[ENHANCEMENT] Query-Aware Cache Selection for Efficient LLM Serving
还没有人认领这个 Issue。
评估
- 难度
- 5/5
- 预计耗时
- 一周以上
- 新手友好度
- 8/100
- Issue 类型
- 功能
- 描述清晰度
- 需要澄清
- 活跃度
- 停滞
调研方向
This is a design proposal rather than a scoped task: it sketches a query-aware caching layer (profiler, policy manager, scheduler integration) for LLM serving, and the payload names no file, module or test in veScale. A newcomer would first have to confirm whether veScale, a PyTorch training framework, even contains the serving scheduler and cache storage the proposal assumes, and whether the referenced TinyServe paper is meant to be ported. Done would mean an agreed design with maintainers before any code; with no comments or assignee after 456 days, that conversation has not started.
由索引模型根据 Issue 内容生成。
描述
Problem Description
I would like to enhance the cache mechanism for LLM serving in Volcengine. The goal is to make the cache usage more adaptive to heterogeneous query workloads by dynamically adjusting reuse strategies.
Proposed Solution
The proposed enhancement is a Query-Aware Cache Selection (QACS) mechanism:
- Lightweight Query Profiling: Efficiently estimate query complexity (length, entropy, semantic redundancy).
- Dynamic Cache Selection: Adaptively choose between aggressive reuse, partial reuse, or fresh recomputation.
- Policy Control: Expose configurable knobs (
reuse_ratio,cache_budget,latency_accuracy_tradeoff) for fine-grained control. - Scheduler Integration: Integrate into the Volcengine serving scheduler for runtime-aware resource allocation.
Alternatives Considered
- Uniform cache reuse across all queries, which is simple but cannot adapt to different workloads.
- Manual configuration at deployment time, which lacks flexibility and incurs higher operational overhead.
The proposed TinyServe provides adaptive, per-query cache control without requiring manual tuning.
Implementation Plan
- Implement a query profiler module (<1ms overhead).
- Add a policy manager to select caching strategies on the fly.
- Integrate with existing cache storage and serving scheduler.
- Provide developer-facing APIs for configuration.
Additional Context
- This enhancement has been validated at scale across multiple LLMs, consistently improving efficiency and robustness.
- Reference: Dong Liu et al., “TinyServe: Query-Aware Cache Selection for Efficient LLM Serving,” ACM Multimedia 2025 (Oral).
- Expected benefits:
- ⚡ Efficiency: Up to 3x latency reduction under mixed workloads.
- 🎯 Quality: Maintains high generation quality even with aggressive cache usage.
- 📈 Scalability: Supports larger models within the same GPU budget.
- 主要语言
- Python
- 星标
- 1k
- 派生
- 65
- 平均合并
- 21 小时 34 分钟
- 30 天内合并 PR
- 1
环境准备
- 提供 Dockerfile 或 Docker Compose 文件
- 没有 Pull Request 模板
- 阅读贡献指南
从这里开始
- 先读完整个 Issue,再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库,在一个分支上完成修改。
- 提交 Pull Request,并在描述里引用这个 Issue 编号。
volcengine/veScale 的其他 Issue
-
难度 1/5 1 小时以内 新手友好度 12/100
volcengine/veScale#75 ·
-
难度 5/5 一周以上 新手友好度 10/100
volcengine/veScale#73 · 1 条评论 · 3 个 reaction ·
-
难度 5/5 一周以上 新手友好度 8/100
volcengine/veScale#72 · 5 个 reaction ·
-
难度 3/5 1-2 天 新手友好度 30/100
volcengine/veScale#68 ·
-
难度 3/5 1-2 天 新手友好度 12/100
volcengine/veScale#63 ·
查看 volcengine/veScale 的全部 Issue
相似的 Issue
-
bug
难度 2/5 1-3 小时 新手友好度 78/100
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 66/100
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 75/100
docling-project/docling-pipelines#669 ·
维护者通常 1 天内回复
-
blocked kind/enhancement
难度 2/5 1-3 小时 新手友好度 78/100
opendatahub-io/agent-eval-harness#329 · 2 条评论 · 1 个 reaction ·
-
cc_wrapper crashes with ValueError on Windows when a path is on another drive可能已有人在做 @TrapsterDK 今天认领。 未关闭
难度 2/5 1-3 小时 新手友好度 79/100
tweag/rules_haskell#2492 ·
维护者通常 1 天内回复