<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0">
  <channel>
    <title>土法炼钢兴趣小组的算法知识备份</title>
    <link>https://quant67.com/</link>
    <description>发布来自土法炼钢兴趣小组的知识、笔记、进展和应用。主题包括数据结构和算法、编程语言、网络安全、密码学等。</description>
    <language>zh-CN</language>
    <lastBuildDate>Sun, 12 Jul 2026 14:54:08 +0000</lastBuildDate>

    <item>
      <title>数据库内核实验索引</title>
      <link>https://quant67.com/post/db/index.html</link>
      <guid>https://quant67.com/post/db/index.html</guid>
      <pubDate>Wed, 22 Apr 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[汇总本站数据库内核文章：PostgreSQL / MySQL InnoDB / 列存引擎 / 数据湖与开放表格式 / 流式数据处理 / 分布式 OLAP 查询引擎 / RocksDB 内核 / 向量检索引擎，以及 LSM-Tree 从零实现实验。]]></description>
    </item>
    
    <item>
      <title>【向量检索引擎】向量引擎全景：算法、RAG 与专用引擎之间的一层</title>
      <link>https://quant67.com/post/db/vector-engine/01-vector-engine-overview/01-vector-engine-overview.html</link>
      <guid>https://quant67.com/post/db/vector-engine/01-vector-engine-overview/01-vector-engine-overview.html</guid>
      <pubDate>Sun, 12 Jul 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[定位专用向量检索引擎相对 ANN 算法、RAG 应用与湖仓格式的分工；以 Milvus 2.6.x 四层架构给出本系列 18 篇地图，并交代从 SIGMOD 2021 系统论文到存算分离演进的谱系。]]></description>
    </item>
    
    <item>
      <title>【向量检索引擎】ANN 算法工程接口：从 HNSW/IVF 到 Knowhere 契约</title>
      <link>https://quant67.com/post/db/vector-engine/02-ann-engine-interface/02-ann-engine-interface.html</link>
      <guid>https://quant67.com/post/db/vector-engine/02-ann-engine-interface/02-ann-engine-interface.html</guid>
      <pubDate>Sun, 12 Jul 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[把 HNSW、IVF、DiskANN、Flat 收成引擎侧的索引类型与参数面；说明 Train/Build/Search 生命周期如何贴着 Segment，并与 db-frontier/08 的算法细节、第 8 篇 Knowhere 实现分层分工。]]></description>
    </item>
    
    <item>
      <title>【向量检索引擎】Collection · Partition · Segment · Channel：Growing 到 Sealed 的状态机</title>
      <link>https://quant67.com/post/db/vector-engine/03-segment-model/03-segment-model.html</link>
      <guid>https://quant67.com/post/db/vector-engine/03-segment-model/03-segment-model.html</guid>
      <pubDate>Sun, 12 Jul 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[钉住 Milvus 2.6.x 的数据模型词汇：Collection/Partition、vchannel/pchannel 与 Streaming Node 绑定，Growing/Sealed 段类型、flush 与 handoff，为后续写路径与读路径提供共同状态机。]]></description>
    </item>
    
    <item>
      <title>【向量检索引擎】Proxy 与 Coordinator：接入面、TSO 与集群大脑</title>
      <link>https://quant67.com/post/db/vector-engine/04-proxy-coordinator/04-proxy-coordinator.html</link>
      <guid>https://quant67.com/post/db/vector-engine/04-proxy-coordinator/04-proxy-coordinator.html</guid>
      <pubDate>Sun, 12 Jul 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[按 Milvus 2.6.x 四层架构拆解无状态 Proxy 的校验与 MPP 归并，以及单活跃 Coordinator 的 DDL/TSO、Streaming 绑定、Query 拓扑与离线任务调度；说明控制面与 Worker 的分工边界。]]></description>
    </item>
    
    <item>
      <title>【向量检索引擎】Streaming Node 与 Woodpecker WAL：实时可搜的日志层</title>
      <link>https://quant67.com/post/db/vector-engine/05-streaming-wal/05-streaming-wal.html</link>
      <guid>https://quant67.com/post/db/vector-engine/05-streaming-wal/05-streaming-wal.html</guid>
      <pubDate>Sun, 12 Jul 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[按 Milvus 2.6.x Streaming Service 拆解 Message/TSO、WAL 组件、Recovery Storage 与 Query Delegator；说明 Woodpecker 零本地盘设计与 MemoryBuffer/QuorumBuffer 模式，并标明官方吞吐数字的引用边界。]]></description>
    </item>
    
    <item>
      <title>【向量检索引擎】对象存储上的 Segment 布局：快照、索引与寻址代价</title>
      <link>https://quant67.com/post/db/vector-engine/06-object-storage-layout/06-object-storage-layout.html</link>
      <guid>https://quant67.com/post/db/vector-engine/06-object-storage-layout/06-object-storage-layout.html</guid>
      <pubDate>Sun, 12 Jul 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[说明 Milvus 2.6.x 对象存储持久化的职责边界（日志快照、标量/向量索引、中间结果），Growing flush 后 Sealed 如何进共享存储，以及按字段拆分 binlog 与按段整合（StorageV2/Parquet）在对象存储 API 次数上的工程含义。]]></description>
    </item>
    
    <item>
      <title>【向量检索引擎】Query Node 与 Segcore：段级 search 如何执行</title>
      <link>https://quant67.com/post/db/vector-engine/07-segcore/07-segcore.html</link>
      <guid>https://quant67.com/post/db/vector-engine/07-segcore/07-segcore.html</guid>
      <pubDate>Sun, 12 Jul 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[按 Milvus 2.6.x Data Processing 与 Architecture 拆解 Query Node 对 Sealed 的加载与段级检索，说明 Streaming Node 上 Growing 路径与 Query Delegator 如何拼成一次 search，并界定 Segcore 与 Knowhere 的层次。]]></description>
    </item>
    
    <item>
      <title>【向量检索引擎】Knowhere：向量索引执行引擎与插件契约</title>
      <link>https://quant67.com/post/db/vector-engine/08-knowhere/08-knowhere.html</link>
      <guid>https://quant67.com/post/db/vector-engine/08-knowhere/08-knowhere.html</guid>
      <pubDate>Sun, 12 Jul 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[按官方 Knowhere 文档说明其在 Milvus 中的位置、相对 Faiss 的扩展（bitset、SIMD 选择、二进制度量）、VecIndex 类层次与 IDMAP/IVF/HNSW 等类型，并与 db-frontier/08 的算法细节分工。]]></description>
    </item>
    
    <item>
      <title>【向量检索引擎】分布式 search 归并：Delegator、多级 reduce 与 GuaranteeTs</title>
      <link>https://quant67.com/post/db/vector-engine/09-distributed-search/09-distributed-search.html</link>
      <guid>https://quant67.com/post/db/vector-engine/09-distributed-search/09-distributed-search.html</guid>
      <pubDate>Sun, 12 Jul 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[按 Milvus 2.6.x Data Processing 与 Architecture 拆解 search 的多级归并树：Proxy → Streaming Node Delegator → Query Node 段级结果；说明 shard 并发、Top-k 合并直觉，以及 GuaranteeTs/ServiceTime 如何让一致性变成等待。]]></description>
    </item>
    
    <item>
      <title>【向量检索引擎】Data Node：compaction 与 index build</title>
      <link>https://quant67.com/post/db/vector-engine/10-datanode-compaction/10-datanode-compaction.html</link>
      <guid>https://quant67.com/post/db/vector-engine/10-datanode-compaction/10-datanode-compaction.html</guid>
      <pubDate>Sun, 12 Jul 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[说明 Milvus 2.6.x 中 Data Node 作为离线 Worker 如何承接 Coordinator 下发的建索引与 compaction，输入输出如何进出对象存储，以及索引堆积与 handoff 如何影响查询新鲜度与资源争用。]]></description>
    </item>
    
    <item>
      <title>【向量检索引擎】混合检索与标量过滤：bitset、选择度与引擎落点</title>
      <link>https://quant67.com/post/db/vector-engine/11-hybrid-filter/11-hybrid-filter.html</link>
      <guid>https://quant67.com/post/db/vector-engine/11-hybrid-filter/11-hybrid-filter.html</guid>
      <pubDate>Sun, 12 Jul 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[在 Milvus 引擎路径上落实 filtered ANN：表达式与 bitset 如何进入 Segcore/Knowhere，pre-filter/post-filter 在段级归并下的失效模式，以及对读 db-frontier/09 的 ACORN 争论而不重写算法全文。]]></description>
    </item>
    
    <item>
      <title>【向量检索引擎】一致性模型：Strong 到 Eventually 与 GuaranteeTs</title>
      <link>https://quant67.com/post/db/vector-engine/12-consistency/12-consistency.html</link>
      <guid>https://quant67.com/post/db/vector-engine/12-consistency/12-consistency.html</guid>
      <pubDate>Sun, 12 Jul 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[按官方 Consistency Level 与 Timestamp 文档说明 Strong/Bounded/Session/Eventually 如何映射到 GuaranteeTs，ServiceTime 等待如何影响尾延迟，并对照 2.6 Growing/Sealed 分流下的「刚写入就能搜到」精确含义。]]></description>
    </item>
    
    <item>
      <title>【向量检索引擎】Delete · Upsert · TTL：软删、覆盖写与过期回收</title>
      <link>https://quant67.com/post/db/vector-engine/13-delete-upsert/13-delete-upsert.html</link>
      <guid>https://quant67.com/post/db/vector-engine/13-delete-upsert/13-delete-upsert.html</guid>
      <pubDate>Sun, 12 Jul 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[按官方文档说明 delete（主键/过滤）、upsert（override/merge）、collection TTL 的语义；结合 bitset 软删与 compaction 物理回收，划清「搜索不可见」与「对象存储已清理」的时间差。]]></description>
    </item>
    
    <item>
      <title>【向量检索引擎】副本、负载与故障恢复：内存副本与 WAL 迁移</title>
      <link>https://quant67.com/post/db/vector-engine/14-replica-failover/14-replica-failover.html</link>
      <guid>https://quant67.com/post/db/vector-engine/14-replica-failover/14-replica-failover.html</guid>
      <pubDate>Sun, 12 Jul 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[按官方 In-Memory Replica 与 Streaming Service 说明副本组、shard/historical 副本、Proxy 缓存与 failover；对照 2.6 Streaming Delegator 多副本，以及 WAL Wait for Ready 与 handoff 负载均衡。]]></description>
    </item>
    
    <item>
      <title>【向量检索引擎】Qdrant 对照：单库路径、payload 过滤与分片副本</title>
      <link>https://quant67.com/post/db/vector-engine/15-qdrant-compare/15-qdrant-compare.html</link>
      <guid>https://quant67.com/post/db/vector-engine/15-qdrant-compare/15-qdrant-compare.html</guid>
      <pubDate>Sun, 12 Jul 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[对照 Qdrant 与本系列 Milvus 主线：Rust 段式存储与 payload 索引、分布式下 Raft 管元数据与分片复制；说明何时单进程/小集群更合适，以及分片语义与 Milvus Channel 的差异。]]></description>
    </item>
    
    <item>
      <title>【向量检索引擎】Lance / LanceDB 对照：格式还是服务</title>
      <link>https://quant67.com/post/db/vector-engine/16-lance-compare/16-lance-compare.html</link>
      <guid>https://quant67.com/post/db/vector-engine/16-lance-compare/16-lance-compare.html</guid>
      <pubDate>Sun, 12 Jul 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[承接 lakehouse/21 的 Lance vs Parquet 实测口径，划清湖原生向量格式与 Milvus 专用引擎的边界：随机 take、ANN 随数据集管理、版本化，以及何时不必上向量集群。]]></description>
    </item>
    
    <item>
      <title>【向量检索引擎】生产排障：召回、延迟、堆积、OOM</title>
      <link>https://quant67.com/post/db/vector-engine/17-troubleshooting/17-troubleshooting.html</link>
      <guid>https://quant67.com/post/db/vector-engine/17-troubleshooting/17-troubleshooting.html</guid>
      <pubDate>Sun, 12 Jul 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[按症状反推到 Segment/索引/过滤/一致性/对象存储/副本的检查清单；只使用本系列已建立的机制，不伪造集群指标，区分金标准召回、水位等待与建索队列。]]></description>
    </item>
    
    <item>
      <title>【向量检索引擎】选型与阅读地图：决策树、RAG 回链与开放问题</title>
      <link>https://quant67.com/post/db/vector-engine/18-engine-selection/18-engine-selection.html</link>
      <guid>https://quant67.com/post/db/vector-engine/18-engine-selection/18-engine-selection.html</guid>
      <pubDate>Sun, 12 Jul 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[用决策树收束 Milvus、Qdrant、Lance/pgvector 的适用边界；回链 llm-infra RAG 与 db-frontier ANN；列出 GPU 索引、多向量、稀疏+稠密等开放问题，完成本系列阅读地图。]]></description>
    </item>
    
    <item>
      <title>向量检索引擎 — 系列规划</title>
      <link>https://quant67.com/post/db/vector-engine/PLAN.html</link>
      <guid>https://quant67.com/post/db/vector-engine/PLAN.html</guid>
      <pubDate>Sun, 12 Jul 2026 14:54:08 +0000</pubDate>
      <description><![CDATA[> 本文是写作规划，不是可发布正文。拆解对象：Milvus 2.6.x 主线（存算分离、Proxy / Coordinator / Streaming·Query·Data Node、Segcore + Knowhere）；Qdrant、Lance / LanceDB 作对照。不写 embedding 模型训练，不写…]]></description>
    </item>
    
    <item>
      <title>【向量检索引擎】Milvus · Segcore · Knowhere · Qdrant · Lance</title>
      <link>https://quant67.com/post/db/vector-engine/index.html</link>
      <guid>https://quant67.com/post/db/vector-engine/index.html</guid>
      <pubDate>Sun, 12 Jul 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[补齐 ANN 算法与 RAG 应用之间的生产级向量引擎层：以 Milvus 2.6.x 为主线拆解 Segment、WAL、Segcore、Knowhere、混合过滤与一致性，并用 Qdrant、LanceDB 对照选型。]]></description>
    </item>
    
    <item>
      <title>全部系列索引</title>
      <link>https://quant67.com/post/series/index.html</link>
      <guid>https://quant67.com/post/series/index.html</guid>
      <pubDate>Wed, 22 Apr 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[本站所有系列专题的总目录。按大模型基础设施、Transformer 与注意力机制、数据与数据库、分布式与存储、架构与金融工程、密码学与安全、操作系统与底层、网络与 I/O、算法与可观测等方向分类整理，每个系列均给出简介与入口链接。]]></description>
    </item>
    
    <item>
      <title>【量化交易】量化交易全景：从信号到订单的工程链路</title>
      <link>https://quant67.com/post/quant/01-intro/01-intro.html</link>
      <guid>https://quant67.com/post/quant/01-intro/01-intro.html</guid>
      <pubDate>Fri, 01 May 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[量化交易不是策略写得好就能赚钱，更难的是把数据、特征、因子、信号、组合、执行、风控、复盘这八段链路在工程上连成一条不漏数据、不串时间、不丢订单的流水线。本文是【量化交易】系列的总目录与读图，给出八段链路的输入输出、失败模式、不变量清单，并用研究流程图把从一个想法到一笔实盘订单之间所有该过的卡点串起来。]]></description>
    </item>
    
    <item>
      <title>【量化交易】因子动物园：价值、动量、质量、低波、规模</title>
      <link>https://quant67.com/post/quant/09-factor-zoo/09-factor-zoo.html</link>
      <guid>https://quant67.com/post/quant/09-factor-zoo/09-factor-zoo.html</guid>
      <pubDate>Fri, 01 May 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[从 Fama-French 三因子到当代发表因子动物园，把价值、动量、质量、低波、规模五个方向拆开对照，给出可复用的 Python 因子构造、中性化、IC 评估与 Fama-MacBeth 流水线。文中 A 股量级与 SVG 均为形态示意，须在自有 PIT 样本上复现后才可引用。]]></description>
    </item>
    
    <item>
      <title>【量化交易】回测陷阱：前视偏差、过拟合、数据窥视</title>
      <link>https://quant67.com/post/quant/20-backtest-pitfalls/20-backtest-pitfalls.html</link>
      <guid>https://quant67.com/post/quant/20-backtest-pitfalls/20-backtest-pitfalls.html</guid>
      <pubDate>Fri, 01 May 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[回测引擎只能保证「语法对」，但真正杀死策略的是「逻辑错、数据脏、推断不严」三件事。本文系统拆解前视偏差（lookahead bias）、过拟合（overfitting）、数据窥视（data snooping）三大陷阱，介绍 Bonferroni、BH-FDR、Family-Wise Error 的多重检验修正，给出 Deflated Sharpe 与概率 Sharpe（PSR）的可运行 Python 实现，配一份 30 条上线前自检清单。]]></description>
    </item>
    
    <item>
      <title>【量化交易】做市策略：Avellaneda-Stoikov 与库存管理</title>
      <link>https://quant67.com/post/quant/25-market-making/25-market-making.html</link>
      <guid>https://quant67.com/post/quant/25-market-making/25-market-making.html</guid>
      <pubDate>Fri, 01 May 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[做市不是猜方向，而是为流动性定价。本文从买卖价差与库存成本的经济学起步，沿 Glosten-Milgrom 走到 Avellaneda-Stoikov，讲清 reservation price、optimal spread、HJB 推导、库存惩罚、对抗逆选择（VPIN、订单流毒性）、加密做市的资金费率与跨所对冲、监管红线（操纵 vs 流动性供给），最后给出可运行的 AS 仿真、双边报价器与 VPIN 触发撤单的 Python 实现。]]></description>
    </item>
    
    <item>
      <title>【量化交易】量化交易系统架构：研究、回测、模拟、实盘四套环境</title>
      <link>https://quant67.com/post/quant/27-trading-system-arch/27-trading-system-arch.html</link>
      <guid>https://quant67.com/post/quant/27-trading-system-arch/27-trading-system-arch.html</guid>
      <pubDate>Fri, 01 May 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[量化策略从 notebook 走到实盘要穿过研究、回测、模拟、实盘四套环境，各写一份代码会在上线第一周从接缝处翻车。本文给出接口驱动的统一架构：用六层抽象（数据、特征、信号、组合、订单、风控）让同一份 Strategy 代码靠依赖注入在四套环境间切换，并落地 PIT 数据、订单状态机、对账、事前风控与灰度发布。]]></description>
    </item>
    
    <item>
      <title>量化交易</title>
      <link>https://quant67.com/post/quant/index.html</link>
      <guid>https://quant67.com/post/quant/index.html</guid>
      <pubDate>Fri, 01 May 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[从因子研究到生产执行的量化交易全栈工程。覆盖市场微结构、数据管线、因子构造、组合优化、回测方法论、执行算法、做市策略、高频架构到生产运维。面向策略研究员与工程师。]]></description>
    </item>
    
    <item>
      <title>【列存引擎内核】列存基础与 ClickHouse 架构</title>
      <link>https://quant67.com/post/db/columnar-engine/01-columnar-basics/01-columnar-basics.html</link>
      <guid>https://quant67.com/post/db/columnar-engine/01-columnar-basics/01-columnar-basics.html</guid>
      <pubDate>Thu, 18 Jun 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[行存 vs 列存的带宽、压缩与向量化三角；ClickHouse Server 进程模型、线程池与 MergeTree 引擎家族地图；src/Storages 与 src/Processors 源码入口。对照 PG 行存与 LSM 写优化路径，版本锚定 ClickHouse 24.x LTS。]]></description>
    </item>
    
    <item>
      <title>【列存引擎内核】MergeTree Part 文件格式</title>
      <link>https://quant67.com/post/db/columnar-engine/02-mergetree-part-format/02-mergetree-part-format.html</link>
      <guid>https://quant67.com/post/db/columnar-engine/02-mergetree-part-format/02-mergetree-part-format.html</guid>
      <pubDate>Thu, 18 Jun 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[ClickHouse MergeTree Part 目录结构：columns.txt、checksums.txt、.bin、.mrk2、primary.idx 语义，Granule 与 Mark 的定位作用，Wide/Compact 布局与 MergeTreeDataPart 源码入口。版本锚定 24.x LTS。]]></description>
    </item>
    
    <item>
      <title>【列存引擎内核】压缩与编码</title>
      <link>https://quant67.com/post/db/columnar-engine/03-compression-encoding/03-compression-encoding.html</link>
      <guid>https://quant67.com/post/db/columnar-engine/03-compression-encoding/03-compression-encoding.html</guid>
      <pubDate>Thu, 18 Jun 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[ClickHouse 列压缩：LZ4、ZSTD、Delta、DoubleDelta、Gorilla 时序编码与列类型关系；CODEC 链顺序、LowCardinality 与 PG TOAST 对照。压缩比须本机实测，本文不编造倍数。]]></description>
    </item>
    
    <item>
      <title>【列存引擎内核】向量化执行引擎</title>
      <link>https://quant67.com/post/db/columnar-engine/04-vectorized-execution/04-vectorized-execution.html</link>
      <guid>https://quant67.com/post/db/columnar-engine/04-vectorized-execution/04-vectorized-execution.html</guid>
      <pubDate>Thu, 18 Jun 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[ClickHouse Block 列向量 batch、IProcessor Pipeline 与 filter/project/aggregate 向量实现；对照 PostgreSQL 火山模型 ExecProcNode。源码入口 src/Processors、src/Columns。24.x LTS。]]></description>
    </item>
    
    <item>
      <title>【列存引擎内核】查询读取路径</title>
      <link>https://quant67.com/post/db/columnar-engine/05-query-read-path/05-query-read-path.html</link>
      <guid>https://quant67.com/post/db/columnar-engine/05-query-read-path/05-query-read-path.html</guid>
      <pubDate>Thu, 18 Jun 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[MergeTree SELECT 读路径：Mark Range 定位 Granule、PREWHERE 与 WHERE、Part 级并行与 max_threads。EXPLAIN indexes=1 解读方法。24.x LTS，无伪造 EXPLAIN 输出。]]></description>
    </item>
    
    <item>
      <title>【列存引擎内核】Merge 与 Mutation</title>
      <link>https://quant67.com/post/db/columnar-engine/06-merge-mutation/06-merge-mutation.html</link>
      <guid>https://quant67.com/post/db/columnar-engine/06-merge-mutation/06-merge-mutation.html</guid>
      <pubDate>Thu, 18 Jun 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[MergeTree 后台 merge 触发与 selector、ReplacingMergeTree/CollapsingMergeTree 语义、ALTER UPDATE/DELETE mutation 异步路径与危害。对照 LSM compaction。源码 MergeTreeDataMergerMutator。24.x LTS。]]></description>
    </item>
    
    <item>
      <title>【列存引擎内核】索引与跳数索引</title>
      <link>https://quant67.com/post/db/columnar-engine/07-index-skipping/07-index-skipping.html</link>
      <guid>https://quant67.com/post/db/columnar-engine/07-index-skipping/07-index-skipping.html</guid>
      <pubDate>Thu, 18 Jun 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[ClickHouse PRIMARY KEY 排序键稀疏索引语义、Data Skipping Index（minmax/set/bloom_filter）与 Mark Range 协同；对照 PostgreSQL B-Tree。24.x LTS。]]></description>
    </item>
    
    <item>
      <title>【列存引擎内核】ReplicatedMergeTree</title>
      <link>https://quant67.com/post/db/columnar-engine/08-replicated-mergetree/08-replicated-mergetree.html</link>
      <guid>https://quant67.com/post/db/columnar-engine/08-replicated-mergetree/08-replicated-mergetree.html</guid>
      <pubDate>Thu, 18 Jun 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[ReplicatedMergeTree 副本协调：Log entry、ClickHouse Keeper/ZooKeeper 路径、副本同步与 recovery。双节点实验步骤（本环境未部署）。24.x LTS 默认推荐 Keeper。]]></description>
    </item>
    
    <item>
      <title>【列存引擎内核】Distributed 引擎与分布式查询路由</title>
      <link>https://quant67.com/post/db/columnar-engine/09-distributed-engine/09-distributed-engine.html</link>
      <guid>https://quant67.com/post/db/columnar-engine/09-distributed-engine/09-distributed-engine.html</guid>
      <pubDate>Thu, 18 Jun 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[ClickHouse Distributed 表的分片键、写入路由与 SELECT 下推；GLOBAL IN/JOIN 的代价与替代方案；与 ReplicatedMergeTree 副本层的关系；对照 PG Citus 的边界。]]></description>
    </item>
    
    <item>
      <title>【列存引擎内核】物化视图与增量管道</title>
      <link>https://quant67.com/post/db/columnar-engine/10-materialized-views/10-materialized-views.html</link>
      <guid>https://quant67.com/post/db/columnar-engine/10-materialized-views/10-materialized-views.html</guid>
      <pubDate>Thu, 18 Jun 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[ClickHouse Materialized View 的触发语义、块级增量与目标表引擎选择；Kafka Engine + MV 典型架构；与 PostgreSQL 触发器/MV 的对照及常见坑。]]></description>
    </item>
    
    <item>
      <title>【列存引擎内核】DuckDB 架构与嵌入式 OLAP</title>
      <link>https://quant67.com/post/db/columnar-engine/11-duckdb-architecture/11-duckdb-architecture.html</link>
      <guid>https://quant67.com/post/db/columnar-engine/11-duckdb-architecture/11-duckdb-architecture.html</guid>
      <pubDate>Thu, 18 Jun 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[DuckDB 进程内嵌入式模型、Storage 的 Row Group 与 Column Segment、Catalog 与 checkpoint；与 ClickHouse Server 部署差异及 pg_duckdb 联邦场景边界。]]></description>
    </item>
    
    <item>
      <title>【列存引擎内核】DuckDB 向量化与 Morsel-Driven Pipeline</title>
      <link>https://quant67.com/post/db/columnar-engine/12-duckdb-pipeline/12-duckdb-pipeline.html</link>
      <guid>https://quant67.com/post/db/columnar-engine/12-duckdb-pipeline/12-duckdb-pipeline.html</guid>
      <pubDate>Thu, 18 Jun 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[DuckDB 向量批执行、morsel-driven 并行、Pipeline 调度与 spill；源码 execution/parallel 路径；对照 ClickHouse Processors 与 PG volcano 模型。]]></description>
    </item>
    
    <item>
      <title>【列存引擎内核】ClickHouse 与 DuckDB 选型决策</title>
      <link>https://quant67.com/post/db/columnar-engine/13-ch-vs-duckdb/13-ch-vs-duckdb.html</link>
      <guid>https://quant67.com/post/db/columnar-engine/13-ch-vs-duckdb/13-ch-vs-duckdb.html</guid>
      <pubDate>Thu, 18 Jun 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[从部署形态、规模、并发、联邦与运维成本五维对比 ClickHouse 与 DuckDB；给出决策树与组合架构，不排名不测虚构 benchmark。]]></description>
    </item>
    
    <item>
      <title>【列存引擎内核】监控与系统表</title>
      <link>https://quant67.com/post/db/columnar-engine/14-monitoring/14-monitoring.html</link>
      <guid>https://quant67.com/post/db/columnar-engine/14-monitoring/14-monitoring.html</guid>
      <pubDate>Thu, 18 Jun 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[ClickHouse system.parts、merges、replicas、query_log 与 replication_queue 的字段语义；parts 数、merge 延迟、内存 tracking 的可观测性基线与告警思路。]]></description>
    </item>
    
    <item>
      <title>【列存引擎内核】经典故障模式</title>
      <link>https://quant67.com/post/db/columnar-engine/15-classic-failures/15-classic-failures.html</link>
      <guid>https://quant67.com/post/db/columnar-engine/15-classic-failures/15-classic-failures.html</guid>
      <pubDate>Thu, 18 Jun 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[Too many parts、merge 跟不上 insert、mutation 堆积、副本延迟与 lost replica、max_memory_usage OOM 的症状链、根因与缓解；附测试环境复现框架。]]></description>
    </item>
    
    <item>
      <title>【列存引擎内核】配置陷阱与容量规划</title>
      <link>https://quant67.com/post/db/columnar-engine/16-config-capacity/16-config-capacity.html</link>
      <guid>https://quant67.com/post/db/columnar-engine/16-config-capacity/16-config-capacity.html</guid>
      <pubDate>Thu, 18 Jun 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[parts_to_throw_insert、merge 线程池、max_bytes_to_merge、merge_max_block_size、磁盘 SSD/HDD 策略与内存预算；MergeTree settings 与服务器级 config 的容量规划方法。]]></description>
    </item>
    
    <item>
      <title>列存引擎内核 — 系列规划</title>
      <link>https://quant67.com/post/db/columnar-engine/PLAN.html</link>
      <guid>https://quant67.com/post/db/columnar-engine/PLAN.html</guid>
      <pubDate>Sat, 11 Jul 2026 16:21:07 +0000</pubDate>
      <description><![CDATA[> 本文是写作规划，不是可发布正文。主选 ClickHouse 作为拆解对象，DuckDB 作为嵌入式/OLAP 对照；不写泛泛的「列存概念」科普。]]></description>
    </item>
    
    <item>
      <title>【列存引擎内核】ClickHouse 与 DuckDB 源码级拆解</title>
      <link>https://quant67.com/post/db/columnar-engine/index.html</link>
      <guid>https://quant67.com/post/db/columnar-engine/index.html</guid>
      <pubDate>Thu, 18 Jun 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[主选 ClickHouse 拆解 MergeTree 存储格式、向量化执行与分布式协调；DuckDB 作为嵌入式 OLAP 对照。覆盖列存文件布局、merge 机制、跳数索引与生产故障模式，面向数据平台工程师与从 PG/MySQL 转 OLAP 的 DBA。]]></description>
    </item>
    
    <item>
      <title>算术编码与 ANS：超越 Huffman</title>
      <link>https://quant67.com/post/algorithms/83-arithmetic-coding/arithmetic-coding.html</link>
      <guid>https://quant67.com/post/algorithms/83-arithmetic-coding/arithmetic-coding.html</guid>
      <pubDate>Mon, 11 May 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[用区间细分、重归一化与 ANS 槽位图讲清：Huffman 的 1 比特天花板如何被算术编码和 ANS 打破。]]></description>
    </item>
    
    <item>
      <title>【RocksDB 内核机制】生产嵌入对照：Flink · TiKV · Kafka Streams</title>
      <link>https://quant67.com/post/db/rocksdb/16-embedded-production/16-embedded-production.html</link>
      <guid>https://quant67.com/post/db/rocksdb/16-embedded-production/16-embedded-production.html</guid>
      <pubDate>Tue, 07 Jul 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[从 RocksDB 内核侧补全 Flink KeyGroup 前缀与增量 checkpoint SST 上传机制，对照 TiKV Region 引擎与 Kafka Streams changelog 容错；不重复 stream/12 作业侧全文，给出 Docker 复现步骤而不伪造 checkpoint 指标。]]></description>
    </item>
    
    <item>
      <title>【数据库研究前沿】流批一体与增量视图：Materialize、RisingWave、Feldera 的 DBSP 理论</title>
      <link>https://quant67.com/post/db-frontier/23-incremental/23-incremental.html</link>
      <guid>https://quant67.com/post/db-frontier/23-incremental/23-incremental.html</guid>
      <pubDate>Mon, 15 Jun 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[以 IVM 历史、Differential Dataflow、DBSP（Z-set 与线性化）为主线，对比 Materialize、RisingWave、Feldera 的架构取舍，划清与 Flink/Kafka Streams 的能力边界，并附 Python Z-set 最小增量 join demo]]></description>
    </item>
    
    <item>
      <title>【流式数据处理】流处理全景：从日志到有状态计算</title>
      <link>https://quant67.com/post/stream-processing/01-stream-overview/01-stream-overview.html</link>
      <guid>https://quant67.com/post/stream-processing/01-stream-overview/01-stream-overview.html</guid>
      <pubDate>Wed, 01 Jul 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[从批、流、微批四维度对比出发，建立「可重放日志 + 有状态计算」心智模型，厘清 Lambda/Kappa 边界与流表对偶，并给出与 lakehouse 入湖侧对称的全系列地图。]]></description>
    </item>
    
  </channel>
</rss>
