SplitFS

Abstract

We present SplitFS, a file system for persistent memory (PM) that reduces software overhead significantly compared to state-of-the-art PM file systems. SplitFS presents a novel split of responsibilities between a user-space library file system and an existing kernel PM file system. The user-space library file system handles data operations by intercepting POSIX calls, memory-mapping the underlying file, and serving the read and overwrites using processor loads and stores. Metadata operations are handled by the kernel PM file system (ext4 DAX). SplitFS introduces a new primitive termed relink to efficiently support file appends and atomic data operations. SplitFS provides three consistency modes, which different applications can choose from, without interfering with each other. SplitFS reduces software overhead by up-to 4× compared to the NOVA PM file system, and 17× compared to ext4 DAX. On a number of micro-benchmarks and applications such as the LevelDB key-value store running the YCSB benchmark, SplitFS increases application performance by up to 2× compared to ext4 DAX and NOVA while providing similar consistency guarantees. CCS Concepts• Information systems → Storage class memory; • Hardware → Non-volatile memory; • Software and its engineering → File systems management;

Anisotropic behavior in the lithiation of a silicon nanopillar

“…In cases of SATA/IDE, the target system employs a hardware controller (i.e., disk controller) to manage their storage interface protocol, so the interface driver usually handles I/O interrupt or system memory management. In contrast, in the case of NVMe, a kernel module (NVMe driver) [11], [54], [55] Arrakis [14], [15], [16] Ishiguro et al [29] Aerie [17] RUMA [56] NVMeDirect [12] Moneta-D [20] Direct-FUSE [18] Strata [30] Breeze [57] Simurgh [25] XFUSE [58] SplitFS [21] HyCache [59] Quill [26] Son et al [60], [61] ZoFS [22] Davram [62] vNVML [27], [28] EvFS [19] Kuco [63] DLFS [64] URFS [65] UMFS [31] DevFS [23] CrossFS [24] FSP [32] directly accesses the PCIe bus over a memory mapped I/O and issues the request to the target SSD by composing an nvme_rw_command.…”

“…By examining prior file system designs with respect to the emerging storage techniques, we divide these works into two categories according to the roles of kernel. As shown in Figure 6, several designs [17], [21], [22], [30] rely on kernel to maintain metadata integrity, security, consistency and etc. The kernel is also in charge of performing permission checks on each data request sent from the user space.…”

“…Though the novel design of Strata [30] casts a spark in the field of constructing file systems that utilize multi-level storage media, it does not fully utilize specific features of new storage techniques (e.g., byte-addressability of NVM). Accordingly, Kadekodi et al [21] presents SplitFS to fully utilize the byte-addressability feature of NVM. Similar to Strata, SplitFS splits file systems and responsibilities into user space and kernel space.…”

“…Considering that the involvement of system software in the I/O data path becomes the main reason for the performance degradation of the storage system, multiple prior work [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24] propose to provide user space with direct access to the underlying storage, referred to as user-space storage. Specifically, many research [11], [13] concentrate on implementing user-space storage driver or I/O framework in order to grant the user applications full access to the underlying storage resources.…”

“…Other prior studies [14], [15], [16], [20], [25], [26], [27], [28] consider to decouple the memory management and the system software such that the address translation and boundary check can be served directly from the user space. Furthermore, there are also multiple new designs of file systems [17], [18], [19], [21], [22], [23], [24], [29], [30], [31], [32] to minimize the involvement of kernel in the file accesses. These work propose to construct user-space file systems, which can take over the tasks of the traditional file system.…”