4 papers
AlignedServe: Orchestrating Prefix-aware Batching to Build a High-throughput and Computing-efficient LLM Serving System
Fengyao Bai, Hongbin Zhang, Zhitao Chen +3
High-throughput inference serving is essential for applications built on large language models (LLMs). Existing serving frameworks reduce request-level and batch-level bubbles thro…
PipeMax: Enhancing Offline LLM Inference on Commodity GPU Servers
Hongbin Zhang, Taosheng Wei, Jiazhi Jiang +3
Offline LLM inference seeks to maximize request processing under fixed budgets, making commodity GPU servers a promising choice. However, prior work typically considers offloading…
TD-Pipe: Temporally-Disaggregated Pipeline Parallelism Architecture for High-Throughput LLM Inference
Hongbin Zhang, Taosheng Wei, Zhenyi Zheng +3
As the model size continuously increases, pipeline parallelism shows great promise in throughput-oriented LLM inference due to its low demand on communications. However, imbalanced…
EcoServe: Enabling Cost-effective LLM Serving with Proactive Intra- and Inter-Instance Orchestration
Jiangsu Du, Hongbin Zhang, Taosheng Wei +4
Existing LLM serving strategies can be categorized based on whether prefill and decode phases are disaggregated: non-disaggregated (NoDG) or fully disaggregated (FuDG). However, th…