SPDK Based User Space NVMe Over TCP Transport Solution
summary
SPDK implements a full user-space NVMe-oF TCP transport (host and target) that interoperates with the Linux kernel stack and is tested end-to-end with 4K random I/O (70% read / 30% write). Benchmarks show 2.0–2.5× throughput versus a pure kernel NVMe-oF setup and 20–30% lower latency when using the SPDK target. Integrations under development include VPP+DPDK user-space TCP and Intel E810 ADQ queue isolation; code is upstreamed and usable today for further tuning and hardware offloads.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 0:25Agenda
- 0:50Why SPDK
- 5:14Target Code Design
- 12:05TCP Transport Support
- 17:04Service and Performance
- 20:09Efficiency Stack
- 20:52Read Command
- 22:51Performance
- 25:10Latency
- 28:44Ongoing Process
- 31:16Interoperability
- 34:47SPDK Advantages
related talks
39mSecured I/O in the Cloud with NVMe-Over-TCP
31mA Practical, Research‑Backed Introduction To Cloud‑Native Kernel‑Bypass Networking
14mPerf Characteristics of Userspace & Kernel-Space Datapaths in Container Networking
52mSuggestions to Enhance DPDK to Enable Migration of User Space Networking Applications to DPDK
30mZXDH DPU Adapter and It's Application
25mDPDK performance optimization for edge computing scenario
23mDPDK in Databases: Why Isn’t It More Common?
17mRearchitecting the TCP Stack for I/O-Offloaded Content Delivery