LegoDiffusion: Micro-Serving Text-to-Image Diffusion Workflows
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866908949411790848 |
|---|---|
| author | Yang, Lingyun Li, Suyi Feng, Tianyu Jiang, Xiaoxiao Di, Zhipeng Lu, Weiyi Liu, Kan Yu, Yinghao Lan, Tao Yang, Guodong Qu, Lin Zhang, Liping Wang, Wei |
| author_facet | Yang, Lingyun Li, Suyi Feng, Tianyu Jiang, Xiaoxiao Di, Zhipeng Lu, Weiyi Liu, Kan Yu, Yinghao Lan, Tao Yang, Guodong Qu, Lin Zhang, Liping Wang, Wei |
| contents | Text-to-image generation executes a diffusion workflow comprising multiple models centered on a base diffusion model. Existing serving systems treat each workflow as an opaque monolith, provisioning, placing, and scaling all constituent models together, which obscures internal dataflow, prevents model sharing, and enforces coarse-grained resource management. In this paper, we make a case for micro-serving diffusion workflows with LegoDiffusion, a system that decomposes a workflow into loosely coupled model-execution nodes that can be independently managed and scheduled. By explicitly managing individual model inference, LegoDiffusion unlocks cluster-scale optimizations, including per-model scaling, model sharing, and adaptive model parallelism. Collectively, LegoDiffusion outperforms existing diffusion workflow serving systems, sustaining up to 3x higher request rates and tolerating up to 8x higher burst traffic. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_08123 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | LegoDiffusion: Micro-Serving Text-to-Image Diffusion Workflows Yang, Lingyun Li, Suyi Feng, Tianyu Jiang, Xiaoxiao Di, Zhipeng Lu, Weiyi Liu, Kan Yu, Yinghao Lan, Tao Yang, Guodong Qu, Lin Zhang, Liping Wang, Wei Distributed, Parallel, and Cluster Computing Artificial Intelligence Text-to-image generation executes a diffusion workflow comprising multiple models centered on a base diffusion model. Existing serving systems treat each workflow as an opaque monolith, provisioning, placing, and scaling all constituent models together, which obscures internal dataflow, prevents model sharing, and enforces coarse-grained resource management. In this paper, we make a case for micro-serving diffusion workflows with LegoDiffusion, a system that decomposes a workflow into loosely coupled model-execution nodes that can be independently managed and scheduled. By explicitly managing individual model inference, LegoDiffusion unlocks cluster-scale optimizations, including per-model scaling, model sharing, and adaptive model parallelism. Collectively, LegoDiffusion outperforms existing diffusion workflow serving systems, sustaining up to 3x higher request rates and tolerating up to 8x higher burst traffic. |
| title | LegoDiffusion: Micro-Serving Text-to-Image Diffusion Workflows |
| topic | Distributed, Parallel, and Cluster Computing Artificial Intelligence |
| url | https://arxiv.org/abs/2604.08123 |