Inference & Serving · core
vLLM & Inference Engines
Request scheduling, model execution, memory management, distributed serving, APIs, and engine architecture.
Mental model
An inference engine is a runtime: it schedules requests, owns model and KV memory, dispatches kernels, exposes metrics, and enforces admission control.
How to study vLLM & Inference Engines
Begin by restating the mental model in your own words, then connect it to a concrete system you have built or operated. Name the mechanism, the constraint it addresses, and the trade-off it introduces. Use Orca: A Distributed Serving System for Transformer-Based Generative Models (OSDI '22), SGLang: Efficient Execution of Structured Language Model Programs, Stanford CS336 — Lecture 10: Inference (Language Modeling from Scratch, Spring 2025) to check details, but close the source before writing your explanation. Retrieval is the learning step; rereading is only preparation.
Next, compare vLLM & Inference Engines with Model Quantization, KV Caching & PagedAttention. Ask what changes in correctness, latency, resource use, operability, and failure recovery. Complete Design exercise: vLLM & Inference Engines and preserve the command, input, output, and one failed attempt as evidence. Finish by explaining the idea without jargon to someone who has not studied the track.
Proof of understanding
- Explain the mechanism from first principles and identify the state it reads or changes.
- Give one situation where the concept is the right choice and one where it is not.
- Predict a realistic failure mode before running the drill, then compare the prediction with evidence.
- Connect the result to a roadmap or build artifact instead of treating the concept as isolated trivia.
Learn from primary sources
- Orca: A Distributed Serving System for Transformer-Based Generative Models (OSDI '22) (doc)
- SGLang: Efficient Execution of Structured Language Model Programs (doc)
- Stanford CS336 — Lecture 10: Inference (Language Modeling from Scratch, Spring 2025) (video)
- Efficient Memory Management for Large Language Model Serving with PagedAttention (paper)
- Fast and Expressive LLM Inference with RadixAttention and SGLang (LMSYS) (article)
- vLLM Documentation (doc)
Practice and explain it back
Design exercise: vLLM & Inference Engines
Request scheduling, model execution, memory management, distributed serving, APIs, and engine architecture. Implement designOutline() returning non-empty values for: requestScheduler, memoryManager, executionBackend. Each value must name a concrete mechanism or decision.
Expected evidence: A design outline with requestScheduler, memoryManager, executionBackend plus an explicit failure mode or trade-off.
Open the interactive drill →Review prompts
- Prefill and decode have very different hardware profiles. Name each and say why it drives the scheduler's design.
Build evidence
Synthesize: Inference & Serving
Build a production mental model for inference engines, memory, kernels, routing, hardware, and serving economics. Produce one working system, benchmark, or evidence-backed design that integrates the path.
- Implements or precisely models the core mechanisms from all three milestones
- Includes at least one injected failure or adversarial case and demonstrates recovery
- Reports quality, latency, resource, reliability, or usability measurements relevant to the domain
- Ships a concise architecture note explaining decisions, trade-offs, and remaining risks
Prerequisites
None assigned yet.