OFFLOADING NVME OVER FABRICS (NVME-OF) TO SMARTNICS ON AN AT-SCALE DISTRIBUTED TESTBED
The rapid growth of cloud deployments has made them foundational to modern com-
puting infrastructure, enabling unprecedented scalability, flexibility, and resource sharing.
However, traditional general-purpose CPUs are burdened with managing critical infrastruc-
ture functions such as storage, security, and networking. SmartNICs have emerged as a
transformative technology, integrating specialized processing cores and accelerators directly
on NIC hardware to offload and accelerate these essential functions. By handling complex
tasks at line speed, SmartNICs enhance throughput, reduce latency, and alleviate CPU
load, effectively addressing the limitations of conventional NICs in cloud environments. This
work explores the offloading of NVMe over Fabrics (NVMe-oF) functionality to SmartNICs,
investigating the performance implications of hardware-accelerated storage management.
The experimental approach of this study leverages SmartNIC technology to measure
performance metrics directly indicative of CPU efficiency, including CPU interrupts, context
switches, and I/O throughput, which demonstrate improved performance through offload-
ing. To investigate if these metrics indicate real-world performance gains for the CPU, this
study benchmarks two database applications, PostgreSQL and MongoDB, and the CPU
performance is measured using completed database transactions and average latency per op-
eration. By comparing the offloaded and non-offloaded configurations, the increase in CPU
efficiency due to SmartNIC offloading in distributed storage infrastructures is quantified.
The number of database operations rises by at least 20% in the least efficient case, with a
corresponding decrease of 10% in database transaction latency when the operations are of-
floaded. These results highlight the potential of SmartNICs in improving CPU performance,
reducing transaction latency, and maximizing resource utilization, ultimately demonstrating
their transformative impact on scalability and performance in modern cloud-native infras-
tructure. This study contributes to understanding the feasibility and benefits of SmartNIC
deployment for NVMe-oF in cloud infrastructures, providing valuable insights into future
cloud optimization strategies
History
Degree Type
- Master of Science
Department
- Computer and Information Technology
Campus location
- West Lafayette