Llm Inference Serving easy
Every request to your support bot begins with the same 2,000-token system prompt, followed by a short user message. TTFT is higher than you want. Which optimization targets this pattern directly?
- A Speculative decoding, which uses a small draft model to process the shared system prompt more quickly.
- B A larger batch size, so multiple requests share a single prefill pass over their combined prompts.
- C Prefix caching, which reuses the stored KV cache for the shared system prompt across all requests.
- D Streaming the response, which shows output to the client before the system prompt finishes prefill.
Answer: C. Prefix caching, which reuses the stored KV cache for the shared system prompt across all requests.
**Prefix caching** stores the KV cache for the shared prefix once, so each request only prefills its short user message, which cuts TTFT and saves compute. The tradeoff is that cached prefixes occupy GPU memory that active requests could use, and they give no benefit when prompts are unique. Speculative decoding is tempting, but it speeds up **decode**, not prefill. Streaming cannot show anything until prefill finishes.
One system design question a day, with spaced repetition so the ones you miss come back.
Get the appRelated questions
- Your 70B-class model servers each handle about 40 concurrent chats with 4K-token contexts. Product raises the maximum context to 32K tokens, and many users start using it. What is the main capacity consequence?
- A 70B-class model (80 layers, grouped-query attention with 8 KV heads of dimension 128) is served in fp16 on one 8-GPU node with about 640 GB of total HBM. Product wants 256 concurrent sessions, each holding a full 32K-token context. What is the binding constraint, and what does it imply?
- Your server reserves a contiguous KV region sized for max_tokens = 8,192 for every request at admission, but the average sequence ends near 600 tokens. The node reports out-of-KV-memory at about 20 concurrent requests while GPU compute sits at 30%. Which change best raises concurrency without restricting legitimate long outputs?
- You must summarize 10 million documents overnight, and no user is waiting on any single result. How should you configure the serving stack compared with your interactive chat deployment?
- You serve a 7B-parameter model in fp16 (about 14 GB of weights) for a chat app. Users say single-conversation token generation is too slow, and you are choosing new GPU hardware. Which hardware spec most directly increases per-sequence decode speed?
- Your RAG workload sends 8K-token prompts and gets ~150-token answers from a 70B-class model. You plan to move prefill and decode onto separate GPU pools so long prefills stop stalling decode. What new resource must you size carefully?