{"id":34420,"date":"2025-03-10T17:18:45","date_gmt":"2025-03-10T09:18:45","guid":{"rendered":"https:\/\/www.yunshan.net\/blog\/?p=34420"},"modified":"2025-03-10T17:18:45","modified_gmt":"2025-03-10T09:18:45","slug":"%e5%85%a8%e6%a0%88%e5%8f%af%e8%a7%82%e6%b5%8b%e6%80%a7%e5%a6%82%e4%bd%95%e6%94%af%e6%8c%81%e9%ab%98%e6%80%a7%e8%83%bd%e8%ae%a1%e7%ae%97%e7%8e%af%e5%a2%83%ef%bc%9f","status":"publish","type":"post","link":"https:\/\/www.yunshan.net\/blog\/archives\/34420.html","title":{"rendered":"\u5168\u6808\u53ef\u89c2\u6d4b\u6027\u5982\u4f55\u652f\u6301\u9ad8\u6027\u80fd\u8ba1\u7b97\u73af\u5883\uff1f"},"content":{"rendered":"<p style=\"text-align: center;\">\n<p><img decoding=\"async\" src=\"https:\/\/www.yunshan.net\/blog\/wp-content\/uploads\/2025\/03\/cSjjc6.jpg\" alt=\"\u5168\u6808\u53ef\u89c2\u6d4b\u6027\u5982\u4f55\u652f\u6301\u9ad8\u6027\u80fd\u8ba1\u7b97\u73af\u5883\uff1f\" ><\/p>\n<\/p>\n<p><strong>\u524d\u8a00<\/strong><br \/>\n\u5728\u5168\u7403\u6570\u5b57\u5316\u8f6c\u578b\u7684\u6d6a\u6f6e\u4e2d\uff0c\u9ad8\u6027\u80fd\u8ba1\u7b97\uff08HPC\uff09\u5df2\u6210\u4e3a\u63a8\u52a8\u79d1\u5b66\u53d1\u73b0\u548c\u5546\u4e1a\u521b\u65b0\u7684\u6838\u5fc3\u5f15\u64ce\u3002\u7136\u800c\uff0c\u968f\u7740\u8ba1\u7b97\u89c4\u6a21\u4ece\u5343\u6838\u7ea7\u5411\u767e\u4e07\u6838\u7ea7\u8dc3\u8fc1\uff0c\u4f20\u7edf\u76d1\u63a7\u624b\u6bb5\u5df2\u96be\u4ee5\u5e94\u5bf9\u65e5\u76ca\u590d\u6742\u7684\u7cfb\u7edf\u8fd0\u7ef4\u6311\u6218\u3002\u5f53\u4e07\u4ebf\u7ea7\u6d6e\u70b9\u8fd0\u7b97\u4e0ePB\u7ea7\u6570\u636e\u541e\u5410\u6210\u4e3a\u5e38\u6001\uff0c\u5982\u4f55\u5b9e\u73b0\u4ece\u82af\u7247\u5230\u96c6\u7fa4\u7684\u5168\u94fe\u8def\u6027\u80fd\u900f\u89c6\uff1f\u8fd9\u6b63\u662f<strong><a href=\"https:\/\/www.yunshan.net\/\"   target=\"_blank\" style=\"color:blue\">\u5168\u6808\u53ef\u89c2\u6d4b<\/a>\u6027<\/strong>\u6280\u672f\u5c55\u73b0\u5176\u72ec\u7279\u4ef7\u503c\u7684\u5173\u952e\u6218\u573a\u3002\u672c\u6587\u5c06\u6df1\u5165\u63a2\u8ba8\u8fd9\u4e00\u6280\u672f\u5982\u4f55\u4e3aHPC\u73af\u5883\u63d0\u4f9b\u7cbe\u51c6\u7684\u201c\u8bca\u65ad\u663e\u5fae\u955c\u201d\u4e0e\u201c\u4f18\u5316\u5bfc\u822a\u4eea\u201d\u3002<\/p>\n<hr>\n<h3 id=\"\u4e00-hpc\u73af\u5883\u7684\u72ec\u7279\u6311\u6218\">\u4e00\u3001HPC\u73af\u5883\u7684\u72ec\u7279\u6311\u6218<\/h3>\n<p>\u9ad8\u6027\u80fd\u8ba1\u7b97\u73af\u5883\u901a\u5e38\u7531\u5f02\u6784\u8ba1\u7b97\u8282\u70b9\uff08CPU\u3001GPU\u3001FPGA\uff09\u3001\u9ad8\u901f\u7f51\u7edc\uff08\u5982InfiniBand\uff09\u3001\u5206\u5e03\u5f0f\u5b58\u50a8\u7cfb\u7edf\u53ca\u590d\u6742\u8c03\u5ea6\u6846\u67b6\u6784\u6210\u3002\u8fd9\u79cd\u67b6\u6784\u5728\u5e26\u6765\u5f3a\u5927\u7b97\u529b\u7684\u540c\u65f6\uff0c\u4e5f\u57cb\u85cf\u7740\u591a\u91cd\u8fd0\u7ef4\u9690\u60a3\uff1a<\/p>\n<ol>\n<li><strong>\u7cfb\u7edf\u590d\u6742\u6027\u6307\u6570\u589e\u957f<\/strong>\uff1a\u8de8\u5c42\u7ec4\u4ef6\u7684\u8026\u5408\u6027\u5bfc\u81f4\u5355\u4e00\u8282\u70b9\u6545\u969c\u53ef\u80fd\u5f15\u53d1\u7ea7\u8054\u6548\u5e94\uff1b\n<\/li>\n<li><strong>\u5b9e\u65f6\u6027\u8981\u6c42\u4e25\u82db<\/strong>\uff1a\u90e8\u5206\u79d1\u5b66\u8ba1\u7b97\u4efb\u52a1\u9700\u5728\u6570\u5c0f\u65f6\u5185\u5b8c\u6210\uff0c\u4efb\u4f55\u6027\u80fd\u6ce2\u52a8\u90fd\u53ef\u80fd\u5f71\u54cd\u7814\u7a76\u8fdb\u5ea6\uff1b\n<\/li>\n<li><strong>\u6545\u969c\u5b9a\u4f4d\u56f0\u96be<\/strong>\uff1a\u4f20\u7edf\u65e5\u5fd7\u5206\u6790\u96be\u4ee5\u5728\u4ebf\u7ea7\u5e76\u53d1\u7ebf\u7a0b\u4e2d\u5feb\u901f\u5b9a\u4f4d\u5f02\u5e38\u6839\u6e90\uff1b\n<\/li>\n<li><strong>\u8d44\u6e90\u5229\u7528\u7387\u74f6\u9888<\/strong>\uff1a\u636eIDC\u7814\u7a76\uff0c\u5168\u7403HPC\u96c6\u7fa4\u5e73\u5747\u5229\u7528\u7387\u4e0d\u8db360%\uff0c\u9690\u6027\u8d44\u6e90\u6d6a\u8d39\u4e25\u91cd\u3002\n<\/li>\n<\/ol>\n<p>\u8fd9\u4e9b\u75db\u70b9\u547c\u5524\u4e00\u79cd\u80fd\u591f\u7a7f\u900f\u786c\u4ef6\u3001\u8f6f\u4ef6\u3001\u7f51\u7edc\u5c42\uff0c\u5e76\u63d0\u4f9b\u5b9e\u65f6\u6d1e\u89c1\u7684\u89c2\u6d4b\u4f53\u7cfb\u3002<\/p>\n<hr>\n<h3 id=\"\u4e8c-\u5168\u6808\u53ef\u89c2\u6d4b\u6027\u7684\u6280\u672f\u5185\u6db5\">\u4e8c\u3001<a href=\"https:\/\/www.yunshan.net\/\"   target=\"_blank\" style=\"color:blue\">\u5168\u6808\u53ef\u89c2\u6d4b<\/a>\u6027\u7684\u6280\u672f\u5185\u6db5<\/h3>\n<p><strong>\u5168\u6808\u53ef\u89c2\u6d4b\u6027<\/strong>\u5e76\u975e\u7b80\u5355\u7684\u76d1\u63a7\u5347\u7ea7\uff0c\u800c\u662f\u901a\u8fc7<strong>\u6307\u6807\uff08Metrics\uff09\u3001\u65e5\u5fd7\uff08Logs\uff09\u3001\u8ffd\u8e2a\uff08Traces\uff09<\/strong>\u7684\u4e09\u7ef4\u6570\u636e\u878d\u5408\uff0c\u6784\u5efa\u7cfb\u7edf\u7684\u201c\u6570\u5b57\u5b6a\u751f\u201d\u3002\u5176\u6838\u5fc3\u4f18\u52bf\u4f53\u73b0\u5728\uff1a<\/p>\n<ul>\n<li><strong>\u5168\u5c40\u53ef\u89c6\u5316<\/strong>\uff1a\u4ece\u7269\u7406\u670d\u52a1\u5668\u529f\u8017\u5230\u5e94\u7528\u7ebf\u7a0b\u72b6\u6001\uff0c\u5b9e\u73b0\u8de8\u5c42\u6570\u636e\u5173\u8054\uff1b\n<\/li>\n<li><strong>\u667a\u80fd<a href=\"https:\/\/www.yunshan.net\/\"   target=\"_blank\" style=\"color:blue\">\u6839\u56e0\u5206\u6790<\/a><\/strong>\uff1a\u7ed3\u5408\u673a\u5668\u5b66\u4e60\uff0c\u81ea\u52a8\u8bc6\u522b\u5f02\u5e38\u6a21\u5f0f\u5e76\u8ffd\u6eaf\u6e90\u5934\uff1b\n<\/li>\n<li><strong>\u9884\u6d4b\u6027\u8fd0\u7ef4<\/strong>\uff1a\u901a\u8fc7\u5386\u53f2\u6570\u636e\u5efa\u6a21\uff0c\u9884\u5224\u786c\u4ef6\u6545\u969c\u6216\u6027\u80fd\u62d0\u70b9\u3002\n<\/li>\n<\/ul>\n<p>\u5728HPC\u573a\u666f\u4e2d\uff0c\u8fd9\u610f\u5473\u7740\u80fd\u591f\u540c\u65f6\u6355\u6349<strong>\u8ba1\u7b97\u8282\u70b9\u7684\u6e29\u5ea6\u66f2\u7ebf<\/strong>\u3001<strong>MPI\u901a\u4fe1\u5ef6\u8fdf<\/strong>\u3001<strong>\u4f5c\u4e1a\u8c03\u5ea6\u961f\u5217\u72b6\u6001<\/strong>\u7b49\u5173\u952e\u4fe1\u53f7\uff0c\u5e76\u5c06\u5176\u8f6c\u5316\u4e3a\u53ef\u64cd\u4f5c\u7684\u4f18\u5316\u5efa\u8bae\u3002<\/p>\n<hr>\n<h3 id=\"\u4e09-\u5168\u6808\u53ef\u89c2\u6d4b\u6027\u5728\u9ad8\u6027\u80fd\u8ba1\u7b97\u4e2d\u7684\u5b9e\u8df5\u8def\u5f84\">\u4e09\u3001\u5168\u6808\u53ef\u89c2\u6d4b\u6027\u5728\u9ad8\u6027\u80fd\u8ba1\u7b97\u4e2d\u7684\u5b9e\u8df5\u8def\u5f84<\/h3>\n<h4 id=\"1-\u57fa\u7840\u67b6\u6784\u5c42\u7684\u6df1\u5ea6\u76d1\u63a7\">1. <strong>\u57fa\u7840\u67b6\u6784\u5c42\u7684\u6df1\u5ea6\u76d1\u63a7<\/strong><\/h4>\n<p>\u901a\u8fc7\u90e8\u7f72<strong>Prometheus<\/strong>\u3001<strong>Grafana<\/strong>\u7b49\u5de5\u5177\uff0c\u5b9e\u65f6\u91c7\u96c6CPU\/GPU\u5229\u7528\u7387\u3001\u5185\u5b58\u5e26\u5bbd\u3001\u7f51\u7edc\u4e22\u5305\u7387\u7b49\u786c\u4ef6\u6307\u6807\u3002\u4f8b\u5982\uff0c\u7f8e\u56fd\u67d0\u56fd\u5bb6\u5b9e\u9a8c\u5ba4\u901a\u8fc7<strong>\u529f\u8017\u5efa\u6a21<\/strong>\uff0c\u53d1\u73b012%\u7684\u8ba1\u7b97\u8282\u70b9\u56e0\u6563\u70ed\u4e0d\u826f\u5bfc\u81f4\u9891\u7387\u964d\u9891\uff0c\u901a\u8fc7\u8c03\u6574\u673a\u67dc\u5e03\u5c40\u540e\uff0c\u4efb\u52a1\u5b8c\u6210\u65f6\u95f4\u7f29\u77ed18%\u3002<\/p>\n<h4 id=\"2-\u5e94\u7528\u6027\u80fd\u7684\u7ec6\u7c92\u5ea6\u5206\u6790\">2. <strong>\u5e94\u7528\u6027\u80fd\u7684\u7ec6\u7c92\u5ea6\u5206\u6790<\/strong><\/h4>\n<p>\u501f\u52a9<strong>eBPF<\/strong>\u6280\u672f\uff0c\u5728\u4e0d\u4fb5\u5165\u4ee3\u7801\u7684\u60c5\u51b5\u4e0b\u8ffd\u8e2a\u79d1\u5b66\u8ba1\u7b97\u5e94\u7528\u7684\u5185\u6838\u6001\u884c\u4e3a\u3002\u67d0\u6c14\u8c61\u6a21\u62df\u56e2\u961f\u4f7f\u7528<strong>\u52a8\u6001\u706b\u7130\u56fe<\/strong>\uff0c\u5b9a\u4f4d\u5230\u6c14\u5019\u6a21\u578b\u4e2d\u7684\u9501\u7ade\u4e89\u95ee\u9898\uff0c\u4f18\u5316\u540e\u5355\u6b21\u6a21\u62df\u8017\u65f6\u51cf\u5c1123%\u3002<\/p>\n<h4 id=\"3-\u6570\u636e\u6d41\u6c34\u7ebf\u7684\u7aef\u5230\u7aef\u4f18\u5316\">3. <strong>\u6570\u636e\u6d41\u6c34\u7ebf\u7684\u7aef\u5230\u7aef\u4f18\u5316<\/strong><\/h4>\n<p>\u5728\u57fa\u56e0\u7ec4\u5b66\u7b49\u6570\u636e\u5bc6\u96c6\u578b\u573a\u666f\u4e2d\uff0c<strong><a href=\"https:\/\/www.yunshan.net\/\"   target=\"_blank\" style=\"color:blue\">\u5206\u5e03\u5f0f\u8ffd\u8e2a<\/a>\u7cfb\u7edf<\/strong>\uff08\u5982Jaeger\uff09\u53ef\u7ed8\u5236\u8de8\u5b58\u50a8\u8282\u70b9\u3001\u8ba1\u7b97\u96c6\u7fa4\u7684\u6570\u636e\u6d41\u56fe\u8c31\u3002\u67d0\u57fa\u56e0\u6d4b\u5e8f\u516c\u53f8\u901a\u8fc7\u5206\u6790<strong>\u5bf9\u8c61\u5b58\u50a8\u7684IO\u8def\u5f84<\/strong>\uff0c\u5c06\u6570\u636e\u5904\u7406\u541e\u5410\u91cf\u63d0\u534740%\u3002<\/p>\n<h4 id=\"4-\u8d44\u6e90\u8c03\u5ea6\u5668\u7684\u667a\u80fd\u51b3\u7b56\u652f\u6301\">4. <strong>\u8d44\u6e90\u8c03\u5ea6\u5668\u7684\u667a\u80fd\u51b3\u7b56\u652f\u6301<\/strong><\/h4>\n<p>\u5c06\u89c2\u6d4b\u6570\u636e\u53cd\u9988\u81f3Slurm\u3001Kubernetes\u7b49\u8c03\u5ea6\u5668\uff0c\u5b9e\u73b0\u52a8\u6001\u8d44\u6e90\u5206\u914d\u3002\u5fb7\u56fd\u67d0\u8d85\u7b97\u4e2d\u5fc3\u901a\u8fc7<strong>\u5b9e\u65f6\u8d1f\u8f7d\u9884\u6d4b\u7b97\u6cd5<\/strong>\uff0c\u4f7f\u96c6\u7fa4\u5229\u7528\u7387\u4ece58%\u63d0\u5347\u81f376%\uff0c\u5e74\u8282\u7701\u7535\u8d39\u8d85200\u4e07\u6b27\u5143\u3002<\/p>\n<hr>\n<h3 id=\"\u56db-\u5178\u578b\u6848\u4f8b\u4e2d\u7684\u4ef7\u503c\u9a8c\u8bc1\">\u56db\u3001\u5178\u578b\u6848\u4f8b\u4e2d\u7684\u4ef7\u503c\u9a8c\u8bc1<\/h3>\n<h4 id=\"\u6848\u4f8b1-\u80fd\u6e90\u52d8\u63a2\u4e2d\u7684\u5e76\u884c\u8ba1\u7b97\u4f18\u5316\">\u6848\u4f8b1\uff1a\u80fd\u6e90\u52d8\u63a2\u4e2d\u7684\u5e76\u884c\u8ba1\u7b97\u4f18\u5316<\/h4>\n<p>\u67d0\u77f3\u6cb9\u516c\u53f8\u4f7f\u7528\u5168\u6808\u89c2\u6d4b\u5e73\u53f0\uff0c\u53d1\u73b0\u5176\u5730\u9707\u53cd\u6f14\u4f5c\u4e1a\u7684<strong>MPI\u901a\u4fe1\u5f00\u9500\u5360\u6bd4\u9ad8\u8fbe35%<\/strong>\u3002\u901a\u8fc7\u91cd\u6784\u4efb\u52a1\u5206\u533a\u7b56\u7565\u5e76\u4f18\u5316\u7f51\u7edc\u7f13\u51b2\u533a\u914d\u7f6e\uff0c\u6574\u4f53\u8ba1\u7b97\u6548\u7387\u63d0\u534728%\uff0c\u5355\u4e2a\u6cb9\u85cf\u6a21\u578b\u6c42\u89e3\u5468\u671f\u4ece14\u5929\u538b\u7f29\u81f310\u5929\u3002<\/p>\n<h4 id=\"\u6848\u4f8b2-ai\u8bad\u7ec3\u96c6\u7fa4\u7684\u6545\u969c\u81ea\u6108\">\u6848\u4f8b2\uff1aAI\u8bad\u7ec3\u96c6\u7fa4\u7684\u6545\u969c\u81ea\u6108<\/h4>\n<p>\u4e00\u5bb6\u81ea\u52a8\u9a7e\u9a76\u4f01\u4e1a\u5728\u5176GPU\u96c6\u7fa4\u4e2d\u90e8\u7f72\u4e86<strong>\u5f02\u5e38\u68c0\u6d4b\u7ba1\u9053<\/strong>\u3002\u5f53\u7cfb\u7edf\u81ea\u52a8\u8bc6\u522b\u5230\u67d0\u6279\u6b21\u663e\u5361\u7684<strong>ECC\u9519\u8bef\u7387\u5f02\u5e38\u6500\u5347<\/strong>\u65f6\uff0c\u7acb\u5373\u9694\u79bb\u6545\u969c\u5361\u5e76\u8fc1\u79fb\u4efb\u52a1\uff0c\u907f\u514d\u4e86\u4e00\u6b21\u53ef\u80fd\u6301\u7eed6\u5c0f\u65f6\u7684\u670d\u52a1\u4e2d\u65ad\u3002<\/p>\n<h4 id=\"\u6848\u4f8b3-\u8de8\u5730\u57df\u8d85\u7b97\u7684\u8d44\u6e90\u534f\u540c\">\u6848\u4f8b3\uff1a\u8de8\u5730\u57df\u8d85\u7b97\u7684\u8d44\u6e90\u534f\u540c<\/h4>\n<p>\u6b27\u6d32\u67d0\u8054\u5408\u8d85\u7b97\u9879\u76ee\u901a\u8fc7\u7edf\u4e00\u89c2\u6d4b\u5e73\u53f0\uff0c\u6574\u5408\u4e94\u5730\u6570\u636e\u4e2d\u5fc3\u7684\u95f2\u7f6e\u7b97\u529b\u3002\u5229\u7528<strong>\u52a8\u6001\u4f18\u5148\u7ea7\u8c03\u6574\u7b97\u6cd5<\/strong>\uff0c\u7d27\u6025\u6c14\u5019\u6a21\u62df\u4efb\u52a1\u7684\u6392\u961f\u7b49\u5f85\u65f6\u95f4\u4ece72\u5c0f\u65f6\u964d\u81f39\u5c0f\u65f6\uff0c\u52a9\u529b\u6781\u7aef\u5929\u6c14\u9884\u8b66\u54cd\u5e94\u3002<\/p>\n<hr>\n<h3 id=\"\u4e94-\u6280\u672f\u6f14\u8fdb\u4e0e\u672a\u6765\u8d8b\u52bf\">\u4e94\u3001\u6280\u672f\u6f14\u8fdb\u4e0e\u672a\u6765\u8d8b\u52bf<\/h3>\n<p>\u968f\u7740HPC\u4e0e\u4e91\u539f\u751f\u6280\u672f\u7684\u878d\u5408\uff0c\u5168\u6808\u53ef\u89c2\u6d4b\u6027\u6b63\u671d\u7740\u4e09\u4e2a\u65b9\u5411\u8fdb\u5316\uff1a<\/p>\n<ol>\n<li><strong>\u91cf\u5b50\u8ba1\u7b97\u517c\u5bb9\u6027<\/strong>\uff1a\u5f00\u53d1\u80fd\u89e3\u6790\u91cf\u5b50\u6bd4\u7279\u72b6\u6001\u7684\u4e13\u7528\u63a2\u9488\uff1b\n<\/li>\n<li><strong>\u8fb9\u7f18-\u4e91\u534f\u540c\u89c2\u6d4b<\/strong>\uff1a\u5728\u8054\u90a6\u5b66\u4e60\u7b49\u573a\u666f\u4e2d\u5b9e\u73b0\u4f4e\u5ef6\u8fdf\u6570\u636e\u540c\u6b65\uff1b\n<\/li>\n<li><strong>\u6570\u5b57\u5b6a\u751f\u4eff\u771f<\/strong>\uff1a\u901a\u8fc7\u5386\u53f2\u6570\u636e\u6784\u5efa\u865a\u62df\u96c6\u7fa4\uff0c\u9884\u6f14\u4f18\u5316\u7b56\u7565\u3002\n<\/li>\n<\/ol>\n<p>MIT\u8ba1\u7b97\u673a\u79d1\u5b66\u5b9e\u9a8c\u5ba4\u7684\u8fd1\u671f\u5b9e\u9a8c\u8868\u660e\uff0c<strong>AI\u9a71\u52a8\u7684\u89c2\u6d4b\u7cfb\u7edf<\/strong>\u5df2\u80fd\u572815\u5206\u949f\u5185\u81ea\u4e3b\u5b8c\u6210\u4f20\u7edf\u56e2\u961f\u97002\u5929\u5206\u6790\u7684\u8c03\u4f18\u4efb\u52a1\uff0c\u9884\u793a\u7740\u4e00\u573a\u667a\u80fd\u8fd0\u7ef4\u9769\u547d\u7684\u5230\u6765\u3002<\/p>\n<hr>\n<p><strong>\u6ce8<\/strong>\uff1a\u672c\u6587\u6570\u636e\u4e0e\u6848\u4f8b\u5747\u6765\u81ea\u516c\u5f00\u5b66\u672f\u8bba\u6587\u3001\u884c\u4e1a\u767d\u76ae\u4e66\u53ca\u6280\u672f\u4f1a\u8bae\u62a5\u544a\uff0c\u90e8\u5206\u7ec6\u8282\u7ecf\u8131\u654f\u5904\u7406\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u524d\u8a00 \u5728\u5168\u7403\u6570\u5b57\u5316\u8f6c\u578b\u7684\u6d6a\u6f6e\u4e2d\uff0c\u9ad8\u6027\u80fd\u8ba1\u7b97\uff08HPC\uff09\u5df2\u6210\u4e3a\u63a8\u52a8\u79d1\u5b66\u53d1\u73b0\u548c\u5546\u4e1a\u521b\u65b0\u7684\u6838\u5fc3\u5f15\u64ce\u3002\u7136\u800c\uff0c\u968f\u7740\u8ba1\u7b97\u89c4\u6a21\u4ece\u5343\u6838\u7ea7\u5411\u767e\u4e07\u6838\u7ea7\u8dc3\u8fc1\uff0c\u4f20\u7edf\u76d1\u63a7\u624b\u6bb5\u5df2\u96be\u4ee5\u5e94\u5bf9\u65e5\u76ca\u590d\u6742\u7684\u7cfb\u7edf\u8fd0\u7ef4\u6311\u6218\u3002\u5f53\u4e07\u4ebf\u7ea7\u6d6e\u70b9\u8fd0\u7b97\u4e0ePB\u7ea7\u6570\u636e\u541e\u5410\u6210\u4e3a\u5e38\u6001\uff0c\u5982\u4f55\u5b9e\u73b0\u4ece\u82af\u7247\u5230\u96c6\u7fa4\u7684\u5168\u94fe\u8def\u6027\u80fd\u900f\u89c6\uff1f\u8fd9\u6b63\u662f\u5168\u6808\u53ef\u89c2\u6d4b\u6027\u6280\u672f\u5c55\u73b0\u5176\u72ec\u7279\u4ef7\u503c\u7684\u5173\u952e\u6218\u573a\u3002\u672c\u6587\u5c06\u6df1\u5165\u63a2\u8ba8\u8fd9\u4e00\u6280\u672f\u5982\u4f55\u4e3aHPC\u73af\u5883\u63d0\u4f9b\u7cbe\u51c6\u7684\u201c\u8bca\u65ad\u663e\u5fae\u955c\u201d\u4e0e\u201c\u4f18\u5316\u5bfc\u822a\u4eea\u201d\u3002 \u4e00\u3001HPC\u73af\u5883\u7684\u72ec\u7279\u6311\u6218 \u9ad8\u6027\u80fd\u8ba1\u7b97\u73af\u5883\u901a\u5e38\u7531\u5f02\u6784\u8ba1\u7b97\u8282\u70b9\uff08CPU\u3001GPU\u3001FPGA\uff09\u3001\u9ad8\u901f\u7f51\u7edc\uff08\u5982InfiniBand\uff09\u3001\u5206\u5e03\u5f0f\u5b58\u50a8\u7cfb\u7edf\u53ca\u590d\u6742\u8c03\u5ea6\u6846\u67b6\u6784\u6210\u3002\u8fd9\u79cd\u67b6\u6784\u5728\u5e26\u6765\u5f3a\u5927\u7b97\u529b\u7684\u540c\u65f6\uff0c\u4e5f\u57cb\u85cf\u7740\u591a\u91cd\u8fd0\u7ef4\u9690\u60a3\uff1a \u7cfb\u7edf\u590d\u6742\u6027\u6307\u6570\u589e\u957f\uff1a\u8de8\u5c42\u7ec4\u4ef6\u7684\u8026\u5408\u6027\u5bfc\u81f4\u5355\u4e00\u8282\u70b9\u6545\u969c\u53ef\u80fd\u5f15\u53d1\u7ea7\u8054\u6548\u5e94\uff1b \u5b9e\u65f6\u6027\u8981\u6c42\u4e25\u82db\uff1a\u90e8\u5206\u79d1\u5b66\u8ba1\u7b97\u4efb\u52a1\u9700\u5728\u6570\u5c0f\u65f6\u5185\u5b8c\u6210\uff0c\u4efb\u4f55\u6027\u80fd\u6ce2\u52a8\u90fd\u53ef\u80fd\u5f71\u54cd\u7814\u7a76\u8fdb\u5ea6\uff1b \u6545\u969c\u5b9a\u4f4d\u56f0\u96be\uff1a\u4f20\u7edf\u65e5\u5fd7\u5206\u6790\u96be\u4ee5\u5728\u4ebf\u7ea7\u5e76\u53d1\u7ebf\u7a0b\u4e2d\u5feb\u901f\u5b9a\u4f4d\u5f02\u5e38\u6839\u6e90\uff1b \u8d44\u6e90\u5229\u7528\u7387\u74f6\u9888\uff1a\u636eIDC\u7814\u7a76\uff0c\u5168\u7403HPC\u96c6\u7fa4\u5e73\u5747\u5229\u7528\u7387\u4e0d\u8db360%\uff0c\u9690\u6027\u8d44\u6e90\u6d6a\u8d39\u4e25\u91cd\u3002 \u8fd9\u4e9b\u75db\u70b9\u547c\u5524\u4e00\u79cd\u80fd\u591f\u7a7f\u900f\u786c\u4ef6\u3001\u8f6f\u4ef6\u3001\u7f51\u7edc\u5c42\uff0c\u5e76\u63d0\u4f9b\u5b9e\u65f6\u6d1e\u89c1\u7684\u89c2\u6d4b\u4f53\u7cfb\u3002 \u4e8c\u3001\u5168\u6808\u53ef\u89c2\u6d4b\u6027\u7684\u6280\u672f\u5185\u6db5 \u5168\u6808\u53ef\u89c2\u6d4b\u6027\u5e76\u975e\u7b80\u5355\u7684\u76d1\u63a7\u5347\u7ea7\uff0c\u800c\u662f\u901a\u8fc7\u6307\u6807\uff08Metrics\uff09\u3001\u65e5\u5fd7\uff08Logs\uff09\u3001\u8ffd\u8e2a\uff08Traces\uff09\u7684\u4e09\u7ef4\u6570\u636e\u878d\u5408\uff0c\u6784\u5efa\u7cfb\u7edf\u7684\u201c\u6570\u5b57\u5b6a\u751f\u201d\u3002\u5176\u6838\u5fc3\u4f18\u52bf\u4f53\u73b0\u5728\uff1a \u5168\u5c40\u53ef\u89c6\u5316\uff1a\u4ece\u7269\u7406\u670d\u52a1\u5668\u529f\u8017\u5230\u5e94\u7528\u7ebf\u7a0b\u72b6\u6001\uff0c\u5b9e\u73b0\u8de8\u5c42\u6570\u636e\u5173\u8054\uff1b \u667a\u80fd\u6839\u56e0\u5206\u6790\uff1a\u7ed3\u5408\u673a\u5668\u5b66\u4e60\uff0c\u81ea\u52a8\u8bc6\u522b\u5f02\u5e38\u6a21\u5f0f\u5e76\u8ffd\u6eaf\u6e90\u5934\uff1b \u9884\u6d4b\u6027\u8fd0\u7ef4\uff1a\u901a\u8fc7\u5386\u53f2\u6570\u636e\u5efa\u6a21\uff0c\u9884\u5224\u786c\u4ef6\u6545\u969c\u6216\u6027\u80fd\u62d0\u70b9\u3002 \u5728HPC\u573a\u666f\u4e2d\uff0c\u8fd9\u610f\u5473\u7740\u80fd\u591f\u540c\u65f6\u6355\u6349\u8ba1\u7b97\u8282\u70b9\u7684\u6e29\u5ea6\u66f2\u7ebf\u3001MPI\u901a\u4fe1\u5ef6\u8fdf\u3001\u4f5c\u4e1a\u8c03\u5ea6\u961f\u5217\u72b6\u6001\u7b49\u5173\u952e\u4fe1\u53f7\uff0c\u5e76\u5c06\u5176\u8f6c\u5316\u4e3a\u53ef\u64cd\u4f5c\u7684\u4f18\u5316\u5efa\u8bae\u3002 \u4e09\u3001\u5168\u6808\u53ef\u89c2\u6d4b\u6027\u5728\u9ad8\u6027\u80fd\u8ba1\u7b97\u4e2d\u7684\u5b9e\u8df5\u8def\u5f84 1. \u57fa\u7840\u67b6\u6784\u5c42\u7684\u6df1\u5ea6\u76d1\u63a7 \u901a\u8fc7\u90e8\u7f72Prometheus\u3001Grafana\u7b49\u5de5\u5177\uff0c\u5b9e\u65f6\u91c7\u96c6CPU\/GPU\u5229\u7528\u7387\u3001\u5185\u5b58\u5e26\u5bbd\u3001\u7f51\u7edc\u4e22\u5305\u7387\u7b49\u786c\u4ef6\u6307\u6807\u3002\u4f8b\u5982\uff0c\u7f8e\u56fd\u67d0\u56fd\u5bb6\u5b9e\u9a8c\u5ba4\u901a\u8fc7\u529f\u8017\u5efa\u6a21\uff0c\u53d1\u73b012%\u7684\u8ba1\u7b97\u8282\u70b9\u56e0\u6563\u70ed\u4e0d\u826f\u5bfc\u81f4\u9891\u7387\u964d\u9891\uff0c\u901a\u8fc7\u8c03\u6574\u673a\u67dc\u5e03\u5c40\u540e\uff0c\u4efb\u52a1\u5b8c\u6210\u65f6\u95f4\u7f29\u77ed18%\u3002 2. \u5e94\u7528\u6027\u80fd\u7684\u7ec6\u7c92\u5ea6\u5206\u6790 \u501f\u52a9eBPF\u6280\u672f\uff0c\u5728\u4e0d\u4fb5\u5165\u4ee3\u7801\u7684\u60c5\u51b5\u4e0b\u8ffd\u8e2a\u79d1\u5b66\u8ba1\u7b97\u5e94\u7528\u7684\u5185\u6838\u6001\u884c\u4e3a\u3002\u67d0\u6c14\u8c61\u6a21\u62df\u56e2\u961f\u4f7f\u7528\u52a8\u6001\u706b\u7130\u56fe\uff0c\u5b9a\u4f4d\u5230\u6c14\u5019\u6a21\u578b\u4e2d\u7684\u9501\u7ade\u4e89\u95ee\u9898\uff0c\u4f18\u5316\u540e\u5355\u6b21\u6a21\u62df\u8017\u65f6\u51cf\u5c1123%\u3002 3. \u6570\u636e\u6d41\u6c34\u7ebf\u7684\u7aef\u5230\u7aef\u4f18\u5316 \u5728\u57fa\u56e0\u7ec4\u5b66\u7b49\u6570\u636e\u5bc6\u96c6\u578b\u573a\u666f\u4e2d\uff0c\u5206\u5e03\u5f0f\u8ffd\u8e2a\u7cfb\u7edf\uff08\u5982Jaeger\uff09\u53ef\u7ed8\u5236\u8de8\u5b58\u50a8\u8282\u70b9\u3001\u8ba1\u7b97\u96c6\u7fa4\u7684\u6570\u636e\u6d41\u56fe\u8c31\u3002\u67d0\u57fa\u56e0\u6d4b\u5e8f\u516c\u53f8\u901a\u8fc7\u5206\u6790\u5bf9\u8c61\u5b58\u50a8\u7684IO\u8def\u5f84\uff0c\u5c06\u6570\u636e\u5904\u7406\u541e\u5410\u91cf\u63d0\u534740%\u3002 4. \u8d44\u6e90\u8c03\u5ea6\u5668\u7684\u667a\u80fd\u51b3\u7b56\u652f\u6301 \u5c06\u89c2\u6d4b\u6570\u636e\u53cd\u9988\u81f3Slurm\u3001Kubernetes\u7b49\u8c03\u5ea6\u5668\uff0c\u5b9e\u73b0\u52a8\u6001\u8d44\u6e90\u5206\u914d\u3002\u5fb7\u56fd\u67d0\u8d85\u7b97\u4e2d\u5fc3\u901a\u8fc7\u5b9e\u65f6\u8d1f\u8f7d\u9884\u6d4b\u7b97\u6cd5\uff0c\u4f7f\u96c6\u7fa4\u5229\u7528\u7387\u4ece58%\u63d0\u5347\u81f376%\uff0c\u5e74\u8282\u7701\u7535\u8d39\u8d85200\u4e07\u6b27\u5143\u3002 \u56db\u3001\u5178\u578b\u6848\u4f8b\u4e2d\u7684\u4ef7\u503c\u9a8c\u8bc1 \u6848\u4f8b1\uff1a\u80fd\u6e90\u52d8\u63a2\u4e2d\u7684\u5e76\u884c\u8ba1\u7b97\u4f18\u5316 \u67d0\u77f3\u6cb9\u516c\u53f8\u4f7f\u7528\u5168\u6808\u89c2\u6d4b\u5e73\u53f0\uff0c\u53d1\u73b0\u5176\u5730\u9707\u53cd\u6f14\u4f5c\u4e1a\u7684MPI\u901a\u4fe1\u5f00\u9500\u5360\u6bd4\u9ad8\u8fbe35%\u3002\u901a\u8fc7\u91cd\u6784\u4efb\u52a1\u5206\u533a\u7b56\u7565\u5e76\u4f18\u5316\u7f51\u7edc\u7f13\u51b2\u533a\u914d\u7f6e\uff0c\u6574\u4f53\u8ba1\u7b97\u6548\u7387\u63d0\u534728%\uff0c\u5355\u4e2a\u6cb9\u85cf\u6a21\u578b\u6c42\u89e3\u5468\u671f\u4ece14\u5929\u538b\u7f29\u81f310\u5929\u3002 \u6848\u4f8b2\uff1aAI\u8bad\u7ec3\u96c6\u7fa4\u7684\u6545\u969c\u81ea\u6108 \u4e00\u5bb6\u81ea\u52a8\u9a7e\u9a76\u4f01\u4e1a\u5728\u5176GPU\u96c6\u7fa4\u4e2d\u90e8\u7f72\u4e86\u5f02\u5e38\u68c0\u6d4b\u7ba1\u9053\u3002\u5f53\u7cfb\u7edf\u81ea\u52a8\u8bc6\u522b\u5230\u67d0\u6279\u6b21\u663e\u5361\u7684ECC\u9519\u8bef\u7387\u5f02\u5e38\u6500\u5347\u65f6\uff0c\u7acb\u5373\u9694\u79bb\u6545\u969c\u5361\u5e76\u8fc1\u79fb\u4efb\u52a1\uff0c\u907f\u514d\u4e86\u4e00\u6b21\u53ef\u80fd\u6301\u7eed6\u5c0f\u65f6\u7684\u670d\u52a1\u4e2d\u65ad\u3002 \u6848\u4f8b3\uff1a\u8de8\u5730\u57df\u8d85\u7b97\u7684\u8d44\u6e90\u534f\u540c \u6b27\u6d32\u67d0\u8054\u5408\u8d85\u7b97\u9879\u76ee\u901a\u8fc7\u7edf\u4e00\u89c2\u6d4b\u5e73\u53f0\uff0c\u6574\u5408\u4e94\u5730\u6570\u636e\u4e2d\u5fc3\u7684\u95f2\u7f6e\u7b97\u529b\u3002\u5229\u7528\u52a8\u6001\u4f18\u5148\u7ea7\u8c03\u6574\u7b97\u6cd5\uff0c\u7d27\u6025\u6c14\u5019\u6a21\u62df\u4efb\u52a1\u7684\u6392\u961f\u7b49\u5f85\u65f6\u95f4\u4ece72\u5c0f\u65f6\u964d\u81f39\u5c0f\u65f6\uff0c\u52a9\u529b\u6781\u7aef\u5929\u6c14\u9884\u8b66\u54cd\u5e94\u3002 \u4e94\u3001\u6280\u672f\u6f14\u8fdb\u4e0e\u672a\u6765\u8d8b\u52bf \u968f\u7740HPC\u4e0e\u4e91\u539f\u751f\u6280\u672f\u7684\u878d\u5408\uff0c\u5168\u6808\u53ef\u89c2\u6d4b\u6027\u6b63\u671d\u7740\u4e09\u4e2a\u65b9\u5411\u8fdb\u5316\uff1a \u91cf\u5b50\u8ba1\u7b97\u517c\u5bb9\u6027\uff1a\u5f00\u53d1\u80fd\u89e3\u6790\u91cf\u5b50\u6bd4\u7279\u72b6\u6001\u7684\u4e13\u7528\u63a2\u9488\uff1b \u8fb9\u7f18-\u4e91\u534f\u540c\u89c2\u6d4b\uff1a\u5728\u8054\u90a6\u5b66\u4e60\u7b49\u573a\u666f\u4e2d\u5b9e\u73b0\u4f4e\u5ef6\u8fdf\u6570\u636e\u540c\u6b65\uff1b \u6570\u5b57\u5b6a\u751f\u4eff\u771f\uff1a\u901a\u8fc7\u5386\u53f2\u6570\u636e\u6784\u5efa\u865a\u62df\u96c6\u7fa4\uff0c\u9884\u6f14\u4f18\u5316\u7b56\u7565\u3002 MIT\u8ba1\u7b97\u673a\u79d1\u5b66\u5b9e\u9a8c\u5ba4\u7684\u8fd1\u671f\u5b9e\u9a8c\u8868\u660e\uff0cAI\u9a71\u52a8\u7684\u89c2\u6d4b\u7cfb\u7edf\u5df2\u80fd\u572815\u5206\u949f\u5185\u81ea\u4e3b\u5b8c\u6210\u4f20\u7edf\u56e2\u961f\u97002\u5929\u5206\u6790\u7684\u8c03\u4f18\u4efb\u52a1\uff0c\u9884\u793a\u7740\u4e00\u573a\u667a\u80fd\u8fd0\u7ef4\u9769\u547d\u7684\u5230\u6765\u3002 \u6ce8\uff1a\u672c\u6587\u6570\u636e\u4e0e\u6848\u4f8b\u5747\u6765\u81ea\u516c\u5f00\u5b66\u672f\u8bba\u6587\u3001\u884c\u4e1a\u767d\u76ae\u4e66\u53ca\u6280\u672f\u4f1a\u8bae\u62a5\u544a\uff0c\u90e8\u5206\u7ec6\u8282\u7ecf\u8131\u654f\u5904\u7406\u3002<\/p>\n","protected":false},"author":4,"featured_media":34421,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[14],"tags":[],"class_list":["post-34420","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-product-news"],"_links":{"self":[{"href":"https:\/\/www.yunshan.net\/blog\/wp-json\/wp\/v2\/posts\/34420","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.yunshan.net\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.yunshan.net\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.yunshan.net\/blog\/wp-json\/wp\/v2\/users\/4"}],"replies":[{"embeddable":true,"href":"https:\/\/www.yunshan.net\/blog\/wp-json\/wp\/v2\/comments?post=34420"}],"version-history":[{"count":1,"href":"https:\/\/www.yunshan.net\/blog\/wp-json\/wp\/v2\/posts\/34420\/revisions"}],"predecessor-version":[{"id":34422,"href":"https:\/\/www.yunshan.net\/blog\/wp-json\/wp\/v2\/posts\/34420\/revisions\/34422"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.yunshan.net\/blog\/wp-json\/wp\/v2\/media\/34421"}],"wp:attachment":[{"href":"https:\/\/www.yunshan.net\/blog\/wp-json\/wp\/v2\/media?parent=34420"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.yunshan.net\/blog\/wp-json\/wp\/v2\/categories?post=34420"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.yunshan.net\/blog\/wp-json\/wp\/v2\/tags?post=34420"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}