Llm Inference Serving easy
Your chat assistant returns each response only after generation completes. A typical answer is 400 tokens at about 40 tokens/s, and users say the app feels slow. What is the main effect of switching to token streaming?
- A Total generation time drops substantially, because the server no longer buffers the full response in memory.
- B GPU throughput increases, because tokens leave the server sooner and free their KV cache slots much earlier.
- C Nothing users would notice, because the model still generates exactly the same number of tokens per answer.
- D Users see the first words after roughly the TTFT instead of about 10 s, while total generation time is unchanged.
Answer: D. Users see the first words after roughly the TTFT instead of about 10 s, while total generation time is unchanged.
Streaming cuts **perceived latency**, not total latency. The model still needs about 10 seconds for 400 tokens, but people read along from the first token. The tradeoff is the **error model**: a failure mid-stream leaves partial output on screen. Buffered responses remain the better choice for machine consumers that must validate complete JSON before acting.
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?