfence instruction enforce order on memory operations be-
fore and after the barrier, but mfence only guarantees all
CPUs have the same view of the memory. It does not impose
any constraints on the order of data writebacks to NVMM.
Intel has proposed new instructions that fix these prob-
lems, including clflushopt (a more efficient version of
clflush), clwb (to explicitly write back a cache line with-
out invalidating it) and PCOMMIT (to force stores out to
NVMM) [26, 79]. NOVA is built with these instructions
in mind. In our evaluation we use a hardware NVMM emu-
lation system that approximates the performance impacts of
these instructions.
Atomicity POSIX-style file system semantics require
many operations to be atomic (i.e., to execute in an “all or
nothing” fashion). For example, the POSIX rename re-
quires that if the operation fails, neither the file with the old
name nor the file with the new name shall be changed or
created [53]. Renaming a file is a metadata-only operation,
but some atomic updates apply to both file system metadata
and data. For instance, appending to a file atomically updates
the file data and changes the file’s length and modification
time. Many applications rely on atomic file system operations
for their own correctness.
Storage devices typically provide only rudimentary guaran-
tees about atomicity. Disks provide atomic sector writes and
processors guarantee only that 8-byte (or smaller), aligned
stores are atomic. To build the more complex atomic up-
dates that file systems require, programmers must use more
complex techniques.
2.3. Building complex atomic operations
Existing file systems use a variety of techniques like journal-
ing, shadow paging, or log-structuring to provide atomicity
guarantees. These work in different ways and incur different
types of overheads.
Journaling Journaling (or write-ahead logging) is widely
used in journaling file systems [24, 27, 32, 71] and
databases [39, 43] to ensure atomicity. A journaling system
records all updates to a journal before applying them and, in
case of power failure, replays the journal to restore the system
to a consistent state. Journaling requires writing data twice:
once to the log and once to the target location, and to im-
prove performance journaling file systems usually only jour-
nal metadata. Recent work has proposed back pointers [17]
and decoupling ordering from durability [16] to reduce the
overhead of journaling.
Shadow paging Several file systems use a copy-on-write
mechanism called shadow paging [20, 8, 25, 54]. Shadow
paging file systems rely heavily on their tree structure to
provide atomicity. Rather than modifying data in-place during
a write, shadow paging writes a new copy of the affected
page(s) to an empty portion of the storage device. Then, it
splices the new pages into the file system tree by updating
the nodes between the pages and root. The resulting cascade
of updates is potentially expensive.
Log-structuring Log-structured file systems (LFSs) [55,
60] were originally designed to exploit hard disk drives’ high
performance on sequential accesses. LFSs buffer random
writes in memory and convert them into larger, sequential
writes to the disk, making the best of hard disks’ strengths.
Although LFS is an elegant idea, implementing it effi-
ciently is complex, because LFSs rely on writing sequentially
to contiguous free regions of the disk. To ensure a consistent
supply of such regions, LFSs constantly clean and compact
the log to reclaim space occupied by stale data.
Log cleaning adds overhead and degrades the performance
of LFSs [3, 61]. To reduce cleaning overhead, some LFS
designs separate hot and cold data and apply different clean-
ing policies to each [69, 70]. SSDs also perform best under
sequential workloads [9, 14], so LFS techniques have been
applied to SSD file systems as well. SFS [38] classifies file
blocks based on their update likelihood, and writes blocks
with similar “hotness” into the same log segment to reduce
cleaning overhead. F2FS [30] uses multi-head logging, writes
metadata and data to separate logs, and writes new data di-
rectly to free space in dirty segments at high disk utilization
to avoid frequent garbage collection.
剩余16页未读，继续阅读
2021-05-13 上传
2021-06-09 上传
2021-05-27 上传
2021-05-27 上传
2021-06-09 上传
2021-06-08 上传
2015-01-18 上传
-Kylin
- 粉丝: 2
- 资源: 3
我的内容管理
展开
最新资源
- 前端协作项目：发布猜图游戏功能与待修复事项
- Spring框架REST服务开发实践指南
- ALU课设实现基础与高级运算功能
- 深入了解STK：C++音频信号处理综合工具套件
- 华中科技大学电信学院软件无线电实验资料汇总
- CGSN数据解析与集成验证工具集：Python和Shell脚本
- Java实现的远程视频会议系统开发教程
- Change-OEM: 用Java修改Windows OEM信息与Logo
- cmnd：文本到远程API的桥接平台开发
- 解决BIOS刷写错误28：PRR.exe的应用与效果
- 深度学习对抗攻击库：adversarial_robustness_toolbox 1.10.0
- Win7系统CP2102驱动下载与安装指南
- 深入理解Java中的函数式编程技巧
- GY-906 MLX90614ESF传感器模块温度采集应用资料
- Adversarial Robustness Toolbox 1.15.1 工具包安装教程
- GNU Radio的供应商中立SDR开发包：gr-sdr介绍
