FLYING SERVING: On-the-Fly Parallelism Switching for Large Language Model Serving
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866915827107758080 |
|---|---|
| author | Gao, Shouwei Yin, Junqi Wang, Feiyi Dong, Wenqian |
| author_facet | Gao, Shouwei Yin, Junqi Wang, Feiyi Dong, Wenqian |
| contents | Production LLM serving must simultaneously deliver high throughput, low latency, and sufficient context capacity under non-stationary traffic and mixed request requirements. Data parallelism (DP) maximizes throughput by running independent replicas, while tensor parallelism (TP) reduces per-request latency and pools memory for long-context inference. However, existing serving stacks typically commit to a static parallelism configuration at deployment; adapting to bursts, priorities, or long-context requests is often disruptive and slow. We present Flying Serving, a vLLM-based system that enables online DP-TP switching without restarting engine workers. Flying Serving makes reconfiguration practical by virtualizing the state that would otherwise force data movement: (i) a zero-copy Model Weights Manager that exposes TP shard views on demand, (ii) a KV Cache Adaptor that preserves request KV state across DP/TP layouts, (iii) an eagerly initialized Communicator Pool to amortize collective setup, and (iv) a deadlock-free scheduler that coordinates safe transitions under execution skew. Across three popular LLMs and realistic serving scenarios, Flying Serving improves performance by up to $4.79\times$ under high load and $3.47\times$ under low load while supporting latency- and memory-driven requests. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_22593 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | FLYING SERVING: On-the-Fly Parallelism Switching for Large Language Model Serving Gao, Shouwei Yin, Junqi Wang, Feiyi Dong, Wenqian Distributed, Parallel, and Cluster Computing Production LLM serving must simultaneously deliver high throughput, low latency, and sufficient context capacity under non-stationary traffic and mixed request requirements. Data parallelism (DP) maximizes throughput by running independent replicas, while tensor parallelism (TP) reduces per-request latency and pools memory for long-context inference. However, existing serving stacks typically commit to a static parallelism configuration at deployment; adapting to bursts, priorities, or long-context requests is often disruptive and slow. We present Flying Serving, a vLLM-based system that enables online DP-TP switching without restarting engine workers. Flying Serving makes reconfiguration practical by virtualizing the state that would otherwise force data movement: (i) a zero-copy Model Weights Manager that exposes TP shard views on demand, (ii) a KV Cache Adaptor that preserves request KV state across DP/TP layouts, (iii) an eagerly initialized Communicator Pool to amortize collective setup, and (iv) a deadlock-free scheduler that coordinates safe transitions under execution skew. Across three popular LLMs and realistic serving scenarios, Flying Serving improves performance by up to $4.79\times$ under high load and $3.47\times$ under low load while supporting latency- and memory-driven requests. |
| title | FLYING SERVING: On-the-Fly Parallelism Switching for Large Language Model Serving |
| topic | Distributed, Parallel, and Cluster Computing |
| url | https://arxiv.org/abs/2602.22593 |