{"id":10006791,"date":"2022-11-03T21:49:18","date_gmt":"2022-11-03T13:49:18","guid":{"rendered":"https:\/\/ityet.com\/?p=10006791"},"modified":"2022-11-03T21:49:18","modified_gmt":"2022-11-03T13:49:18","slug":"%e6%95%b0%e6%8d%ae%e5%ba%93%e7%ad%89%e5%80%bc%e6%9f%a5%e8%af%a2%e4%b8%8e%e7%bb%9f%e8%ae%a1%e4%bf%a1%e6%81%af","status":"publish","type":"post","link":"https:\/\/itmax.cn\/?p=10006791","title":{"rendered":"\u6570\u636e\u5e93\u7b49\u503c\u67e5\u8be2\u4e0e\u7edf\u8ba1\u4fe1\u606f"},"content":{"rendered":"<p>\u7b80\u4ecb\uff1a\u00a0\u7edf\u8ba1\u4fe1\u606f\u662f\u4e3a\u4f18\u5316\u5668\u7684 cost \u4f30\u7b97\u63d0\u4f9b\u6570\u636e\u652f\u6491\uff0c\u5176\u4e2d\u5f88\u91cd\u8981\u7684\u4e00\u70b9\u9700\u6c42\u4fbf\u662f\u7b49\u503c\u67e5\u8be2(EQUALS, IN \u7b49) \u573a\u666f\u4e0b\u7684\u57fa\u6570\u4f30\u7b97\u3002\u6982\u5ff5\u7edf\u8ba1\u4fe1\u606f\u662f\u4e3a\u4f18\u5316\u5668\u7684 cost \u4f30\u7b97\u63d0\u4f9b\u6570\u636e\u652f\u6491\uff0c\u5176\u4e2d\u5f88\u91cd\u8981\u7684\u4e00\u70b9\u9700\u6c42\u4fbf\u662f\u7b49\u503c\u67e5\u8be2(EQUALS, IN \u7b49) \u573a\u666f\u4e0b\u7684\u57fa\u6570\u4f30\u7b97\u3002\u8003\u8651\u4ee5\u4e0b CaseCREATE TABLE <code>mc_tac_template<\/code> (<br \/>  <code>ID<\/code> BIGINT ,<br \/>  <code>NAME<\/code> varchar(50) NOT NULL,<br \/>  <code>GENDER<\/code> varchar(10) NOT NULL,<br \/>  PRIMARY KEY (<code>ID<\/code>),<br \/>  KEY <code>KEY_NAME<\/code> USING BTREE (<code>NAME<\/code>),<br \/>  KEY <code>KEY_GENDER<\/code> USING BTREE (<code>GENDER<\/code>)<br \/>) <\/p>\n<p>SELECT * FROM User WHERE Name=&#8217;wy&#8217; AND Gender=&#8217;female&#8217;\u6211\u4eec\u7684\u5ba2\u6237\u540c\u65f6\u4e3a\u6027\u522b\u548c\u59d3\u540d\u8bbe\u7f6e\u4e86\u7d22\u5f15\uff0c\u8fd9\u65f6\u4f18\u5316\u5668\u5c31\u8981\u57fa\u4e8e\u5404\u4e2a\u6761\u4ef6\u8fc7\u6ee4\u884c\u6570\u591a\u5be1\u6765\u4f30\u7b97\u9009\u62e9\u4e0d\u540c\u7d22\u5f15\u7684\u4ee3\u4ef7\u5f00\u9500\u3002\u5c31\u8fd9\u4e2a case \u800c\u8a00\uff0c\u5982\u679c\u9009\u5230\u4e86 gender \u7684\u7d22\u5f15\uff0c\u67e5\u8be2\u8fc7\u7a0b\u5f88\u53ef\u80fd\u5c31\u4e0d\u592a\u5999\u4e86\u3002\u5982\u679c\u7edf\u8ba1\u4fe1\u606f\u53ef\u4ee5\u63d0\u4f9b\u7cbe\u786e\u57fa\u6570\uff0c\u4f8b\u5982 Name=&#8217;wy&#8217; \u53ef\u4ee5\u8fc7\u6ee4\u51fa5\u884c\uff0c Gender=&#8217;female&#8217; \u53ea\u80fd\u8fc7\u6ee4\u51fa\u4e00\u534a\u7684\u6570\u636e\uff0c \u4f18\u5316\u5668\u5c31\u4f1a\u77e5\u9053\uff0c\u6b64\u65f6\u9009\u62e9 KEY_NAME \u662f\u4ee3\u4ef7\u66f4\u5c0f\u7684\u9009\u62e9\u3002\u4e3a\u4e86\u652f\u6491\u8fd9\u91cc\u7684\u4f30\u7b97\u8fc7\u7a0b\uff0c\u7edf\u8ba1\u4fe1\u606f\u9700\u8981\u51b3\u5b9a\u4f7f\u7528\u4ec0\u4e48\u6837\u7684\u7b97\u6cd5\uff0c\u8fd9\u4e5f\u51b3\u5b9a\u4e86\u5982\u4f55\u6536\u96c6\uff0c\u5982\u4f55\u4f30\u7b97\u3002\u7b97\u6cd5\u6570\u636e\u5e93\u4e2d\u7684\u57fa\u6570\u4f30\u7b97\u7c7b\u4f3c\u4e8e Data Sketching\uff0c\u7b49\u503c\u573a\u666f\u4e0b\u7684\u57fa\u6570\u4f30\u7b97\u4e00\u822c\u6709\u4e24\u79cd\u9009\u62e9\uff0c\u57fa\u4e8e sample \u6216\u57fa\u4e8e\u5168\u91cf\u6570\u636e\u3002SAMPLE\u5f53\u6570\u636e\u91cf\u589e\u957f\u65f6\uff0c sample \u662f\u4e00\u79cd\u975e\u5e38\u6709\u6548\u63d0\u5347\u4f30\u7b97\u6548\u7387\u7684\u624b\u6bb5\u3002\u4f46 sample \u5fc5\u987b\u81f3\u5c11\u9762\u5bf9\u4ee5\u4e0b\u95ee\u9898\u25cf \u91c7\u6837\u7684\u968f\u673a\u5316\u25cf \u91c7\u6837\u6570\u636e\u4e0e\u6574\u4f53\u6570\u636e\u95f4\u7684\u5dee\u5f02\u4f30\u7b97\u25cb \u591a\u5c11 sample \u91cf\u624d\u80fd\u786e\u4fdd\u8bef\u5dee\u7387\u25cf \u5f53\u65b0\u6570\u636e\u51fa\u73b0\u65f6\uff0c\u5982\u4f55\u91c7\u6837\u91c7\u6837\u968f\u673a\u5316\u662f\u4e2a\u5f88\u73b0\u5b9e\u7684\u95ee\u9898\uff0c\u4f8b\u5982\u5b58\u50a8\u4e2d\u7684\u91c7\u6837\u6709\u4e24\u79cd\u505a\u6cd5\uff0c \u4e00\u79cd\u6309\u9875\u91c7\u6837\uff0c\u4e00\u79cd\u662f\u6bcf\u9875\u90fd\u91c7\u4e00\u90e8\u5206\u6570\u636e\u3002\u7b2c\u4e00\u79cd\u65b9\u6848\u867d\u7136\u975e\u5e38\u9ad8\u6548\uff0c\u4f46\u968f\u673a\u6027\u592a\u5dee\u5f88\u5bb9\u6613\u5bfc\u81f4\u4f30\u7b97\u7684\u6548\u679c\u975e\u5e38\u5dee\uff0c\u7b2c\u4e8c\u79cd\u65b9\u6848\u867d\u7136\u4f30\u7b97\u6548\u679c\u9ad8\u4e00\u4e9b\uff0c\u4f46\u5728 IO \u4e0a\u76f8\u5f53\u4e8e\u5168\u91cf\u626b\u63cf\u4e86\u4e00\u904d\u3002<br \/><img decoding=\"async\" referrerpolicy=\"no-referrer\" src=\"https:\/\/segmentfault.com\/img\/remote\/1460000042737951\" alt=\"\u56fe\u7247\" title=\"\u56fe\u7247\" loading=\"lazy\"><br \/>\u5728 Every Row Counts: Combining Sketches and Sampling for Accurate Group-By Result Estimates \u8fd9\u7bc7\u8bba\u6587\u4e2d\uff0c\u4f5c\u8005\u6bd4\u8f83\u4e86\u57fa\u4e8e Sample \u7684 GEE\/AE \u4e0e\u9762\u5411\u5168\u91cf\u6570\u636e\u7684 HyperLogLog sketch \u7b97\u6cd5\uff0cHyperLogLog \u7684\u7cbe\u786e\u5ea6\u5b8c\u80dc\u3002\u5f53\u7136\u8fd9\u4ec5\u4ec5\u662f\u5728\u8ba1\u7b97 NDV \u7684\u573a\u666f\u3002\u5f53\u6211\u4eec\u9700\u8981\u67d0\u4e2a\u5355\u4e00\u786e\u5b9a\u6570\u636e\u5728\u96c6\u7fa4\u4e2d\u7684\u7279\u5f81\u65f6\uff0c sample \u90fd\u5f88\u96be\u56de\u7b54\u8be5\u7c7b\u95ee\u9898\u3002\u4e3e\u4e2a\u4f8b\u5b50\uff0c\u6570\u636e\u7684\u5b58\u5728\u95ee\u9898\uff0c\u5373\u67d0\u6570\u636e\u662f\u5426\u5728\u96c6\u7fa4\u4e2d\u5b58\u5728\uff0c\u5982\u679c\u8be5\u6570\u636e\u4e0d\u5728 sample \u96c6\u4e2d\u5b58\u5728\uff0c\u90a3\u65e2\u6709\u53ef\u80fd\u6574\u4f53\u4e2d\u4e5f\u4e0d\u5b58\u5728\uff0c\u4e5f\u6709\u53ef\u80fd\u53ea\u662f\u6ca1\u6709\u88ab sample \u5230\u3002\u518d\u4f8b\u5982\u901a\u8fc7\u91c7\u6837\u4f30\u7b97 ndv \u65f6\uff0c\u5982\u679c\u91c7\u6837\u6570\u636e(\u5047\u8bbe 100000 \u884c) \u4e2d ndv \u63a5\u8fd1 100000\uff0c\u90a3\u4e48\u6574\u4f53\u6570\u636e(\u5047\u8bbe 10000000\u884c)\u7684 ndv \u6709\u53ef\u80fd\u662f 100000\uff0c\u6216\u8005\u4e5f\u6709\u53ef\u80fd\u662f 10000000\u3002NDV: number of the different valueSample \u7684\u6570\u636e\u96c6\u66f4\u9002\u5408\u7528\u4e8e\u89e3\u7b54\u8303\u56f4\u6761\u4ef6\u4e0b\u6570\u636e\u7684\u7279\u5f81\u95ee\u9898\u3002\u4f8b\u5982\u901a\u8fc7 sample \u7684\u6570\u636e\u6765\u6784\u5efa\u76f4\u65b9\u56fe\uff0c\u7528\u4e8e\u5904\u7406\u8303\u56f4\u6761\u4ef6\u7684\u67e5\u8be2\u95ee\u9898\uff0c\u6216\u8005\u901a\u8fc7 sample \u96c6\u4f30\u7b97\u591a\u8868\u7b49\u503c Join \u7684\u57fa\u6570\u95ee\u9898\u7b49\u3002\u8fd9\u4e9b\u95ee\u9898\u4e0d\u5728\u672c\u6587\u7684\u8ba8\u8bba\u8303\u56f4\u5185\uff0c\u6682\u4e0d\u5c55\u5f00\u3002\u975e Sample\u25cf frequency\uff1a\u7b49\u503c\u573a\u666f\u7684\u57fa\u6570\u4f30\u7b97\u5c31\u662f\u83b7\u53d6\u67d0\u503c\u5728\u96c6\u5408\u4e2d\u7684\u6570\u91cf\uff0c\u4e5f\u5c31\u662f\u4e00\u4e2a\u83b7\u53d6 frequency \u7684\u95ee\u9898CountMin SketchCountMin Sketch \u901a\u5e38\u7528\u4e8e\u89e3\u51b3 frequency \u95ee\u9898\u3002\u5b83\u6709\u70b9\u50cf Bloom filter\uff0c\u540c\u6837\u9700\u8981\u4e00\u4e2a\u6570\u7ec4\u4ee5\u53ca\u4e00\u4e2a Hash \u51fd\u6570\u7684\u96c6\u5408\uff0c\u5f53\u904d\u5386\u6570\u636e\u65f6\uff0c\u6bcf\u4e2a\u6570\u636e\u4f1a\u88ab\u5404\u4e2a Hash \u51fd\u6570 Map \u5230\u6570\u7ec4\u4e2d\u7684\u67d0\u4e2a\u4f4d\u7f6e\uff0c\u5e76\u4f7f\u8be5\u4f4d\u7f6e\u7684\u8ba1\u6570\u5668\u52a01\u3002\u5728\u4f30\u7b97\u73af\u8282\uff0c\u53d6\u8be5\u6570\u636e\u5bf9\u5e94\u6240\u6709\u8ba1\u6570\u5668\u6240\u5f97\u7684\u6700\u5c0f\u503c\u3002\u53ef\u4ee5\u9884\u89c1\u5b83\u4e00\u5b9a\u4f1a\u786e\u4fdd\u4f30\u7b97\u503c\u5927\u4e8e\u7b49\u4e8e\u771f\u5b9e\u503c\uff0c\u7136\u540e\u518d\u901a\u8fc7\u4e00\u4e9b\u624b\u6bb5\u786e\u4fdd\u8bef\u5dee\u7387\u3002\u5177\u4f53\u7684\u539f\u7406\u8fd9\u91cc\u4e0d\u518d\u8d58\u8ff0\u3002CountMin Sketch \u5177\u4f53\u539f\u7406\u89c1 An Improved Data Stream Summary:The Count-Min Sketch and its Applications<br \/><img decoding=\"async\" referrerpolicy=\"no-referrer\" src=\"https:\/\/segmentfault.com\/img\/remote\/1460000042737952\" alt=\"\u56fe\u7247\" title=\"\u56fe\u7247\" loading=\"lazy\"><br \/>CountMin Sketch \u6700\u5927\u7684\u95ee\u9898\u662f\u5f53 Sketch \u5927\u5c0f\u4e0d\u5408\u9002\u65f6\uff0c\u5f88\u5bb9\u6613\u56e0 Hash collision \u51fa\u73b0\u8fc7\u5ea6\u81a8\u80c0\u7684\u4f30\u7b97\u3002\u4e0d\u7ba1\u662f\u589e\u5927 rows \uff08\u5373 Hash \u51fd\u6570\uff09\u6570\u91cf\u6216 columns \uff08\u5373 Counter\uff09\u7684\u6570\u91cf\uff0c\u90fd\u53ef\u4ee5\u51cf\u5c11\u9519\u8bef\u7387\u3002CountMin Sketch \u5728\u5904\u7406\u5927\u91cf\u91cd\u590d\u6570\u636e\u65f6\u5f88\u6709\u4f18\u52bf\uff0c\u4f46\u5728\u5904\u7406\u975e\u503e\u659c\u6570\u636e\u65f6\uff0c\u5f88\u5bb9\u6613\u4ea7\u751f\u8fc7\u4e8e\u81a8\u80c0\u7684\u7ed3\u679c\u3002\u5f53\u7136\uff0c\u6700\u5f00\u59cb CountMin \u4e5f\u662f\u7528\u4e8e\u5904\u7406\u70ed\u70b9\u6570\u636e\u83b7\u53d6\u8fd9\u79cd\u573a\u666f\u3002\u4f46\u5728\u6570\u636e\u5206\u5e03\u6bd4\u8f83\u5747\u5300\u7684\u573a\u666f\u5e76\u4e0d\u5341\u5206\u53cb\u597d\u3002HyperLogLog+TopN\u5728 PolarDB-X \u4e2d\uff0c\u6211\u4eec\u6743\u8861\u4e4b\u540e\u4f7f\u7528 HyperLogLog+TopN \u7684\u65b9\u5f0f\u5904\u7406 frequency \u95ee\u9898\u3002\u5177\u4f53\u7684\u505a\u6cd5\u662f\u8fd9\u6837\u7684\u3002HyperLogLog \u7528\u4e8e\u4f30\u7b97\u51fa\u51c6\u786e\u7684 NDV \u503c\uff0cTopN \u7528\u4e8e\u5904\u7406\u6570\u636e\u503e\u659c\u3002\u5177\u4f53\u7684\u7b97\u6cd5\u5927\u81f4\u5982\u4e0b\uff1aif (topN.get(i) != 0) {return (long)(topN.get(i) \/ sampleRate);}return (rowcount \/ cardinality);HyperLogLog \u9700\u8981\u626b\u63cf\u5168\u91cf\u6570\u636e\uff0c TopN \u53ef\u4ee5\u901a\u8fc7 Sample \u6570\u636e\u6784\u5efa\u3002\u8fd9\u79cd\u6570\u636e\u7ed3\u6784\u5728\u5904\u7406\u6570\u636e\u7684\u5206\u5e03\u6574\u4f53\u8f83\u4e3a\u5e73\u5747\uff0c\u4ec5\u6709\u5c11\u6570\u503e\u659c\u4e25\u91cd\u7684\u6570\u636e\u573a\u666f\u4e2d\u975e\u5e38\u4f18\u79c0\u3002\u901a\u8fc7 rowcount \/ cardinality \u7a33\u5b9a\u5927\u591a\u6570\u975e\u503e\u659c\u6570\u636e\uff0c\u901a\u8fc7 TopN \u89e3\u51b3\u503e\u659c\u6570\u636e\uff0c\u4f46 TopN \u7684 Sample \u53ef\u80fd\u4f1a\u5bfc\u81f4\u503e\u659c\u6570\u636e\u51fa\u73b0\u8f83\u5927\u504f\u5dee\u3002\u4ece\u4f18\u5316\u5668\u7684\u9700\u6c42\u8003\u8651\uff0c\u80fd\u591f\u901a\u8fc7\u4f30\u7b97\u503c\u660e\u663e\u7684\u533a\u522b\u51fa\u503e\u659c\u4e0e\u975e\u503e\u659c\u6570\u636e\uff0c\u8fd9\u5df2\u7ecf\u53ef\u4ee5\u6ee1\u8db3 Plan \u7684\u9009\u62e9\u9700\u6c42\u3002\u6d4b\u8bd5\u6211\u4eec\u57fa\u4e8e\u6a21\u62df\u7684\u968f\u673a\u6570\u636e\uff0c\u5bf9\u6bd4\u4e86 CountMin Sketch \u4e0e HyperLogLog+TopN \uff0c\u5728\u51e0\u4e2a\u7279\u5b9a\u573a\u666f\u4e0b\u7684\u8868\u73b0\u3002\u4e3b\u8981\u89c2\u5bdf\u7684\u6570\u636e\u6709\u4ee5\u4e0b\u4e24\u9879\uff1a\u25cf \u6700\u5927\u8bef\u5dee\uff1a\u4f30\u7b97\u503c\u4e0e\u771f\u5b9e\u503c\u6700\u5927\u5dee\u5f02\u503c\u25cf \u8bef\u5dee\u5e73\u65b9\u5dee\uff1a\u4ee3\u8868\u8bef\u5dee\u8303\u56f4\u7684\u7a33\u5b9a\u7a0b\u5ea6\u6570\u636e\u573a\u666f\u6709\u4ee5\u4e0b\u4e09\u79cd\uff1abound = 100000<br \/>\/\/ \u5747\u8861:<br \/>for (int i = 0; i &lt; size; i++) {<\/p>\n<pre><code>rs[i] = r.nextInt(bound);<\/code><\/pre>\n<p>}<\/p>\n<p>\/\/ \u4e00\u534a\u6570\u636e\u503e\u659c\u5230 1000 \u4e2a slot \u4e2d<br \/>if (r.nextBoolean()) {<\/p>\n<pre><code>rs[i] = r.nextInt(1000);<\/code><\/pre>\n<p>} else {<\/p>\n<pre><code>rs[i] = r.nextInt(bound);<\/code><\/pre>\n<p>}<\/p>\n<p>\/\/ \u4e00\u534a\u6570\u636e\u503e\u659c\u81f3 100 \u4e2a slot \u4e2d<br \/>if (r.nextBoolean()) {<\/p>\n<pre><code>rs[i] = r.nextInt(100);<\/code><\/pre>\n<p>} else {<\/p>\n<pre><code>rs[i] = r.nextInt(bound);<\/code><\/pre>\n<p>}\u7b97\u6cd5\u521d\u59cb\u5316\u53c2\u6570\u5982\u4e0b\uff1a\u25cf CountMin Sketch \u521d\u59cb\u5316\u53c2\u6570 epsOfTotalCount = 0.0001,confidence=0.99\uff0c\u5f62\u6210\u4e86 7*20000 \u7684 long \u6570\u7ec4\uff0csketch\u5927\u7ea6\u53601M\u7a7a\u95f4\u25cf TopN \u56fa\u5b9a sample 10W \u4e2a\u6570\u636e\uff0c\u4fdd\u7559\u7ea6\u4e0d\u8d85\u8fc75000\u4e2a\u6700\u5927\u503c\u3002 HyperLogLog \u7cbe\u786e\u5ea6\u57281.04\/sqrt(2^16) = 0.004\uff0c \u7a7a\u95f4\u5927\u7ea6\u572850K\u4ee5\u4e0b\u6d4b\u8bd5\u7684\u5927\u81f4\u7ed3\u679c\u5982\u4e0b:<br \/><img decoding=\"async\" referrerpolicy=\"no-referrer\" src=\"https:\/\/segmentfault.com\/img\/remote\/1460000042737953\" alt=\"\u56fe\u7247\" title=\"\u56fe\u7247\" loading=\"lazy\"><\/p>\n<p><img decoding=\"async\" referrerpolicy=\"no-referrer\" src=\"https:\/\/segmentfault.com\/img\/remote\/1460000042737954\" alt=\"\u56fe\u7247\" title=\"\u56fe\u7247\" loading=\"lazy\"><br \/>\u5728\u6700\u5927\u8bef\u5dee\u4e2d\uff0cCountMin \u4e3b\u8981\u6765\u81ea\u4e8e\u5747\u8861\u6570\u636e\u7684\u81a8\u80c0\u3002Hll + TopN \u5728 10W \u7684\u573a\u666f\u6700\u5927\u8bef\u5dee\u6765\u81ea\u4e8e\u5355\u4e00\u503c\u88ab\u4f30\u7b97\u6210\u4e86 rowcount\/cardinality\uff0c \u4f46\u8fd9\u4e2a\u8bef\u5dee\u975e\u5e38\u5c0f\u3002\u5728100W \u548c 1000W \u7684\u573a\u666f\uff0cHll + TopN \u7684\u6700\u5927\u8bef\u5dee\u8fdc\u5927\u4e8e CountMin\uff0c\u4f46\u4ece\u6574\u4f53\u4e0a\u5224\u65ad\uff0cHLL+TopN\u7684\u8bef\u5dee\u5e73\u65b9\u5dee\u76f8\u6bd4\u800c\u65b9\u66f4\u52a0\u7a33\u5b9a\uff0c\u8fd9\u4e9b\u8bef\u5dee\u4e3b\u8981\u6765\u81ea\u4e8e TopN \u4e2d\u672a\u8bb0\u5f55\u7684 Skew \u6570\u636e\uff08\u672a Sample \u5230\u6216 TopN \u5bb9\u91cf\u4e0d\u8db3\uff09\u3002\u6574\u4f53\u4e0a\u6bd4\u8f83\u6765\u770b\uff0cCountMin Sketch \u968f\u6570\u636e\u91cf\u589e\u957f\uff0c\u9488\u5bf9\u503e\u659c\u6570\u636e\u7684\u4f30\u7b97\u4f1a\u66f4\u51c6\u786e\uff0c\u4f46\u975e\u503e\u659c\u6570\u636e\u7684\u4f30\u7b97\u4f1a\u51fa\u73b0\u8f83\u4e25\u91cd\u504f\u5dee\uff0c\u8fd9\u4f53\u73b0\u5728\u8bef\u5dee\u5e73\u65b9\u5dee\u5728\u5927\u6570\u636e\u573a\u666f\u4e00\u76f4\u4fdd\u6301\u8f83\u9ad8\u3002Hll+TopN \u975e\u503e\u659c\u6570\u636e\u7684\u4f30\u7b97\u4e2d\u975e\u5e38\u7a33\u5b9a\u4e14\u51c6\u786e\uff0c\u503e\u659c\u6570\u636e\u7684\u4f30\u7b97\u4f1a\u968f\u6570\u636e\u91cf\u589e\u957f\u6216 sample \u7684\u975e\u968f\u673a\u6027\u589e\u52a0\u800c\u6709\u8f83\u4e25\u91cd\u81a8\u80c0\u3002\u4ece\u7b49\u503c\u6761\u4ef6\u57fa\u6570\u4f30\u7b97\u7684\u573a\u666f\u8003\u8651\uff0c\u5982\u679c\u8981\u786e\u4fdd\u503e\u659c\u6570\u636e\u7684\u7cbe\u786e\u5ea6\uff0c\u5e94\u8be5\u8981\u9009\u62e9 CountMin Sketch\uff0c\u5982\u679c\u8981\u4fdd\u8bc1\u975e\u503e\u659c\u6570\u636e\u7684\u7cbe\u786e\u5ea6\uff0c\u503e\u659c\u6570\u636e\u53ea\u8981\u80fd\u8bc6\u522b\u51fa\u5c31\u597d\uff0c\u91c7\u7528 Hll+TopN \u662f\u66f4\u52a0\u5b9e\u7528\u7684\u9009\u62e9\u3002\u6ce8\u610f\u4ee5\u4e0a\u6d4b\u8bd5\u7ed3\u679c\u4ec5\u9488\u5bf9\u6a21\u62df\u7684\u6570\u636e\u573a\u666f\u3002\u91c7\u96c6\u867d\u7136 CountMin \u548c HyperLogLog \u56de\u7b54\u7684\u95ee\u9898\u4e0d\u540c(\u4e00\u4e2a\u662f frequency, \u4e00\u4e2a\u662f ndv )\uff0c\u4f46\u672c\u8d28\u4e0a\u90fd\u662f\u57fa\u4e8e Sketch \u6765\u63cf\u8ff0\u6574\u4f53\u6570\u636e\u3002\u57fa\u4e8e Sketch \u7684\u7b97\u6cd5\u9664\u4e86\u5185\u5b58\u5360\u7528\u5c0f\u4ee5\u5916\uff0c\u8fd8\u6709\u4e2a\u975e\u5e38\u5927\u7684\u597d\u5904\u662f\u591a\u4e2a Sketch \u53ef\u4ee5\u5408\u5e76\uff0c\u5e76\u7528\u4e8e\u63cf\u8ff0\u591a\u5206\u7247\u6570\u636e\u3002\u5927\u90e8\u5206\u7edf\u8ba1\u4fe1\u606f\u6a21\u5757\u9700\u8981\u6536\u96c6\u7684\u6570\u636e\u662f\u4e0d\u9700\u8981\u8003\u8651\u591a\u5206\u7247\u5e26\u6765\u7684\u5f71\u54cd\u7684\uff0c\u6bd4\u5982\u8bf4\u6700\u5927\u6700\u5c0f\u503c\uff0crowcount\uff0cnull \u503c\u6570\u91cf\uff0c\u5217\u957f\u5ea6\u7b49\u7b49\u3002\u4f46 ndv \u4e0d\u540c\uff0c\u591a\u4e2a\u5206\u7247\u7684 ndv \u503c\u65e0\u6cd5\u7b80\u5355\u76f8\u52a0\u5f97\u51fa\u3002\u4f46 HyperLogLog \u7684 Sketch \u6570\u636e\u53ef\u4ee5\u901a\u8fc7\u591a\u5206\u7247\u5408\u5e76\u7684\u65b9\u5f0f\u4f30\u7b97\u51fa\u6574\u4f53\u7684 ndv \u503c\u3002CountMin \u7684 Sketch \u4e5f\u53ef\u4ee5\u901a\u8fc7\u5408\u5e76\u5904\u7406\u3002CountMin \u548c HyperLogLog \u5982\u679c\u60f3\u8981\u8f83\u9ad8\u7684\u7cbe\u786e\u5ea6\uff0c\u5fc5\u987b\u8981\u626b\u63cf\u5168\u91cf\u7684\u6570\u636e\uff0c\u4f46\u8fd9\u4f1a\u81f3\u5c11\u7ed9 IO \u5e26\u6765\u5de8\u5927\u7684\u5f71\u54cd\u3002 Sketch \u7c7b\u578b\u7684\u7b97\u6cd5\u53ef\u4ee5\u9488\u5bf9\u5404\u4e2a\u5206\u7247\u6570\u636e\u5efa\u7acb\u4e0d\u540c\u7684 Sketch\uff0c\u7136\u540e\u4ec5\u626b\u63cf\u6709\u6570\u636e\u53d8\u66f4\u5206\u7247\u5e76\u66f4\u65b0\u5176 Sketch\uff0c\u8fd9\u6837\u907f\u514d\u4e86\u626b\u63cf\u5168\u91cf\u6570\u636e\u7684\u540c\u65f6\uff0c\u4ecd\u7136\u53ef\u4ee5\u5728\u5168\u5c40\u5f97\u5230\u7cbe\u786e\u7684\u4f30\u7b97\u503c\u3002<br \/><img decoding=\"async\" referrerpolicy=\"no-referrer\" src=\"https:\/\/segmentfault.com\/img\/remote\/1460000042737955\" alt=\"\u56fe\u7247\" title=\"\u56fe\u7247\" loading=\"lazy\"><br \/>\u5728 Oracle 11g \u4e2d\uff0c\u5bf9\u4e0d\u6d3b\u8dc3\u7684\u6570\u636e\u5206\u7247\u5206\u6210\u4e00\u7ec4\u7edf\u4e00\u8fdb\u884c sketch\uff0c\u8fd9\u53ef\u4ee5\u6781\u5927\u51cf\u5c11 sketch \u5360\u7528\u7684\u5185\u5b58\u7a7a\u95f4\u7ed3\u8bedThe approximate approach is often faster and more efficient&#8211;BY GRAHAM CORMODE<br \/>\u5230\u76ee\u524d\u4e3a\u6b62\uff0c\u6570\u636e\u7279\u5f81\u9886\u57df\u5df2\u7136\u5c55\u5f00\u4e86\u5f88\u957f\u65f6\u95f4\u7684\u7814\u7a76\uff0c\u975e\u5e38\u591a\u7684\u5de5\u5177\u88ab\u5f00\u53d1\u51fa\u6765\u3002\u672c\u6587\u4e3b\u8981\u7b80\u5355\u4ecb\u7ecd\u4e86\u5176\u4e2d\u4e00\u90e8\u5206\u57fa\u4e8e Sketch \u7684\u6280\u672f\u5728\u6570\u636e\u5e93\u7b49\u503c\u67e5\u8be2\u573a\u666f\u4e2d\u7684\u533a\u522b\u4e0e\u5f71\u54cd\u3002\u53ef\u4ee5\u770b\u5230\uff0c\u8fd9\u4e9b\u6280\u672f\u63d0\u4f9b\u4e86\u4e0d\u540c\u573a\u666f\u4e0b\u89e3\u51b3\u7279\u5b9a\u95ee\u9898\u7684\u80fd\u529b\uff0c\u4f46\u5e76\u975e\u94f6\u5f39\uff0c\u8c28\u614e\u6743\u8861\u7684\u9009\u62e9\u5de5\u5177\uff0c\u624d\u80fd\u5e26\u6765\u66f4\u4f73\u7684\u4f18\u5316\u6548\u679c\u3002\u53c2\u8003\u25cf Efficient and Scalable Statistics Gathering for Large Databases in Oracle 11g\u2217\u25cf An Improved Data Stream Summary:The Count-Min Sketch and its Applications\u25cf Every Row Counts: Combining Sketches and Sampling for Accurate Group-By Result Estimates\u25cf HyperLogLog in Practice: Algorithmic Engineering of a State of The Art Cardinality Estimation Algorithm\u25cf Towards Estimation Error Guarantees for Distinct Values\u25cf Data Sketching\u25cf New Sampling-Based Summary Statistics for Improving Approximate Query Answers\u25cf A Comparison of Five Probabilistic View-Size Estimation Techniques in OLAP\u25cf\u00a0<a href=\"https:\/\/link.segmentfault.com\/?enc=YsFRiWHebEPNF4RrJNXw9Q%3D%3D.Hua%2Fb0SfmDUIv1eBeedTCn%2BsavU%2BWoivy%2F3qbYqYeY%2B%2BHUN8nHpWdplO3ONht%2F%2B4wQM9Nwjpbovjl7%2FaKEnTTg%3D%3D\" rel=\"nofollow\">http:\/\/content.research.neust&#8230;<\/a>\u6b22\u8fce\u5173\u6ce8PolarDB-X\u77e5\u4e4e\u673a\u6784\u53f7\uff0c\u9605\u8bfb\u66f4\u591a\u6280\u672f\u597d\u6587\u3002\u539f\u6587\u94fe\u63a5\uff1a<a href=\"https:\/\/link.segmentfault.com\/?enc=NGaZIqRoX5%2FJFxje%2F8lwCQ%3D%3D.Qq%2FdF1F4061dtRtX211Y1UWHdlLr2k8%2BsFWPmynwV%2BxkPIAT%2F8ikwQvW18goAoEN\" rel=\"nofollow\">https:\/\/click.aliyun.com\/m\/10&#8230;<\/a>\u672c\u6587\u4e3a\u963f\u91cc\u4e91\u539f\u521b\u5185\u5bb9\uff0c\u672a\u7ecf\u5141\u8bb8\u4e0d\u5f97\u8f6c\u8f7d\u3002<\/p>\n<p><a href=\"https:\/\/segmentfault.com\/a\/1190000042737949\">https:\/\/segmentfault.com\/a\/1190000042737949<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u7b80\u4ecb\uff1a\u00a0\u7edf\u8ba1\u4fe1\u606f\u662f\u4e3a\u4f18\u5316\u5668\u7684 cost \u4f30\u7b97\u63d0\u4f9b\u6570\u636e\u652f\u6491\uff0c\u5176\u4e2d\u5f88\u91cd\u8981\u7684\u4e00\u70b9\u9700\u6c42\u4fbf\u662f\u7b49\u503c\u67e5\u8be2(EQUALS, I [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[4],"tags":[],"class_list":["post-10006791","post","type-post","status-publish","format-standard","hentry","category-share"],"jetpack_sharing_enabled":true,"jetpack_featured_media_url":"","_links":{"self":[{"href":"https:\/\/itmax.cn\/index.php?rest_route=\/wp\/v2\/posts\/10006791","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/itmax.cn\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/itmax.cn\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/itmax.cn\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/itmax.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=10006791"}],"version-history":[{"count":0,"href":"https:\/\/itmax.cn\/index.php?rest_route=\/wp\/v2\/posts\/10006791\/revisions"}],"wp:attachment":[{"href":"https:\/\/itmax.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=10006791"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/itmax.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=10006791"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/itmax.cn\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=10006791"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}