---
title: "Efficient Memory Management for Large Language Model Serving with PagedAttention"
description: "PagedAttention connects a familiar systems idea—virtual-memory paging—to one of LLM serving's hardest constraints: a large, dynamic KV cache. Read it to see how memory layout, sharing, and allocation policy turn directly into higher batching capacity and serving throughput."
canonical_url: "https://fanout.sh/daily/2026-07-21-pagedattention"
md_url: "https://fanout.sh/daily/2026-07-21-pagedattention.md"
last_updated: "2026-07-21"
access: "public"
---

# Efficient Memory Management for Large Language Model Serving with PagedAttention

PagedAttention connects a familiar systems idea—virtual-memory paging—to one of LLM serving's hardest constraints: a large, dynamic KV cache. Read it to see how memory layout, sharing, and allocation policy turn directly into higher batching capacity and serving throughput.

## Paper details

- Authors: Kwon et al.

- Venue: SOSP 2023

- Track: Inference

- Difficulty: Hard

- Reading time: 55 min

- Original paper: https://arxiv.org/abs/2309.06180

## What you will learn

- Why contiguous KV-cache allocation wastes memory through fragmentation and over-reservation.

- How fixed-size KV blocks let vLLM allocate, share, and reclaim cache memory on demand.

- Why better memory utilization raises serving throughput without changing the model itself.

## Continue exploring

[Explore PagedAttention](https://fanout.sh/knowledge-graph?node=ai-research%3Aconcept%3Apagedattention): Open Fanout's PagedAttention node and connect the paper to vLLM, KV-cache management, batching, and high-throughput inference.

---
This representation contains public Fanout content only. Protected Pro lessons, account data, billing, checkout, and pricing are not included.

Browse the public content map: https://fanout.sh/sitemap.md