🤖 AI Summary
This paper addresses key challenges in building heterogeneous computing systems through DPU/SmartNIC–CPU co-design. We conduct a systematic survey of over 100 representative works published between 2018 and 2024. Methodologically, we propose the first comprehensive taxonomy for DPU–CPU collaborative computing, categorizing research along three dimensions: hardware architectures (e.g., NVIDIA BlueField, Pensando), programming models (e.g., eBPF, DPDK, SPDK), and offloading mechanisms with coordinated scheduling techniques. Our analysis identifies driving forces behind technological evolution, fundamental bottlenecks—including memory consistency, inter-device communication latency, and software stack fragmentation—and emerging trends toward tighter hardware–software integration. As a result, we construct a domain knowledge graph spanning architectural principles, software stacks, and application scenarios (e.g., AI/ML acceleration, cloud data centers). This work establishes an authoritative benchmark and methodological foundation for co-designed DPU hardware/software development, performance modeling, and domain-specific adaptation.
📝 Abstract
The emergence of new, off-path smart network cards (SmartNICs), known generally as Data Processing Units (DPU), has opened a wide range of research opportunities. Of particular interest is the use of these and related devices in tandem with their host's CPU, creating a heterogeneous computing system with new properties and strengths to be explored, capable of accelerating a wide variety of workloads. This survey begins by providing background information to this new field, such as discussing its origins, its motivations and challenges, listing a few of the current market offerings for DPUs, and providing some brief information about the major programming languages and frameworks for using them. Then, we review and categorize a number of recent works in the field, covering a wide variety of studies, benchmarks, and application areas such as in data center infrastructure, commercial uses, and AI and ML acceleration.