Keywords: LLM inference; batch inference; relational data analytics
TL;DR: This paper presents techniques to optimize LLM inference in relational batch data analytics, achieving significant improvements in end-to-end latency and cost reductions while preserving query semantics.
Abstract: Batch data analytics is a growing application for Large Language Models (LLMs). LLMs enable users to perform a wide range of natural language tasks, such as classification, entity extraction, and translation, over large datasets. However, LLM inference is highly costly and slow: for example, an NVIDIA L4 GPU running Llama3-8B can only process 6 KB of text per second, taking about a month to handle 15 GB of data; processing a similar amount of data costs around $18K on OpenAI’s GPT-4o. In this paper, we propose novel techniques that can significantly reduce the cost of LLM calls for relational data analytics workloads. Our key contribution is developing efficient algorithms for reordering the rows and the fields within each row of an input table to maximize key-value (KV) cache reuse when performing LLM serving. As such, our approach can be easily applied to existing analytics systems and serving platforms. Our evaluation shows that our solution can yield up to 3.4× improvement in job completion time on a benchmark of diverse LLM-based queries using Llama 3 models. Our solution also achieves a 32% cost savings under OpenAI and Anthropic pricing models.
Supplementary Material: pdf
Submission Number: 252
Loading