Llm Inference Serving medium
Your autoscaler adds inference replicas when queue depth rises, but a new replica takes about 6 minutes to pull ~140 GB of weights, load them to GPUs, and warm up. Most traffic spikes last 3 to 4 minutes. What change actually helps absorb these spikes?
- A Lower the queue-depth threshold so scale-up triggers earlier, at the first sign of an incoming spike.
- B Keep warm standby headroom and pre-stage weights on local NVMe, accepting the cost of idle GPUs.
- C Switch the scaling signal to GPU utilization, which reacts faster than queue depth during a spike.
- D Scale up by adding tensor-parallel GPUs to running replicas instead of launching new replicas.
Answer: B. Keep warm standby headroom and pre-stage weights on local NVMe, accepting the cost of idle GPUs.
When provisioning time exceeds spike duration, no reactive trigger can help. You need **capacity already loaded**, plus faster weight loading for sustained growth, which trades paying for idle GPUs against SLO misses. An earlier threshold (A) still delivers capacity 6 minutes later, and resharding a live model (D) requires a reload anyway. Reactive scaling is fine for slow diurnal ramps measured in hours.
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?