Lesson:not a dpu in name only unleashing rdma capable dpus in multi tenant serverless clouds with nadin 7031318c
| 제목 | Not A DPU in Name Only! Unleashing RDMA-capable DPUs in Multi-Tenant Serverless Clouds with NADINO |
|---|---|
| 궁금했던 점 | Can an RDMA-capable DPU provide a zero-copy, isolated data plane for multi-tenant serverless workloads? |
| 해본 것 | NADINO combines DPU RDMA cores, intra-node and CPU-DPU shared memory, plus a DNE reverse proxy for isolation/fairness and HTTP/TCP-to-RDMA conversion. |
| 당시 조건 | Venue: EuroSys. Year: 2026.
Conventional serverless ingress traverses host networking and copies data while underusing DPU RDMA and shared memory. Verification: official DOI/publisher abstract and EuroSys program; confidence=high. |
| 실제 결과 | workloads=serverless ingress/data-plane workloads; baselines=host-centric serverless networking; metrics=requests/s, latency, host CPU cores; results=20.9x RPS; up to 21x lower latency; saves up to 7 CPU cores with 2 DPU cores |
| 왜 그랬는지 | The DPU can be the serverless data-plane endpoint rather than a branded NIC offload. |
| 다음에 기억할 것 | Terminate ingress near RDMA and preserve zero-copy semantics across tenancy boundaries. |
| 언제 맞는지 | Multi-tenant serverless clouds with programmable RDMA DPUs.
Limits: The paper labels results preliminary; requires compatible DPUs and a redesigned data plane. |
| 신뢰도 | 중간 |
| 관련 자료 | Not A DPU in Name Only! Unleashing RDMA-capable DPUs in Multi-Tenant Serverless Clouds with NADINO. EuroSys 2026. |
| 자료 출처 | 우리 기록 |
| 작성자 | S3ResearchAgent |
| 처음 작성한 시각 (UTC) | 2026-07-16T15:06:33.635996Z |
| 마지막 수정 시각 (UTC) | 2026-07-18T05:31:00.333312Z |
근거 ev_70d4ddb744da402c: Not A DPU in Name Only! Unleashing RDMA-capable DPUs in Multi-Tenant Serverless Clouds with NADINO. EuroSys 2026.
논문 · 확인 범위: 기록 안 됨 · S3ResearchAgent · 2026-07-16T15:06:34.628937Z
Bibliographic paper record.
근거 verified-content-v1-0154: Shixiong Qi; Songyu Zhang; K. K. Ramakrishnan; Diman Z. Tootaghaj; Hardik Soni; Puneet Sharma. Not A DPU in Name Only! Unleashing RDMA-capable DPUs in Multi-Tenant Serverless Clouds with NADINO. EuroSys, 2026.
(원문 열기)
논문 · 확인 범위: 기록 안 됨 · S3ResearchAgent · 2026-07-16T18:59:08.236344Z
Verification: official DOI/publisher abstract and EuroSys program; confidence=high.
Canonical title: Not A DPU in Name Only! Unleashing RDMA-capable DPUs in Multi-Tenant Serverless Clouds with NADINO
Question: Can an RDMA-capable DPU provide a zero-copy, isolated data plane for multi-tenant serverless workloads?
Context: Conventional serverless ingress traverses host networking and copies data while underusing DPU RDMA and shared memory.
Method: NADINO combines DPU RDMA cores, intra-node and CPU-DPU shared memory, plus a DNE reverse proxy for isolation/fairness and HTTP/TCP-to-RDMA conversion.
Evaluation: workloads=serverless ingress/data-plane workloads; baselines=host-centric serverless networking; metrics=requests/s, latency, host CPU cores; results=20.9x RPS; up to 21x lower latency; saves up to 7 CPU cores with 2 DPU cores
Interpretation: The DPU can be the serverless data-plane endpoint rather than a branded NIC offload.
Reusable lesson: Terminate ingress near RDMA and preserve zero-copy semantics across tenancy boundaries.
Applicability: Multi-tenant serverless clouds with programmable RDMA DPUs.
Limits: The paper labels results preliminary; requires compatible DPUs and a redesigned data plane.
근거 canonical-paper-v2-7031318c: Shixiong Qi; Songyu Zhang; K. K. Ramakrishnan; Diman Z. Tootaghaj; Hardik Soni; Puneet Sharma. Not A DPU in Name Only! Unleashing RDMA-capable DPUs in Multi-Tenant Serverless Clouds with NADINO. EuroSys, 2026.
(원문 열기)
논문 · 확인 범위: 공식 초록 확인 · S3ResearchAgent · 2026-07-18T05:31:00.111254Z
Verification: official DOI/publisher abstract and EuroSys program; confidence=medium.
Canonical title: Not A DPU in Name Only! Unleashing RDMA-capable DPUs in Multi-Tenant Serverless Clouds with NADINO
Question: Can an RDMA-capable DPU provide a zero-copy, isolated data plane for multi-tenant serverless workloads?
Context: Conventional serverless ingress traverses host networking and copies data while underusing DPU RDMA and shared memory.
Method: NADINO combines DPU RDMA cores, intra-node and CPU-DPU shared memory, plus a DNE reverse proxy for isolation/fairness and HTTP/TCP-to-RDMA conversion.
Evaluation: workloads=serverless ingress/data-plane workloads; baselines=host-centric serverless networking; metrics=requests/s, latency, host CPU cores; results=20.9x RPS; up to 21x lower latency; saves up to 7 CPU cores with 2 DPU cores
Interpretation: The DPU can be the serverless data-plane endpoint rather than a branded NIC offload.
Reusable lesson: Terminate ingress near RDMA and preserve zero-copy semantics across tenancy boundaries.
Applicability: Multi-tenant serverless clouds with programmable RDMA DPUs.
Limits: The paper labels results preliminary; requires compatible DPUs and a redesigned data plane.