阿里量化分析LoRA适配器共现规律，首个适配器带来66.1%延迟开销
Co-occurrence Patterns of LoRA Adapters in Production Diffusion Model Inference Services
阿里拿生产数据量化LoRA适配器共现规律，66.1%延迟开销这类数字，做推理服务的会想看。
阿里基于生产数据集GenTD26，用图论框架构建LoRA适配器共现网络，从静态结构和动态演化两个维度做量化刻画。数据显示首个适配器引入66.1%的执行延迟开销，之后边际成本递减。共现关系由基础模型驱动，90.6%的多适配器请求中所有适配器共享同一主导基础模型，66.2%的显著共现边连接同模型适配器对。适配器生态呈核心-边缘双极结构，top-10热门模型在12小时窗口内的流失率达54.5%。论文据此提出基于top-k共现统计的预加载策略，k=3时可覆盖81.0%的测试集共现对，为缓存预热、自适应调度和GPU显存管理提供数据依据。
Co-occurrence Patterns of LoRA Adapters in Production Diffusion Model Inference Services
Low-rank adaptation (LoRA) has become a key technology for serving large-scale personalized large language models and diffusion models in the cloud. However, the co-occurrence patterns, resource contention relationships, and evolutionary regularities of adapters under production inference workloads have not been systematically or quantitatively studied. Based on GenTD26, Alibaba's production diffusion model inference dataset, this paper adopts a graph-theoretic framework to construct an adapter co-occurrence network and conducts a characterization from both static structure and dynamic evolution. Our main findings are as follows. (1) The co-occurrence network is extremely sparse, and adapter usage frequency follows a significant heavy-tailed distribution. (2) Introducing the first adapter incurs a 66.1% execution-latency overhead, with diminishing marginal costs afterwards. (3) Co-occurrence relationships are driven by base models: in 90.6% of multi-adapter requests, all adapters share the same dominant base model; 66.2% of significant co-occurrence edges connect same-model adapter pairs; and in 85.8% of multi-adapter requests, all adapter pairs form significant co-occurrence edges. (4) The adapter ecosystem exhibits a core-periphery bipolar structure, with a weekly Jaccard similarity of 0.696 at the model level and a churn rate of 54.5% for the top-10 hottest models within a 12-hour window. Based on these findings, we propose a preloading strategy built on top-k co-occurrence statistics; offline experiments show that it covers 81.0% of test-set co-occurrence pairs at k=3, and sensitivity analyses across frequency thresholds and time windows verify the robustness of the conclusions. These results provide a data-driven basis for cache preloading, adaptive scheduling, and GPU memory management in LoRA inference services.
