泊头市做网站连云港市连云区建设局网站

泊头市做网站,连云港市连云区建设局网站,姜堰网站开发,网站和网页的概念数据价值的不断升级#xff0c;是过去三十年来数据库演进的核心驱动力。而 AI 的崛起#xff0c;将这一需求推向新的高度#xff1a;数据不仅要能被 看 到#xff0c;更要能被 理解 和 创造------ 这一点已在基于大语言模型#xff0…数据价值的不断升级是过去三十年来数据库演进的核心驱动力。而 AI 的崛起将这一需求推向新的高度数据不仅要能被 看 到更要能被 理解 和 创造------ 这一点已在基于大语言模型LLM为核心的代码生成、智能对话等应用中得以验证。这一背景下由自主 AI 智能体Agent驱动的分析已成为典型范式。 智能体能够独立推理、实时分析数据甚至主动触发行动。这意味着分析模式正从被动报告转向主动决策处理模式也从以查询为中心转向以语义和响应为中心。这一转变对数据基础设施提出巨大挑战工作负载已从 少量用户、繁重查询、慢容忍度 转变为 海量用户智能体、轻量级 / 迭代查询、零延迟容忍度。如果数据库系统无法满足高并发低延迟的查询需求那么其上构建的 AI 智能体就会变得缓慢、笨拙尤其是在一些信息检索的领域产生幻觉给人误导性的结果。因此面向智能体的高并发和低延迟处理能力已不再是可选项而是决定数据仓库能否支撑 AI 时代的生存基石。查询吞吐QPS全面领先进入 AI 时代Apache Doris 继续保持技术领先。4.0 版本实现了与 AI 能力的深度融合增强 AI 原生支持并基于混合搜索技术统一处理结构化过滤、文本搜索、向量语义搜索突破数仓功能界限升级为企业核心的 AI 分析中枢为智能决策和创新实践提供稳定、高效的底层数据支持。不可忽视的是Apache Doris 一直以实时极速著称在性能和吞吐量方面均处于领先水平。因此在 AI 时代这一能力依旧强悍能够高效支持面向 Agent 分析的高并发分析。为了更直观的展示这些能力我们对最当下流行几款数据系统进行评估结果显示结果显示Apache Doris 在每种设置下的表现均优于其他系统。1.1 基础配置我们对 SelectDB基于 Apache Doris 内核构建的现代实时数据仓库、Snowflake 和 Clickhouse Cloud 进行了性能及吞吐量的比较。评测基于 SSB-FLAT、SSB、TPC-H 这三个测试集并借助 Apache JMeter一款开源软件应用程序旨在对功能行为进行负载测试并测量性能进行负载测试。具体测试方法为启动 10/30/50 个线程并按顺序提交查询每个查询运行 3 分钟然后获取每个查询的 QPS。为确保测试的准确性和公平性我们尽可能保证配置规模和定价的一致性。由于各平台对计算资源的命名不尽相同以下是相关配置的简要说明SelectDB 和 Clickhouse Cloud用户可以根据 CPU 核心数选择预期的集群规模。本次评估 SelectDB 和 Clickhouse Cloud 均选择了 128 核集群。Snowflake集群按大小超小、小、中、大、超大衡量。本次评估选择超大X-Large尺寸集群约等于 128 核集群。1.2 测试及结果结论先行在三个基准测试集中SelectDB 在不同并行度10/30/50下的性能及吞吐量均优于 SnowFlake 和 Clickhouse。其中 SSB-FLAT 是一个纯宽表基准测试而 SSB 和 TPC-H 则是包含了表连接的复杂查询测试。通常情况下Clickhouse 在扫描单个宽表时通常表现更快Snowflake 以其更好的弹性扩缩容能力而著称SelectDB 则兼具二者并且在复杂查询和单表查询的场景都进行了针对性的优化。SelectDB 凭借强大的优化器能够重写复杂查询凭借高效的执行引擎来执行查询从而能够在各个并行度的基准测试中表现出了远优于其他系统的并发处理能力。SSB-FLATSSB-FLAT 旨在衡量系统查询单张宽表的能力。在该基准测试中SSB 中所有表被转换为一个非规范化的扁平表且不涉及连接操作。在 10、30、50 三种并行度下SelectDB 均展现出比 Snowflake 和 ClickHouse 更高的 QPS 相比 SnowflakeSelectDB 的 QPS 分别达到其 6.38 倍、7.28 倍、7.39 倍相比 ClickHouseSelectDB 的 QPS 分别达到其 6.92 倍、5.66 倍、4.76 倍。下图直观展示了这一性能对比结果。SSB专为评估数据库对星型模型的查询优化能力而设计。该基准结构简明包含四个查询集、四个维度表和一个简单的汇总层次。在该测试集下在 10、30、50 三种并发条件下SelectDB 的 QPS 分别是 Snowflake 的 6.37 倍、5.98 倍、5.17 倍性能表现显著领先。由于 ClickHouse 在当前测试中无法完整支持 SSB 所需的连接操作未能产生有效可比结果因此在图中将其结果设为 0。TPC-HTPC-H 是业界广泛采用的决策支持系统基准测试。它包含一系列面向业务的即席查询与并发数据更新任务其查询语句与测试数据均经过严谨设计具备广泛的行业代表性。该基准旨在评估系统处理大规模数据、执行复杂查询并辅助关键业务决策的能力。在 10、30、50 三种并发度下SelectDB 的 QPS 分别达到 Snowflake 的 3.10 倍、2.16 倍与 1.71 倍持续保持性能领先。由于 ClickHouse 在部分 TPC-H 查询尤其是 Q20、Q21、Q22中无法完全支持所需的连接操作未能获得有效的可比结果因此在图表中将其设为 0 。完整测试结果可从 SelectDB 官网获取https://www.selectdb.com/blog/1580Apache Doris 为何能够领先承接前文基准测试中展现出的卓越吞吐性能接下来介绍为何 Apache Doris 在高并发查询上能全面领先其他同类型产品其背后有哪些能力或技术支持其能力并非源于单一优化手段而是通过多层协同 ------ 比如高效的数据裁剪、Pipeline 执行模式、向量化执行引擎等共同构筑了支撑海量请求并发的技术基石。下面我们将对其中的几项关键技术进行原理解析。2.1 数据裁剪如何高效处理数据是实时数据仓库中的核心主题之一。在 Apache Doris 中过滤掉不必要的数据只读取最小的数据子集这被称为 数据裁剪是查询加速的主要手段之一。2.1.1 谓词过滤在 Apache Doris 中就生成过滤器的时间而言可将其分为两类静态过滤器和动态过滤器。我们将查询执行前生成的过滤器称为静态过滤器。例如假设用户要查询所有价格大于 10 的饮料 10这一谓词过滤器就可在 SQL 解析阶段推导出来。对于包含内等值连接的查询只有探测侧与构建侧匹配的行才应该被读取。因此这些过滤器只能在构建哈希表之后生成称为动态过滤器。现在我们探讨 Apache Doris 中的静态过滤器 ------ 谓词过滤。对于一张普通的表其列可分为分区列、键列和值列三种类型。针对不同类型的列过滤方式也各不相同对于分区列的谓词 FE 可直接根据元数据判断需要访问哪些分区从而直接在分区级别进行数据裁剪这是最高效的数据裁剪方式。关于键Key列的谓词由于数据在段内是按键列顺序组织的只需根据谓词条件生成键列的上下边界再通过二分查找即可定位需要读取的数据行范围。关于普通列的谓词每个列数据文件都会维护包含最大值 / 最小值的元数据因此可以通过比较谓词条件和元数据来过滤列文件。然后读取剩余列文件并执行谓词计算过滤掉所有不匹配谓词的行。完成谓词过滤后系统获得所有匹配查询条件的行索引。随后只需按行索引加载对应的数据行即可。2.1.2 LIMIT 裁剪另一种数据裁剪的方法是 LIMIT 裁剪。在查询时限定返回行数是常见使用方式具体来说由于限制条件会被下推至查询执行过程中一旦满足该行数限制查询即可提前终止。2.1.3 TopK 裁剪TopK 查询在 BI 查询中广泛使用。简单来说TopK 查询是指根据某些列的顺序检索前 K 个结果与 LIMIT 裁剪类似。但如果使用最基本的方法对数据进行全排序然后取前 K 个结果扫描数据所带来的开销非常大。因此在 Apache Doris 中TopK 通常通过堆排序方法实现。A. 标准堆排序方法处理 TopK 查询的直观方法是标准堆排序方法。核心是维护一个最小堆以实现降序排序。当新数据入堆时会即时更新堆内容。此过程中不在堆排序范围中的数据将被丢弃这意味着无需维护不必要的数据。扫描完成后堆中现有数据便是我们所需的全部结果。B. 理论最优解堆排序的理论最优解指通过扫描数据获取正确结果所需的最小数据量。在 Doris 中数据在段内按键列顺序存储。因此当 TopK 查询的结果按键列排序时我们只需读取每个段的前 K 行然后进行归并排序即可得到最终结果。如果排序结果基于普通列理论最优的方法应是读取每个段的排序数据进行排序并根据排序结果检索相应的数据行而无需读取所有数据进行排序。那么在堆排序过程中如果能够应用一些特殊的优化方法只扫描满足查询条件的数据查询执行的效率将得到极大提升。因此Doris 针对 TopK 查询主要进行了以下优化首先在数据扫描线程中先对数据局部截断然后通过全局协调器对数据进行最终排序并根据排序结果进行全局截断。因此Doris 的 TopK 查询执行过程实际上分为两个阶段第一阶段按照上述方案读取排序列执行局部排序和全局排序得到符合条件的数据的行号。第二阶段根据第一阶段得到的行号读取除排序列之外的所需列从而得到最终输出结果。2.1.4 JOIN 裁剪JOIN 是数据库系统中最耗时的操作数据量越少JOIN 的开销就越低。若暴力执行 JOIN即计算笛卡尔积时间复杂度为 OM*N其中 M 和 N 分别为两个表的大小。因此我们通常选择 Hash Join 作为更高效的连接方法。在 Hash Join 中我们选择较小的数据表作为构建端基于其数据构建哈希表然后用另一侧的表作为探测端来查找哈希表。理想情况下若忽略内存访问的影响构建和探测单行的复杂度为 O1整个哈希连接的复杂度为 OM N。由于探测端的数据通常较大减少探测端数据的读取和计算显得尤为重要。Apache Doris 支持 JOIN 裁剪能够对探测侧数据进行有效裁剪。由于哈希表中构建侧数据的值是确定的可以根据数据量的大小选择合适的 JOIN 裁剪方式。2.2 Pipeline 执行引擎Apache Doris Pipeline 执行引擎的设计目标是能够在查询执行遇到阻塞算子例如Join 和 Shuffle 算子中的磁盘 IO、网络 IO时在用户态主动出让 CPU。这些阻塞算子被称为 Pipeline Breaker。因此每个执行线程可以专注于计算密集型任务尽量减少上下文切换的开销。同时 Pipeline Breaker 的存在使得数据能够均匀重新分布每条 Pipeline 可以独立设置并行度。例如在单线程情况下从两个分片加载数据的扫描算子可以将数据分发到所有具有 N 并行度的下游算子。通过 Pipeline 执行引擎用户可以更高效地处理数据具体收益包括引入本地交换优化充分利用 CPU 资源实现数据均匀分布最大限度减少数据倾斜同时并行性不再受分片数量的限制。多个并发任务共享状态减少额外的初始化开销如表达式和常量变量。所有流水线任务的阻塞条件通过 Dependency 进行封装任务执行逻辑由外部事件如 RPC 完成触发消除阻塞轮询线程的开销。用户可获得更直观的查询 Profile。2.3 向量化执行引擎向量化查询执行是指通过批量处理数据而非逐行处理来提升查询性能的方法。该方法充分利用现代 CPU 架构的优势借助单指令多数据流SIMD操作和循环展开等技术显著提高了 CPU 的数据处理效率。在 Apache Doris 中向量化执行引擎为实际应用场景带来了显著的查询性能提升。数据压缩、循环计算等操作也因此得到大幅加速。结论在本文中我们探讨了 AI 时代数据仓库的现状与前景我们认识到数据在训练和推理中发挥着关键作用。针对这一挑战面向 AI 时代设计的 Apache Doris 4.0 版本应运而生该版本原生支持 MCP Server、向量检索、检索增强生成RAG及 AI 函数等功能。并在查询延迟、吞吐量和成本效益方面均显著优于同类产品成为 AI 时代理想的数据仓库解决方案。附录TPC-H (Parallelism 10)场景SelectDB v4.0.5.1Snowflake v9.36.2ClickHouse v25.4QPSMINAVGMAXError %QPSMINAVGMAXError %QPSMINAVGMAXError %TOTAL5.22-41.06s-0.00%2.24-1min20s-0.00%0.722-3min24s-34.50%Q11.47.05s7.09s7.15s0.00%1.17.64s8.39s9.06s0.00%0.4086.48s15.22s24.50s0.00%Q249.8182ms192ms200ms0.00%25.6301ms343ms389ms0.00%2.321.81s3.37s4.32s0.00%Q310.2946ms965ms985ms0.00%0.7866.43s9.60s12.72s0.00%0.6696.44s12.98s14.95s0.00%Q412.5778ms795ms803ms0.00%1.245.74s6.46s8.08s0.00%0.60697ms8.30s16.50s10.00%Q55.231.89s1.90s1.91s0.00%26.2310ms335ms381ms0.00%0.5368.37s14.03s18.67s90.00%Q664.5148ms152ms154ms0.00%11.2700ms753ms890ms0.00%1.352.44s4.47s7.42s0.00%Q710.4938ms949ms957ms0.00%28.3301ms318ms353ms0.00%0.2999.14s22.88s33.45s10.00%Q89.651.01s1.03s1.03s0.00%20.1272ms341ms496ms0.00%0.32213.48s19.17s31.07s30.00%Q91.865.24s5.26s5.37s0.00%13.3422ms561ms749ms0.00%0.517.97s14.10s19.59s100.00%Q102.773.40s3.51s3.61s0.00%16.9284ms415ms590ms0.00%0.9334.91s8.61s10.72s0.00%Q1124.2349ms399ms413ms0.00%15.6315ms440ms638ms0.00%17246ms51ms58ms100.00%Q1224408ms412ms416ms0.00%1.644.11s4.90s6.09s0.00%0.7013.35s9.00s14.27s0.00%Q132.414.06s4.09s4.15s0.00%0.54113.38s15.74s18.47s0.00%0.22711.24s29.73s44.14s10.00%Q1422.7298ms331ms438ms0.00%1.574.09s4.55s6.36s0.00%2.042.03s3.55s4.91s0.00%Q1510.8924ms926ms929ms0.00%1.714.75s5.52s5.86s0.00%17550ms53ms56ms100.00%Q169.631.03s1.03s1.04s0.00%1.943.38s4.19s5.14s0.00%5.22814ms1.45s1.91s0.00%Q1716.1320ms564ms622ms0.00%2.522.65s3.17s3.98s0.00%0.3947.36s16.28s25.36s10.00%Q181.745.06s5.49s5.74s0.00%0.9884.19s6.88s10.12s0.00%1.243.40s5.53s8.04s0.00%Q1915.9344ms575ms627ms0.00%2.332.57s3.48s4.29s0.00%0.4085.65s14.63s24.51s0.00%Q2022.5394ms429ms445ms0.00%32.5274ms292ms308ms0.00%20428ms41ms48ms100.00%Q212.83.36s3.46s3.57s0.00%27.4310ms325ms365ms0.00%21731ms38ms45ms100.00%Q226.411.51s1.52s1.56s0.00%3.592.32s2.53s2.79s0.00%56.2120ms154ms177ms100.00%TPC-H (Parallelism 30)场景SelectDB v4.0.5.1Snowflake v9.36.2ClickHouse v25.4QPSMINAVGMAX99thError %QPSMINAVGMAX99thError %QPSMINAVGMAX99thError %TOTAL5.01-2min9s--0.00%2.83-2min27s--0.00%1.31-4min59s--48.80%Q11.4220.62s20.86s21.08s21.08s0.00%1.935.13s10.27s15.56s15.56s0.00%0.81412.55s25.40s36.84s36.84s0.00%Q258.6424ms483ms511ms511ms0.00%56.2411ms455ms532ms532ms0.00%1.751.86s11.77s17.17s17.17s0.00%Q37.084.04s4.17s4.24s4.24s0.00%0.9925.21s17.51s30.23s30.23s0.00%1.116.12s15.95s26.93s26.93s60.00%Q411.32.53s2.58s2.65s2.65s0.00%1.534.75s12.14s19.60s19.60s0.00%1.073.15s16.96s28.06s28.06s63.30%Q54.686.35s6.36s6.41s6.41s0.00%63.8397ms431ms468ms468ms0.00%0.99138ms16.91s30.27s30.27s90.00%Q678.9208ms345ms380ms380ms0.00%40368ms548ms748ms748ms0.00%1.982.70s11.06s15.15s15.15s0.00%Q78.832.99s3.29s3.40s3.40s0.00%87.7268ms301ms341ms341ms0.00%0.6317.47s22.91s47.56s47.56s46.70%Q89.013.22s3.30s3.33s3.33s0.00%89.6269ms296ms333ms333ms0.00%0.56811.59s28.21s52.86s52.86s80.00%Q91.6417.97s18.16s18.28s18.28s0.00%89.3260ms295ms332ms332ms0.00%1.023.87s12.35s29.31s29.31s100.00%Q102.6510.62s11.07s11.29s11.29s0.00%55.2396ms448ms541ms541ms0.00%0.9964.97s23.08s30.11s30.11s36.70%Q1123.91.14s1.22s1.25s1.25s0.00%70.6305ms374ms425ms425ms0.00%25459ms80ms117ms117ms100.00%Q1223.31.23s1.25s1.28s1.28s0.00%2.152.88s8.29s13.96s13.96s0.00%1.686.67s14.11s17.90s17.90s0.00%Q132.3612.24s12.63s12.71s12.71s0.00%0.56813.63s33.12s52.78s52.78s0.00%0.4969.88s25.77s1min0s1min0s10.00%Q1433.9781ms837ms883ms883ms0.00%2.622.73s7.12s11.44s11.44s0.00%2.164.17s9.53s13.90s13.90s0.00%Q15112.63s2.70s2.72s2.72s0.00%2.477.09s10.31s12.15s12.15s0.00%50037ms47ms57ms57ms100.00%Q1612.22.22s2.39s2.45s2.45s0.00%2.382.85s7.54s12.60s12.60s0.00%5.441.47s3.69s5.51s5.51s0.00%Q1715.91.02s1.72s1.88s1.88s0.00%2.552.95s7.27s11.78s11.78s0.00%1.287.32s12.33s23.42s23.42s80.00%Q181.7415.51s16.78s17.28s17.28s0.00%1.053.96s16.46s28.69s28.69s0.00%1.27.56s18.18s25.09s25.09s6.70%Q1916746ms1.38s1.87s1.87s0.00%2.42.39s7.68s12.48s12.48s0.00%0.6886.71s30.09s43.61s43.61s0.00%Q2024.9926ms1.10s1.20s1.20s0.00%89.3268ms303ms336ms336ms0.00%53627ms38ms49ms49ms100.00%Q212.5511.40s11.60s11.77s11.77s0.00%85284ms315ms351ms351ms0.00%54537ms43ms53ms53ms100.00%Q226.394.57s4.64s4.69s4.69s0.00%4.093.08s5.11s7.33s7.33s0.00%57166ms318ms511ms511ms100.00%TPC-H (Parallelism 50)场景SelectDB v4.0.5.1Snowflake v9.36.2ClickHouse v25.4QPSMINAVGMAX99thError %QPSMINAVGMAX99thError %QPSMINAVGMAX99thError %TOTAL4.84-3min43s--0.00%2.88-3min43s--0.00%1.68-6min28s--59.30%Q11.4334.43s34.80s35.03s35.03s0.00%1.995.11s15.00s25.11s25.11s0.00%1.0124.68s40.55s49.63s49.63s0.00%Q261.8314ms764ms807ms807ms0.00%88.2405ms469ms564ms564ms0.00%1.8314.03s19.12s27.25s27.25s18.00%Q36.767.20s7.34s7.39s7.39s0.00%0.9754.58s27.89s51.26s51.26s0.00%1.177.17s21.34s42.60s42.60s88.00%Q410.54.54s4.66s4.75s4.75s0.00%1.524.94s18.82s32.85s32.85s0.00%1.175.86s18.59s42.57s42.57s88.00%Q53.6112.89s13.76s13.84s13.84s0.00%78.7413ms491ms632ms632ms0.00%1.225.85s25.81s40.96s40.96s98.00%Q673.2426ms666ms682ms682ms0.00%41.4430ms794ms1.21s1.21s0.00%1.987.17s18.78s25.29s25.29s10.00%Q77.045.73s6.90s7.10s7.10s0.00%124261ms332ms401ms401ms0.00%0.6967.26s25.94s1min12s1min12s84.00%Q88.74612ms5.55s5.72s5.72s0.00%77.2319ms560ms646ms646ms0.00%0.91411.85s27.73s54.73s54.73s84.00%Q91.630.98s31.04s31.22s31.22s0.00%129265ms323ms386ms386ms0.00%1.157.56s20.05s43.44s43.44s100.00%Q102.6116.46s18.72s19.17s19.17s0.00%71.5409ms510ms695ms695ms0.00%1.2715.21s26.13s39.38s39.38s62.00%Q1121.91.08s2.14s2.29s2.29s0.00%96.2329ms445ms520ms520ms0.00%32951ms94ms137ms137ms100.00%Q1223.62.01s2.10s2.12s2.12s0.00%2.212.92s12.78s22.57s22.57s0.00%1.7410.36s23.25s28.63s28.63s0.00%Q132.3221.29s21.49s21.52s21.52s0.00%0.56813.94s51.08s1min28s1min28s0.00%0.95529.81s37.50s52.32s52.32s6.00%Q1432.91.39s1.45s1.51s1.51s0.00%2.642.92s10.89s18.90s18.90s0.00%2.346.55s16.27s21.29s21.29s0.00%Q15114.51s4.53s4.54s4.54s0.00%2.610.88s15.50s19.25s19.25s0.00%47633ms48ms63ms63ms100.00%Q1611.73.91s4.17s4.26s4.26s0.00%2.472.84s11.40s20.21s20.21s0.00%6.251.46s5.42s7.96s7.96s0.00%Q1715.6872ms2.71s3.20s3.20s0.00%2.642.38s10.77s18.94s18.94s0.00%1.547.51s11.89s32.45s32.45s94.00%Q181.7727.02s28.04s28.28s28.28s0.00%1.064.06s25.82s47.06s47.06s0.00%2.155.87s13.32s23.26s23.26s64.00%Q1915.51.21s2.25s3.23s3.23s0.00%2.563.01s11.21s19.50s19.50s0.00%0.9963.80s35.56s50.17s50.17s8.00%Q2026.9775ms1.59s1.85s1.85s0.00%123265ms343ms403ms403ms0.00%69429ms43ms71ms71ms100.00%Q212.3820.09s20.66s20.97s20.97s0.00%123278ms334ms403ms403ms0.00%53826ms45ms67ms67ms100.00%Q226.357.80s7.83s7.87s7.87s0.00%4.183.31s7.70s11.97s11.97s0.00%52.6212ms499ms904ms904ms100.00%Star Schema Benchmark(Parallelism 10)场景SelectDB v4.0.5.1Snowflake v9.36.2ClickHouse v25.4QPSMINAVGMAX99thError %QPSMINAVGMAX99thError %QPSMINAVGMAX99thError %TOTAL14.9-8.14s--0.00%2.34-40.57s--0.00%0.692-1min59s--4.60%Q1.128.7335ms339ms349ms349ms0.00%2.981.58s2.79s3.35s3.35s0.00%1.541.67s3.86s6.51s6.51s0.00%Q1.295.277ms95ms102ms102ms0.00%4.62.09s2.12s2.17s2.17s0.00%2.891.63s2.40s3.46s3.46s0.00%Q1.391.783ms96ms108ms108ms0.00%4.612.10s2.13s2.17s2.17s0.00%5.85878ms1.60s1.71s1.71s0.00%Q2.112.3791ms808ms816ms816ms0.00%2.481.82s2.95s4.03s4.03s0.00%0.6122.92s9.53s16.35s16.35s0.00%Q2.210.9596ms802ms921ms921ms0.00%2.272.13s3.30s4.41s4.41s0.00%1.982.29s3.97s5.06s5.06s0.00%Q2.315.6271ms511ms639ms639ms0.00%3.611.67s2.13s2.77s2.77s0.00%1.172.26s6.06s8.54s8.54s0.00%Q3.16.451.53s1.55s1.55s1.55s0.00%1.572.70s4.40s6.38s6.38s0.00%0.21317.16s27.91s46.85s46.85s20.00%Q3.213.3527ms683ms751ms751ms0.00%2.551.80s2.78s3.93s3.93s0.00%0.3545.80s15.31s28.23s28.23s0.00%Q3.315.4299ms516ms649ms649ms0.00%3.531.73s2.19s2.83s2.83s0.00%0.7195.43s8.59s13.90s13.90s0.00%Q3.452.1176ms183ms192ms192ms0.00%3.221.80s2.31s3.10s3.10s0.00%0.7795.41s9.78s12.84s12.84s0.00%Q4.16.491.53s1.53s1.54s1.54s0.00%1.32.42s4.93s7.68s7.68s0.00%0.5575.24s11.54s17.95s17.95s0.00%Q4.215.2602ms636ms655ms655ms0.00%1.322.82s5.27s7.60s7.60s0.00%0.6724.63s9.44s14.88s14.88s40.00%Q4.323.8351ms389ms420ms420ms0.00%1.961.68s3.27s5.10s5.10s0.00%0.8668.33s9.46s11.55s11.55s0.00%Star Schema Benchmark(Parallelism 30)场景SelectDB v4.0.5.1Snowflake v9.36.2ClickHouse v25.4QPSMINAVGMAX99thError %QPSMINAVGMAX99thError %QPSMINAVGMAX99thError %TOTAL15.3-22.82s--0.00%2.56-1min30s--0.00%1.04-4min0s--22.60%Q1.152.3521ms558ms573ms573ms0.00%4.42.73s4.94s6.81s6.81s0.00%4.383.64s4.80s6.85s6.85s0.00%Q1.2148116ms183ms201ms201ms0.00%5.012.74s4.26s5.98s5.98s0.00%4.87892ms3.93s6.15s6.15s0.00%Q1.3155114ms168ms191ms191ms0.00%4.862.84s4.41s6.17s6.17s0.00%4.311.69s3.95s6.96s6.96s0.00%Q2.112.21.86s2.41s2.46s2.46s0.00%2.71.95s6.48s11.09s11.09s0.00%1.032.98s19.16s29.18s29.18s0.00%Q2.210.9686ms2.41s2.75s2.75s0.00%2.672.16s6.61s11.23s11.23s0.00%1.984.91s11.58s15.14s15.14s0.00%Q2.316.3346ms1.23s1.83s1.83s0.00%3.851.79s4.76s7.79s7.79s0.00%1.984.39s10.53s15.14s15.14s0.00%Q3.16.334.65s4.72s4.74s4.74s0.00%1.742.51s9.82s17.28s17.28s0.00%0.46515.57s36.32s1min04s1min04s33.30%Q3.213.6861ms1.55s2.21s2.21s0.00%2.641.91s6.53s11.37s11.37s0.00%0.6486.21s30.55s46.30s46.30s43.30%Q3.316.2367ms1.29s1.86s1.86s0.00%3.592.01s5.11s8.36s8.36s0.00%0.72911.55s29.18s41.17s41.17s33.30%Q3.461.3234ms435ms486ms486ms0.00%3.571.87s5.09s8.40s8.40s0.00%0.5385.43s34.58s55.80s55.80s40.00%Q4.15.974.69s5.00s5.03s5.03s0.00%1.352.57s12.25s22.17s22.17s0.00%0.67511.49s26.55s44.47s44.47s0.00%Q4.215.21.28s1.77s1.97s1.97s0.00%1.452.38s11.53s20.72s20.72s0.00%1.34.82s15.25s23.01s23.01s73.30%Q4.326.11.05s1.10s1.15s1.15s0.00%1.982.00s8.56s15.11s15.11s0.00%1.388.01s14.11s21.76s21.76s70.00%Star Schema Benchmark(Parallelism 50)场景SelectDB v4.0.5.1Snowflake v9.36.2ClickHouse v25.4QPSMINAVGMAX99thError %QPSMINAVGMAX99thError %QPSMINAVGMAX99thError %TOTAL13.6-43.51s--0.00%2.63-2min18s--0.00%1.48-5min28s--35.70%Q1.160488ms775ms832ms832ms0.00%4.733.28s6.84s10.57s10.57s0.00%5.043.89s7.64s9.92s9.92s0.00%Q1.2158107ms246ms314ms314ms0.00%5.142.46s6.05s9.71s9.71s0.00%6.363.31s5.61s7.78s7.78s0.00%Q1.3172123ms232ms287ms287ms0.00%5.112.28s6.21s9.78s9.78s0.00%7.074.13s5.44s7.04s7.04s0.00%Q2.111.8550ms4.11s4.22s4.22s0.00%2.792.11s9.89s17.90s17.90s0.00%1.313.25s29.38s38.17s38.17s0.00%Q2.210.3631ms4.30s4.83s4.83s0.00%2.751.98s9.99s18.17s18.17s0.00%1.7211.50s21.07s28.98s28.98s0.00%Q2.315.8379ms2.31s3.15s3.15s0.00%3.871.81s7.13s12.94s12.94s0.00%2.1410.06s17.25s23.37s23.37s0.00%Q3.14.7510.11s10.48s10.52s10.52s0.00%1.683.01s16.35s29.75s29.75s0.00%0.88626.90s42.88s56.42s56.42s68.00%Q3.213.2432ms3.05s3.79s3.79s0.00%2.701.88s10.18s18.53s18.53s0.00%0.95931.67s38.50s52.11s52.11s72.00%Q3.316448ms2.10s3.12s3.12s0.00%3.781.94s7.50s13.22s13.22s0.00%0.9524.44s41.67s52.60s52.60s70.00%Q3.463.7330ms664ms784ms784ms0.00%3.632.08s7.82s13.76s13.76s0.00%0.87318.55s42.54s57.24s57.24s66.00%Q4.14.722.96s10.23s10.59s10.59s0.00%1.392.63s19.22s35.90s35.90s0.00%0.88824.83s40.72s56.29s56.29s0.00%Q4.214.61.61s3.09s3.43s3.43s0.00%1.522.50s17.77s32.99s32.99s0.00%2.2611.63s17.34s22.08s22.08s94.00%Q4.325.41.86s1.92s1.97s1.97s0.00%2.111.83s12.68s23.74s23.74s0.00%1.8410.39s17.65s27.19s27.19s94.00%Star Schema Benchmark FLAT (Parallelism 10)场景SelectDB v4.0.5.1Snowflake v9.36.2ClickHouse v25.4QPSMINAVGMAX99thError %QPSMINAVGMAX99thError %QPSMINAVGMAX99thError %TOTAL33.90-3.74s--0.00%5.31-28.66s--0.00%4.90-20.63s--0.00%Q1.135.10283ms284ms285ms285ms0.00%4.901.09s2.04s4.38s4.27s0.00%6.201.36s1.50s1.61s1.61s0.00%Q1.2200.0043ms45ms48ms48ms0.00%8.42639ms1.19s2.23s1.73s0.00%31.9078ms250ms312ms312ms0.00%Q1.3100.0091ms95ms100ms100ms0.00%3.891.51s2.56s4.43s3.38s0.00%41.00213ms226ms243ms243ms0.00%Q2.122.20433ms447ms450ms450ms0.00%5.871.15s1.70s4.05s2.67s0.00%1.811.54s3.96s5.53s5.53s0.00%Q2.225.90377ms381ms386ms386ms0.00%3.941.17s2.53s3.67s3.44s0.00%3.43992ms1.95s2.92s2.92s0.00%Q2.337.50247ms255ms266ms266ms0.00%7.71585ms1.29s2.59s1.92s0.00%5.00992ms1.77s2.00s2.00s0.00%Q3.115.60631ms635ms640ms640ms0.00%2.952.35s3.38s6.72s4.51s0.00%4.601.43s1.96s2.17s2.17s0.00%Q3.227.90346ms351ms358ms358ms0.00%3.571.90s2.79s4.85s3.79s0.00%2.911.14s2.84s3.44s3.44s0.00%Q3.338.50187ms250ms260ms260ms0.00%7.64698ms1.31s3.27s1.89s0.00%5.16626ms1.27s1.94s1.94s0.00%Q3.4147.0052ms63ms68ms68ms0.00%9.38671ms1.06s3.00s1.63s0.00%127.0067ms72ms77ms77ms0.00%Q4.117.50564ms568ms572ms572ms0.00%4.881.28s2.05s4.29s3.07s0.00%2.342.60s3.30s4.28s4.28s0.00%Q4.247.60199ms207ms210ms210ms0.00%2.872.46s3.47s6.21s4.24s0.00%11.00408ms683ms906ms906ms0.00%Q4.359.20156ms161ms168ms168ms0.00%3.032.38s3.29s6.39s4.35s0.00%9.43530ms862ms1.06s1.06s0.00%Star Schema Benchmark FLAT (Parallelism 30)场景SelectDB v4.0.5.1Snowflake v9.36.2ClickHouse v25.4QPSMINAVGMAX99thError %QPSMINAVGMAX99thError %QPSMINAVGMAX99thError %TOTAL38.80-9.78s--0.00%5.33-1min25s--0.00%6.85-37.84s--0.00%Q1.175.40366ms383ms398ms398ms0.00%4.853.93s6.13s12.76s11.80s0.00%17.501.31s1.51s1.71s1.71s0.00%Q1.2462.0052ms58ms65ms65ms0.00%8.621.96s3.46s5.14s4.22s0.00%137.0081ms115ms210ms210ms0.00%Q1.3142.00145ms191ms206ms206ms0.00%3.884.30s7.63s9.72s9.23s0.00%127.0067ms119ms236ms236ms0.00%Q2.126.301.11s1.13s1.14s1.14s0.00%5.913.86s5.03s7.86s6.53s0.00%4.072.41s5.05s7.36s7.36s0.00%Q2.227.401.07s1.08s1.09s1.09s0.00%3.963.65s7.47s10.14s8.97s0.00%4.951.99s4.11s6.02s6.02s0.00%Q2.341.00526ms696ms727ms727ms0.00%7.832.19s3.81s5.59s4.70s0.00%4.231.01s4.59s7.09s7.09s0.00%Q3.115.701.85s1.88s1.91s1.91s0.00%2.974.09s9.90s12.52s11.31s0.00%5.191.82s3.96s5.78s5.78s0.00%Q3.229.60990ms1.01s1.01s1.01s0.00%3.573.42s8.25s10.47s9.32s0.00%3.783.07s5.17s7.94s7.94s0.00%Q3.342.20376ms680ms709ms709ms0.00%7.702.08s3.86s5.47s4.59s0.00%6.781.11s2.72s4.42s4.42s0.00%Q3.4236.00115ms122ms126ms126ms0.00%9.461.83s3.15s4.67s3.83s0.00%283.0053ms70ms106ms106ms0.00%Q4.118.201.63s1.64s1.65s1.65s0.00%4.902.84s6.05s8.26s6.87s0.00%2.821.34s6.97s10.63s10.63s0.00%Q4.252.40549ms557ms571ms571ms0.00%2.883.04s10.21s12.32s11.59s0.00%13.50405ms1.59s2.21s2.21s0.00%Q4.372.30197ms350ms415ms415ms0.00%3.033.45s9.71s12.05s11.20s0.00%9.71921ms1.89s3.07s3.07s0.00%Star Schema Benchmark FLAT (Parallelism 50)场景SelectDB v4.0.5.1Snowflake v9.36.2ClickHouse v25.4QPSMINAVGMAX99thError %QPSMINAVGMAX99thError %QPSMINAVGMAX99thError %TOTAL39.40-15.68s--0.00%5.33-2min19s--0.00%8.27-56.86s--0.00%Q1.191.60386ms519ms544ms544ms0.00%4.828.46s10.18s18.32s17.06s0.00%18.601.84s2.19s2.69s2.69s0.00%Q1.2543.0045ms70ms89ms89ms0.00%8.692.28s5.69s7.48s6.50s0.00%214.0081ms126ms173ms173ms0.00%Q1.3152.00132ms273ms326ms326ms0.00%3.891.97s12.53s16.64s14.17s0.00%260.0056ms85ms140ms140ms0.00%Q2.126.401.84s1.88s1.90s1.90s0.00%5.934.07s8.27s11.84s10.06s0.00%4.043.52s8.90s12.31s12.31s0.00%Q2.227.601.79s1.80s1.81s1.81s0.00%3.973.83s12.26s14.62s13.42s0.00%6.123.90s6.37s8.10s8.10s0.00%Q2.340.9088ms1.01s1.22s1.22s0.00%7.882.25s6.26s8.07s7.12s0.00%4.853.87s6.86s10.31s10.31s0.00%Q3.116.103.08s3.10s3.11s3.11s0.00%2.974.25s16.23s19.38s18.49s0.00%5.833.78s6.18s8.54s8.54s0.00%Q3.229.401.68s1.69s1.70s1.70s0.00%3.573.46s13.57s16.39s15.06s0.00%5.023.19s7.53s9.96s9.96s0.00%Q3.342.60299ms860ms1.17s1.17s0.00%7.702.02s6.39s8.76s7.32s0.00%9.971.16s4.18s4.97s4.97s0.00%Q3.4266.0062ms178ms187ms187ms0.00%9.551.58s5.17s7.16s5.86s0.00%382.0047ms69ms97ms97ms0.00%Q4.117.902.71s2.78s2.79s2.79s0.00%4.882.89s10.01s11.91s11.33s0.00%3.855.04s8.81s12.94s12.94s0.00%Q4.253.00921ms934ms942ms942ms0.00%2.893.73s16.67s19.11s18.41s0.00%16.601.02s2.34s3.01s3.01s0.00%Q4.373.90318ms592ms676ms676ms0.00%3.053.63s15.82s19.09s17.55s0.00%10.20