BlendServe: Optimizing Offline Inference with Resource-Aware Batching

Offline batch inference is gaining popularity as a cost-effective solution for latency-insensitive tasks, such as model evaluation and data curation. As the latency objective is highly relaxed, maximizing throughput becomes the primary goal in offline inference. Previous studies focused solely on optimizing throughput within a batch. However, the diverse resource demands (compute-intensive vs. memory-intensive) across a wide range of applications make these approaches less effective, as imbalanced resource demands between batches restrict optimization opportunities.

Our insight for achieving optimal throughput is to reorder requests into batches that mix compute-and memoryintensive workloads to maximize resource overlap. However, such a request schedule can conflict with the schedule that maximizes prefix sharing, a widely-used performance optimization, causing suboptimal inference throughput. In this paper, we first build a performance model to analyze request resource demands. Based on it, we design BlendServe, which harmonizes both resource overlapping and prefix sharing to maximize throughput. BlendServe organizes all requests using a resource-aware prefix tree and proposes a dual scanning algorithm to obtain the request schedule. Our evaluation on various models and workloads shows that BlendServe can achieve up to 90% of the optimal throughput.