xNVMe: Unleashing Storage Hardware-Software Co-design
Fuente:
arXiv
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913575742734336 |
|---|---|
| author | Lund, Simon A. F. Shah, Vivek |
| author_facet | Lund, Simon A. F. Shah, Vivek |
| contents | NVMe SSD hardware has witnessed widespread deployment as commodity and enterprise hardware due to its high performance and rich feature set. Despite the open specifications of various NVMe protocols by the NVMe Express group and NVMe being of software abstractions to program the underlying hardware. The myriad storage I/O paths such as POSIX storage API, ad-hoc OS mechanisms, and userspace I/O libraries have different syntax and semantics that complicate software development and stand in the way of mass adoption and evolution of the NVMe ecosystem. To unify the diverse I/O storage paths, we built xNVMe that exposes a single message-passing API to support both asynchronous and synchronous communication with NVMe devices. xNVMe provides various command sets to support diverse storage I/O paths in different OS (e.g., Linux, FreeBSD, Windows, and MacOS) and userspace libraries (e.g., SPDK) with minimal overhead. xNVMe is an Open Source project and has gained traction amongst various industry stakeholders. In this paper, we elaborate on the lessons that we have learned in the project during its evolution. We also provide some ongoing and future work planned for the project. We hope the database and storage systems community can join in the effort to both extend xNVMe and leverage it as a building block for innovative co-design of storage systems on modern NVMe hardware. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2411_06980 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | xNVMe: Unleashing Storage Hardware-Software Co-design Lund, Simon A. F. Shah, Vivek Operating Systems Databases Distributed, Parallel, and Cluster Computing NVMe SSD hardware has witnessed widespread deployment as commodity and enterprise hardware due to its high performance and rich feature set. Despite the open specifications of various NVMe protocols by the NVMe Express group and NVMe being of software abstractions to program the underlying hardware. The myriad storage I/O paths such as POSIX storage API, ad-hoc OS mechanisms, and userspace I/O libraries have different syntax and semantics that complicate software development and stand in the way of mass adoption and evolution of the NVMe ecosystem. To unify the diverse I/O storage paths, we built xNVMe that exposes a single message-passing API to support both asynchronous and synchronous communication with NVMe devices. xNVMe provides various command sets to support diverse storage I/O paths in different OS (e.g., Linux, FreeBSD, Windows, and MacOS) and userspace libraries (e.g., SPDK) with minimal overhead. xNVMe is an Open Source project and has gained traction amongst various industry stakeholders. In this paper, we elaborate on the lessons that we have learned in the project during its evolution. We also provide some ongoing and future work planned for the project. We hope the database and storage systems community can join in the effort to both extend xNVMe and leverage it as a building block for innovative co-design of storage systems on modern NVMe hardware. |
| title | xNVMe: Unleashing Storage Hardware-Software Co-design |
| topic | Operating Systems Databases Distributed, Parallel, and Cluster Computing |
| url | https://arxiv.org/abs/2411.06980 |