Skip to content
Draft
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 14 additions & 0 deletions kv_cache_manager/optimizer/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,13 @@

## 概述

## 文档导航

- 优化器使用与能力总览:`kv_cache_manager/optimizer/README.md`
- 分析脚本说明:`kv_cache_manager/optimizer/analysis/script/README.md`
- 混合 workload 分层优化问题定义:`kv_cache_manager/optimizer/docs/mixed_workload_hierarchical_strategy.md`


KVCacheManager Optimizer 是一个独立的缓存优化分析模块,通过回放 trace 数据来模拟缓存读写操作,评估不同驱逐策略和配置对缓存命中率的影响,并为 KVCacheManager 主程序提供参数优化能力。

核心功能包括:
Expand Down Expand Up @@ -375,6 +382,13 @@ bazel run //kv_cache_manager/optimizer/analysis/script:tradeoff_analysis_run_by_

### 概述

## 文档导航

- 优化器使用与能力总览:`kv_cache_manager/optimizer/README.md`
- 分析脚本说明:`kv_cache_manager/optimizer/analysis/script/README.md`
- 混合 workload 分层优化问题定义:`kv_cache_manager/optimizer/docs/mixed_workload_hierarchical_strategy.md`


Optimizer 只接受标准格式的trace文件。使用独立的Python工具将各种trace格式转换为标准格式。

### 标准格式
Expand Down
16 changes: 15 additions & 1 deletion kv_cache_manager/optimizer/analysis/script/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -285,4 +285,18 @@ script/
├── optimizer_runner.py # optimizer 运行封装
├── csv_loader.py # CSV 加载 + 容量列表
└── plot_utils.py # 绘图风格 + Pareto 绘图
```
```
---

## 5. 面向混合 workload 的后续演进

当前脚本集已覆盖单次仿真、容量 tradeoff、树导出与 lifecycle 分析。
在 Super App 混合流量场景下,建议将脚本输出继续用于“workload 识别 → 分层策略决策”闭环,详见:

- `kv_cache_manager/optimizer/docs/mixed_workload_hierarchical_strategy.md`

该文档定义了:

- 面向 KV Cache 行为的 workload 分解问题
- Role-based 与 Cost-based 的职责分工
- 从离线分析到在线决策的落地路线
Loading