Large language models (LLMs) often incorporate multiple text chunks in their inputs to provide the necessary contexts. To speed up the prefill of the long LLM inputs, one can pre-compute the KV cache of a text and re-use the KV cache when the context is reused as the prefix of another LLM input. However, the reused text chunks are not always the input prefix, which makes precomputed KV caches not directly usable since they ignore the text's cross-attention with the preceding texts. Thus, the benefits of reusing KV caches remain largely unrealized.

RAG (Retrieval Augmented Generation) allows LLMs (large language models) to generate better responses with external knowledge, but using more external knowledge causes higher response delay. Prior work focuses either on reducing the response delay (e.g., better scheduling of RAG queries) or on maximizing quality (e.g., tuning the RAG workflow), but they fall short in systematically balancing the tradeoff between the delay and quality of RAG responses. To balance both quality and response delay, this paper presents METIS, the first RAG system that jointly schedules queries and adapts the key RAG configurations of each query, such as the number of retrieved text chunks and synthesis methods. Using four popular RAG-QA datasets, we show that compared to the state-of-the-art RAG optimization schemes, METIS reduces the generation latency by 1.64 – 2.54× without sacrificing generation quality.

As large language models (LLMs) take on complex tasks, their inputs are supplemented with longer contexts that incorporate domain knowledge. Yet using long contexts is challenging as nothing can be generated until the whole context is processed by the LLM. While the context-processing delay can be reduced by reusing the KV cache of a context across different inputs, fetching the KV cache, which contains large tensors, over the network can cause high extra network delays.

Generalizing machine learning (ML) models for network traffic dynamics tends to be considered a lost cause. Hence for every new task, we design new models and train them on model-specific datasets closely mimicking the deployment environments. Yet, an ML architecture called Transformer has enabled previously unimaginable generalization in other domains. Nowadays, one can download a model pre-trained on massive datasets and only fine-tune it for a specific task and context with comparatively little time and data. These fine-tuned models are now state-of-the-art for many benchmarks.