[{"data":1,"prerenderedAt":464},["ShallowReactive",2],{"$fFIKozURKdo1bwIDi2eGUNbLmwf1Kj5ABWv6QzTi26fY":3,"$fI79wSYuuuDTO6RTdSFlqSkAODCckeNXsPC8Fs34KYIg":25,"$f2WRIUoEiKLec8qerHkiEMrSMbyD1fy9tJBwJmscWQDI":233,"$fXATp09J0q46YT-U2AnrLPheVyoEZthlEDIP_p5hInb0":241,"$f-swQOr5RZ89Tl31CpWDpjPVdeKzduMHHnRG3pfiuwhM":261,"$fSI6AGHX4Ow3Lv-t1ECo2cGR4rMrUzVD0AUMyx6dl0l8":270,"$fPMsAScDrDJ0a8gP6P_5kFfdXr_P7bma8vR8Z8rv6G_A":367},{"data":4,"status":23,"timestamp":24},{"tenantId":5,"id":6,"name":7,"logo":8,"url":9,"desc":10,"content":11,"status":6,"userId":5,"createdAt":12,"updatedAt":13,"languages":14},2,1,"万博智云新","\u002Ffile\u002F2\u002F2024\u002F3\u002Fa357ffd7d263.png","https:\u002F\u002Fwww.oneprocloud.com.cn",null,"{\"header\":{\"zh-cn\":{\"id\":\"secion-zqo8p4\",\"templateName\":\"HeaderView\",\"name\":\"HeaderView\",\"model\":{\"cols\":[{\"width\":20,\"elements\":[{\"id\":\"ele-1duwbw\",\"type\":\"image\",\"attrs\":{\"src\":\"\u002Ffile\u002F2\u002F2025\u002F1\u002Fe48ddd99f554.png?i=7416&name=中文logo反白.png\"},\"style\":{\"width\":129,\"height\":29.555136986301367,\"left\":\"\",\"top\":\"\",\"boxShadow\":\"transparent 0px 0px 0px 0px\",\"margin\":\"15.416% 0% 0% 0.834%\"},\"styleMobile\":{\"top\":\"\",\"left\":\"\",\"margin\":\"33.749% 0% 0% 13.751%\"}}],\"id\":\"col-48azgy\"},{\"width\":65,\"elements\":[{\"id\":\"ele-g6ejqi\",\"type\":\"nav\",\"styleHover\":{},\"style\":{\"marginTop\":0,\"marginLeft\":0,\"height\":100},\"styleMobile\":{\"width\":15.207999999999998,\"left\":\"\",\"top\":\"\",\"margin\":\"-0.048% 0% 0% 104.776%\"}}],\"id\":\"col-mxd7cv\"},{\"width\":15,\"style\":{},\"elements\":[],\"id\":\"col-227wm7\"}],\"setting\":{\"floatType\":\"preset-2\",\"style\":{\"backgroundColor\":\"rgb(0,0,0)\",\"padding\":0,\"color\":\"#fff\",\"height\":\"100px\",\"boxShadow\":\"transparent 0px 0px 0px 0px\"},\"rollStyle\":{\"boxShadow\":\"transparent 0px 0px 0px 0px\"},\"hoverStyle\":{\"boxShadow\":\"transparent 0px 0px 0px 0px\"}}},\"type\":\"header\",\"eles\":[]},\"en\":{\"id\":\"secion-hx89la\",\"templateName\":\"HeaderView\",\"name\":\"HeaderView\",\"model\":{\"cols\":[{\"width\":20,\"elements\":[{\"id\":\"ele-17kxx9\",\"type\":\"image\",\"attrs\":{\"src\":\"\u002Fimages\u002Flogo.png\"},\"style\":{\"width\":202,\"height\":28,\"margin\":\"15% 0% 0% 0%\"}}],\"id\":\"col-22ih6h\"},{\"width\":65,\"elements\":[{\"id\":\"ele-1w00xz\",\"type\":\"nav\",\"styleHover\":{},\"style\":{\"marginTop\":0,\"marginLeft\":0,\"height\":100},\"arrow\":\"eicon eicon-arrow\"}],\"id\":\"col-o5eytg\"},{\"width\":15,\"style\":{},\"elements\":[],\"id\":\"col-vtlgv4\"}],\"setting\":{\"floatType\":\"\",\"style\":{\"backgroundColor\":\"rgb(0,0,0)\",\"padding\":0,\"color\":\"#fff\",\"height\":\"100px\"},\"rollStyle\":{},\"hoverStyle\":{}}},\"type\":\"header\",\"eles\":[]}},\"footer\":{\"zh-cn\":{\"id\":\"secion-x10vnz\",\"templateName\":\"FooterView\",\"name\":\"FooterView\",\"model\":{\"gap\":30,\"columns\":3,\"flexWrap\":\"wrap\",\"col\":{\"style\":{\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"padding\":\"0 0 40px 0\"},\"elements\":[{\"id\":\"ele-it6x05\",\"type\":\"text\",\"content\":\"版权所有 ©2021 万博智云信息科技（上海）有限公司 \",\"style\":{\"width\":30.333,\"color\":\"var(--el-text-color-secondary)\",\"left\":0,\"top\":0,\"fontSize\":15,\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"margin\":\"0% 0% 0% 0.251%\"},\"styleMobile\":{\"width\":82.17999999999999,\"left\":\"\",\"top\":\"\",\"margin\":\"1.25% 0% 0% 6.737%\"}},{\"id\":\"ele-1eoa87\",\"type\":\"image\",\"attrs\":{\"src\":\"\u002Ffile\u002F2\u002F2024\u002F12\u002F870012aae24f.png?i=1353&name=Q5TAj77st3G4idFZw77YbGkW.png\"},\"style\":{\"width\":20,\"height\":20,\"left\":\"\",\"top\":\"\",\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"padding\":\"0 0 0px 0\",\"margin\":\"9.583px 0px 0px 5.751px\"},\"link\":{\"type\":\"url\",\"target\":\"_blank\",\"url\":\"https:\u002F\u002Fwww.beian.gov.cn\u002Fportal\u002FregisterSystemInfo?recordcode=31011702005563\"},\"styleMobile\":{\"top\":\"\",\"left\":\"\",\"margin\":\"-4.042% 0% 0% 8.251%\"}},{\"id\":\"ele-1h8m2s\",\"type\":\"text\",\"content\":\"沪公网安备31011502404785号\",\"style\":{\"width\":18.083,\"left\":\"\",\"top\":\"\",\"fontSize\":15,\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"color\":\"#909399\",\"margin\":\"-1.667% 0% 0% 2.5%\"},\"link\":{\"type\":\"url\",\"target\":\"_blank\",\"url\":\"https:\u002F\u002Fwww.beian.gov.cn\u002Fportal\u002FregisterSystemInfo?recordcode=31011502404785\"},\"styleMobile\":{\"width\":47.875,\"left\":\"\",\"top\":\"\",\"margin\":\"-4.167% 0% 0% 13.542%\"}},{\"id\":\"ele-j69h58\",\"type\":\"text\",\"content\":\"沪ICP备 19015066号-2  \",\"style\":{\"width\":17.333,\"left\":\"\",\"top\":\"\",\"fontSize\":15,\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"color\":\"#909399\",\"margin\":\"-1.917% 0% 0% 20.499%\"},\"link\":{\"type\":\"url\",\"target\":\"_blank\",\"url\":\"https:\u002F\u002Fbeian.miit.gov.cn\"},\"styleMobile\":{\"width\":37,\"left\":\"\",\"top\":\"\",\"margin\":\"-4.763% 0% 0% 62.388%\"}}],\"id\":\"col-1f3wnk\"},\"cols\":[{\"elements\":[{\"id\":\"ele-oytr0m\",\"type\":\"image\",\"attrs\":{\"src\":\"\u002Ffile\u002F2\u002F2025\u002F1\u002Fe48ddd99f554.png?i=7416&name=中文logo反白.png\"},\"style\":{\"width\":126,\"height\":28.867808219178077,\"left\":0,\"top\":0,\"marginTop\":null,\"marginLeft\":null}},{\"id\":\"ele-nzytr9\",\"type\":\"text\",\"content\":\"地址：上海市浦东新区峨山路101号C2栋\",\"style\":{\"width\":100,\"color\":\"#c8c9cc\",\"fontSize\":15,\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"margin\":\"10% 0% 0% 0%\"}},{\"id\":\"ele-icuoxq\",\"type\":\"text\",\"content\":\"邮箱：enquiry@oneprocloud.com\",\"style\":{\"width\":100,\"color\":\"var(--el-color-info-light-5)\",\"fontSize\":15,\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"margin\":\"2% 0% 0% 0%\"}},{\"id\":\"ele-4fjtdq\",\"type\":\"text\",\"content\":\"电话：180 1760 6559\",\"style\":{\"width\":100,\"color\":\"var(--el-color-info-light-5)\",\"fontSize\":15,\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"margin\":\"2% 0% 0% 0%\"}},{\"id\":\"ele-az0kym\",\"type\":\"image\",\"attrs\":{\"src\":\"\u002Fimages\u002Fmodule\u002Ffooter\u002Fb-logo.png\"},\"style\":{\"width\":25,\"height\":24,\"margin\":\"8% 0% 0% 0%\"},\"link\":{\"type\":\"url\",\"target\":\"_self\",\"url\":\"https:\u002F\u002Fspace.bilibili.com\u002F1246058384\"}},{\"id\":\"ele-1ytu9l\",\"type\":\"image\",\"attrs\":{\"src\":\"\u002Fimages\u002Fmodule\u002Ffooter\u002FCSDN.png\"},\"style\":{\"width\":67,\"height\":18,\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"margin\":\"-5% 0% 0% 10%\"},\"link\":{\"type\":\"url\",\"target\":\"_self\",\"url\":\"https:\u002F\u002Fblog.csdn.net\u002FOneProCloud?type=blog\"},\"styleMobile\":{\"top\":\"\",\"left\":\"\",\"margin\":\"-4.111% 0% 0% 9.333%\"}},{\"id\":\"ele-hhqnxe\",\"type\":\"image\",\"attrs\":{\"src\":\"\u002Fimages\u002Fmodule\u002Ffooter\u002Fdouyin.png\"},\"style\":{\"width\":22,\"height\":25,\"margin\":\"-6% 0% 0% 30%\"},\"styleMobile\":{\"top\":\"\",\"left\":\"\",\"margin\":\"-4.667% 0% 0% 27.334%\"}},{\"id\":\"ele-1gqb5m\",\"type\":\"image\",\"attrs\":{\"src\":\"\u002Fimages\u002Fmodule\u002Ffooter\u002Fweixin.png\"},\"style\":{\"width\":28,\"height\":23,\"margin\":\"-6% 0% 0% 40%\"},\"styleMobile\":{\"top\":\"\",\"left\":\"\",\"margin\":\"-4.889% 0% 0% 36.666%\"}},{\"id\":\"ele-5kcwd8\",\"type\":\"image\",\"attrs\":{\"src\":\"\u002Fimages\u002Fmodule\u002Ffooter\u002Fzhihu.png\"},\"style\":{\"width\":24,\"height\":24,\"margin\":\"-6% 0% 0% 50%\"},\"link\":{\"type\":\"url\",\"target\":\"_self\",\"url\":\"https:\u002F\u002Fwww.zhihu.com\u002Forg\u002Fmo-bo-zhi-yun-84\"},\"styleMobile\":{\"top\":\"\",\"left\":\"\",\"margin\":\"-5.333% 0% 0% 47.111%\"}}],\"id\":\"col-1fttw9\"},{\"elements\":[{\"id\":\"ele-mwg0s5\",\"type\":\"text\",\"content\":\"产品&方案\",\"style\":{\"width\":100,\"fontSize\":18,\"fontWeight\":\"bold\",\"color\":\"var(--el-color-white)\"}},{\"id\":\"ele-b972f7\",\"type\":\"rect\",\"style\":{\"width\":50,\"height\":20,\"borderColor\":\"var(--el-color-primary)\",\"borderTopWidth\":1,\"borderLeftWidth\":0,\"borderRightWidth\":0,\"borderBottomWidth\":0,\"borderStyle\":\"solid\"}},{\"id\":\"ele-1190bs\",\"type\":\"text\",\"content\":\"HyperBDR\u003Cbr>云容灾\",\"style\":{\"width\":50,\"color\":\"var(--el-color-info-light-5)\",\"fontSize\":15,\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"left\":0,\"top\":0,\"margin\":\"2.474% 0% 0% 0%\"},\"link\":{\"type\":\"page\",\"target\":\"_self\",\"languageId\":1,\"languageCode\":\"zh-cn\",\"pageId\":4}},{\"id\":\"ele-jliiet\",\"type\":\"text\",\"content\":\"云上保护\",\"style\":{\"width\":50,\"color\":\"var(--el-color-info-light-5)\",\"left\":0,\"top\":0,\"fontSize\":15,\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"margin\":\"-12.58% 0% 0% 50.526%\"},\"link\":{\"type\":\"page\",\"target\":\"_self\",\"languageId\":1,\"languageCode\":\"zh-cn\",\"pageId\":66}},{\"id\":\"ele-1jwt20\",\"type\":\"text\",\"content\":\"HyperMotion\u003Cbr \u002F>云迁移\",\"style\":{\"width\":50,\"color\":\"var(--el-color-info-light-5)\",\"fontSize\":15,\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"left\":0,\"top\":0,\"margin\":\"8.789% 0% 0% 0%\"},\"link\":{\"type\":\"page\",\"target\":\"_self\",\"languageId\":1,\"languageCode\":\"zh-cn\",\"pageId\":3}},{\"id\":\"ele-1usquk\",\"type\":\"text\",\"content\":\"金融行业容灾方案\",\"style\":{\"width\":50,\"color\":\"var(--el-color-info-light-5)\",\"left\":0,\"top\":0,\"fontSize\":15,\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"margin\":\"-14.684% 0% 0% 50.789%\"},\"link\":{\"type\":\"page\",\"target\":\"_self\",\"languageId\":1,\"languageCode\":\"zh-cn\",\"pageId\":42}},{\"id\":\"ele-1gmjkk\",\"type\":\"text\",\"content\":\"EasyBill\u003Cbr>账单易\",\"style\":{\"width\":50,\"color\":\"var(--el-color-info-light-5)\",\"fontSize\":15,\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"left\":0,\"top\":0,\"margin\":\"10.105% 0% 0% 0.263%\"},\"link\":{\"type\":\"page\",\"target\":\"_self\",\"languageId\":1,\"languageCode\":\"zh-cn\",\"pageId\":6},\"lock\":true}],\"id\":\"col-1kph22\"},{\"elements\":[{\"id\":\"ele-1vopaw\",\"type\":\"rss-input\",\"style\":{\"width\":100},\"btnText\":\"点击发送\",\"inputTip\":\"邮箱地址\",\"checkTitle\":\"\u003Cp>\u003Cspan style=\\\"font-size:12px;\\\">提交此表格即表示您同意万博智云将根据隐私政策处理您的个人信息。\u003C\u002Fspan>\u003C\u002Fp>\",\"emailEmptyTip\":\"请输入邮箱\",\"submitSuccTip\":\"提交成功\"},{\"id\":\"ele-1ohmoz\",\"type\":\"text\",\"content\":\"了解万博智云的一手产品和解决方案讯息，关注最新的新闻动态，接受我们的最新资讯推送\",\"style\":{\"width\":100,\"color\":\"var(--el-color-info-light-5)\",\"fontSize\":15,\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"margin\":\"6% 0% 0% 0%\"}}],\"id\":\"col-216gqq\"}]},\"style\":{\"backgroundColor\":\"var(--el-color-black)\",\"padding\":\"70px 0px 0px 0px\"},\"type\":\"footer\",\"eles\":[],\"modelMobile\":{\"columns\":1,\"gap\":30},\"styleMobile\":{\"padding\":\"40px 15px 0px 15px\"}},\"en\":{\"id\":\"secion-170ey8\",\"templateName\":\"FooterView\",\"name\":\"FooterView\",\"model\":{\"gap\":30,\"columns\":3,\"flexWrap\":\"wrap\",\"col\":{\"style\":{},\"elements\":[{\"id\":\"ele-112o6p\",\"type\":\"text\",\"content\":\" ©2024 OnePro Cloud Limited |        \",\"style\":{\"width\":35.833,\"color\":\"#a8a8a8\",\"left\":0,\"top\":0,\"fontSize\":14,\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"margin\":\"-1.917% 0% 0% 32.333%\"}},{\"id\":\"ele-t74tet\",\"type\":\"text\",\"content\":\"Privacy |\",\"style\":{\"width\":4.834000000000003,\"left\":0,\"top\":0,\"fontSize\":14,\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"color\":\"#A8A8A8\",\"zIndex\":51,\"margin\":\"-1.75% 0% 0% 50.25%\"},\"link\":{\"type\":\"page\",\"target\":\"_blank\",\"languageId\":2,\"languageCode\":\"en\",\"pageId\":102}},{\"id\":\"ele-3v0xxc\",\"type\":\"text\",\"content\":\" Cookies |\",\"style\":{\"width\":5.417000000000002,\"left\":0,\"top\":0,\"fontSize\":14,\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"color\":\"#A8A8A8\",\"margin\":\"-1.75% 0% 0% 54.916%\"},\"link\":{\"type\":\"page\",\"target\":\"_blank\",\"languageId\":2,\"languageCode\":\"en\",\"pageId\":103}}],\"id\":\"col-157n4k\"},\"cols\":[{\"elements\":[{\"id\":\"ele-uxr8yb\",\"type\":\"image\",\"attrs\":{\"src\":\"\u002Ffile\u002F2\u002F2024\u002F11\u002F6adadc64700b.png\"},\"style\":{\"width\":183,\"height\":24,\"margin\":\"1% 0% 0% 0%\"}},{\"id\":\"ele-1sthrh\",\"type\":\"text\",\"content\":\"Adress：Unit 701A, 7\u002FF, Railway Plaza, 39 Chatham Road South, Tsim Sha Tsui, Hong Kong\",\"style\":{\"width\":100,\"color\":\"rgb(194, 194, 194)\",\"margin\":\"10% 0% 0% 0%\"}},{\"id\":\"ele-dvsxo1\",\"type\":\"text\",\"content\":\"Email：enquiry@oneprocloud.com\",\"style\":{\"width\":100,\"color\":\"rgb(194, 194, 194)\",\"fontSize\":16,\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"margin\":\"2% 0% 0% 0%\"}},{\"id\":\"ele-k6p458\",\"type\":\"text\",\"content\":\"\u003Cbr>\",\"style\":{\"width\":100,\"color\":\"rgb(194, 194, 194)\",\"margin\":\"2% 0% 0% 0%\"}},{\"id\":\"ele-gmn4n9\",\"type\":\"image\",\"attrs\":{\"src\":\"\u002Ffile\u002F2\u002F2024\u002F11\u002Ffb24853c2fe4.png\"},\"style\":{\"width\":48,\"height\":43,\"left\":0,\"top\":0,\"margin\":\"9.474% 0% 0% -0.263%\"},\"link\":{\"type\":\"url\",\"target\":\"_blank\",\"url\":\"https:\u002F\u002Fwww.instagram.com\u002Fonepro.cloud\u002Fprofilecard\u002F?igsh=cnMyOWxpcXBldHNx\"}},{\"id\":\"ele-5trm5k\",\"type\":\"image\",\"attrs\":{\"src\":\"\u002Ffile\u002F2\u002F2024\u002F11\u002F8930955f1892.png\"},\"style\":{\"width\":48,\"height\":43,\"left\":0,\"top\":0,\"margin\":\"-11.316% 0% 0% 14.736%\"},\"link\":{\"type\":\"url\",\"target\":\"_blank\",\"url\":\"https:\u002F\u002Fwww.linkedin.com\u002Fcompany\u002Fonepro-cloud\u002F\"}},{\"id\":\"ele-cirko5\",\"type\":\"image\",\"attrs\":{\"src\":\"\u002Ffile\u002F2\u002F2024\u002F11\u002Fb604492d79e5.png\"},\"style\":{\"width\":43,\"height\":44.21759570181331,\"left\":0,\"top\":0,\"margin\":\"-12.105% 0% 0% 29.473%\"},\"link\":{\"type\":\"url\",\"target\":\"_blank\",\"url\":\"https:\u002F\u002Fwww.youtube.com\u002F@OneProCloud\"}}],\"style\":{\"height\":257},\"id\":\"col-ovom5n\"},{\"elements\":[{\"id\":\"ele-72jyoi\",\"type\":\"text\",\"content\":\"Products                    Solutions\",\"style\":{\"width\":100,\"fontSize\":18,\"fontWeight\":\"bold\",\"color\":\"rgb(194, 194, 194)\",\"left\":0,\"top\":0,\"marginTop\":0,\"marginLeft\":0}},{\"id\":\"ele-1ijd0u\",\"type\":\"rect\",\"style\":{\"width\":50,\"height\":20,\"borderColor\":\"var(--el-color-primary)\",\"borderTopWidth\":1,\"borderLeftWidth\":0,\"borderRightWidth\":0,\"borderBottomWidth\":0,\"borderStyle\":\"solid\"}},{\"id\":\"ele-1xn13v\",\"type\":\"text\",\"content\":\"HyperBDR\u003Cbr>\",\"style\":{\"width\":50,\"color\":\"rgb(194, 194, 194)\",\"left\":0,\"top\":0,\"margin\":\"4.842% 0% 0% 0.263%\"},\"link\":{\"type\":\"page\",\"target\":\"_blank\",\"languageId\":2,\"languageCode\":\"en\",\"pageId\":76}},{\"id\":\"ele-1uw5fq\",\"type\":\"text\",\"content\":\"VMware Migration\",\"style\":{\"width\":50,\"color\":\"rgb(194, 194, 194)\",\"left\":0,\"top\":0,\"fontSize\":16,\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"margin\":\"-6.263% 0% 0% 49.737%\"},\"link\":{\"type\":\"page\",\"target\":\"_blank\",\"languageId\":2,\"languageCode\":\"en\",\"pageId\":83}},{\"id\":\"ele-1fztq8\",\"type\":\"text\",\"content\":\"HyperMotion\u003Cbr>\u003Cbr>\",\"style\":{\"width\":50,\"color\":\"rgb(194, 194, 194)\",\"fontSize\":16,\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"left\":0,\"top\":0,\"margin\":\"6.684% 0% 0% 0.263%\"},\"link\":{\"type\":\"page\",\"target\":\"_blank\",\"languageId\":2,\"languageCode\":\"en\",\"pageId\":75}},{\"id\":\"ele-1t96gd\",\"type\":\"text\",\"content\":\"Industry DR Solution\",\"style\":{\"width\":50,\"color\":\"rgb(194, 194, 194)\",\"left\":0,\"top\":0,\"fontSize\":16,\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"margin\":\"-11.526% 0% 0% 48.947%\"},\"link\":{\"type\":\"page\",\"target\":\"_blank\",\"languageId\":2,\"languageCode\":\"en\",\"pageId\":85}},{\"id\":\"ele-knzb4p\",\"type\":\"text\",\"content\":\"EasyBill\",\"style\":{\"width\":50,\"color\":\"rgb(194, 194, 194)\",\"left\":0,\"top\":0,\"fontSize\":16,\"boxShadow\":\"transparent 0. 0. 71px 71px 0px 0px\",\"margin\":\"5.369% 0% 0% 0.263%\"},\"link\":{\"type\":\"page\",\"target\":\"_blank\",\"languageId\":2,\"languageCode\":\"en\",\"pageId\":78}}],\"id\":\"col-205765\"},{\"elements\":[{\"id\":\"ele-1t5dsb\",\"type\":\"rss-input\",\"style\":{\"width\":100},\"btnText\":\"Subscribe\",\"inputTip\":\"Email\",\"checkTitle\":\"\u003Cp>\u003Cspan style=\\\"color:#9b59b6;font-size:12px;\\\">By submitting this form, you acknowledge and agree that OnePro will process your personal information in accordance with the Privacy Policy.\u003C\u002Fspan>\u003C\u002Fp>\",\"emailEmptyTip\":\"Email can not be empty\",\"submitSuccTip\":\"Successful\"},{\"id\":\"ele-ek93s0\",\"type\":\"text\",\"content\":\"Stay informed with first hand updates on OnePro's products and solutions, follow the latest news, and subscribe to our notifications.\",\"style\":{\"width\":94.737,\"color\":\"rgb(194, 194, 194)\",\"left\":0,\"top\":0,\"margin\":\"8.684% 0% 0% 3.421%\"}}],\"id\":\"col-1jh3yz\"}]},\"style\":{\"backgroundColor\":\"var(--el-color-black)\"},\"type\":\"footer\",\"eles\":[]}},\"global\":{\"languages\":[{\"title\":\"Cloud Migration, Cloud Disaster Recovery, Fully Automated Migration and DR Solutions | OnePro Cloud\",\"languageId\":2,\"keywords\":\"Remote Disaster Recovery, Cloud Backup and Recovery, VMware  Backup and Recovery, Hybrid Cloud Disaster Recovery, Continuous Data Protection, CDP, VMware Migration, Application Migration, Agentless Migration\",\"description\":\"A leading provider of multi-cloud infrastructure tools, offering enterprises capabilities such as migration and disaster recovery, billing management, compute orchestration, and cost optimization, empowering businesses to accelerate their digital transformation journey.\",\"favicon\":\"\u002Ffile\u002F2\u002F2024\u002F12\u002F46998dbba181.png?i=220&name=万博智云英文LOGO_方.png\"}],\"zh-cn\":{\"seo\":{\"title\":\"云迁移, 云容灾, 全自动整机迁移和容灾方案 | 万博智云官网\",\"languageId\":1,\"description\":\"将物理机、VMware、OpenStack和云主机持续复制到目标云，在故障或演练时自动创建资源并恢复业务。\",\"keywords\":\"云迁移、云容灾、 数据迁移、容灾备份、混合云容灾、 跨云容灾、VMware迁移、VMware容灾、主备容灾、跨云容灾、万博智云\",\"favicon\":\"\u002Ffile\u002F2\u002F2024\u002F12\u002F46998dbba181.png?i=220&name=万博智云英文LOGO_方.png\"},\"modules\":[\"nav\",\"language\",\"title\"],\"theme\":{\"colorPrimary\":\"rgb(146, 33, 100)\"},\"setting\":{}}}}","2024-03-26 14:27:44","2026-09-08 10:14:31",[15,19],{"id":6,"name":16,"code":17,"isDefault":18,"siteId":6},"中文简体","zh-cn",true,{"id":5,"name":20,"code":21,"isDefault":22,"siteId":6},"EN","en",false,200,1790563554905,{"data":26,"status":23,"timestamp":232},[27,35,98,157,184,192],{"tenantId":5,"id":6,"name":28,"type":6,"url":29,"sortIndex":30,"isHome":18,"status":6,"navVisible":6,"publishedAt":31,"contentUpdatedAt":32,"parentId":10,"siteId":6,"languageId":6,"createdAt":33,"updatedAt":32,"siteName":7,"children":34},"首页","\u002F",10,"2026-09-10 11:26:33","2026-09-10 11:33:32","2024-03-25 17:32:13",[],{"tenantId":5,"id":5,"name":36,"type":5,"url":10,"sortIndex":37,"isHome":22,"status":6,"navVisible":6,"publishedAt":10,"contentUpdatedAt":10,"parentId":10,"siteId":6,"languageId":6,"createdAt":38,"updatedAt":39,"siteName":7,"children":40},"产品",11,"2024-03-25 17:32:55","2024-04-08 10:57:42",[41,49,57,66,90],{"tenantId":5,"id":42,"name":43,"type":6,"url":44,"sortIndex":30,"isHome":22,"status":6,"navVisible":6,"publishedAt":45,"contentUpdatedAt":46,"parentId":5,"siteId":6,"languageId":6,"createdAt":47,"updatedAt":46,"siteName":7,"children":48},3,"HyperMotion云迁移","\u002Fproducts\u002Fhypermotion","2026-09-14 14:18:37","2026-09-14 14:38:24","2024-03-25 17:35:17",[],{"tenantId":5,"id":50,"name":51,"type":6,"url":52,"sortIndex":53,"isHome":22,"status":6,"navVisible":6,"publishedAt":13,"contentUpdatedAt":54,"parentId":5,"siteId":6,"languageId":6,"createdAt":55,"updatedAt":54,"siteName":7,"children":56},4,"HyperBDR云容灾","\u002Fproducts\u002Fhyperbdr",20,"2026-09-14 14:39:51","2024-03-25 17:35:25",[],{"tenantId":5,"id":58,"name":59,"type":6,"url":60,"sortIndex":61,"isHome":22,"status":6,"navVisible":5,"publishedAt":62,"contentUpdatedAt":63,"parentId":5,"siteId":6,"languageId":6,"createdAt":64,"updatedAt":63,"siteName":7,"children":65},107,"价格页面","\u002Fprice",26,"2026-09-09 15:19:27","2026-09-10 11:27:27","2025-12-12 17:26:55",[],{"tenantId":5,"id":67,"name":68,"type":6,"url":69,"sortIndex":70,"isHome":22,"status":6,"navVisible":6,"publishedAt":71,"contentUpdatedAt":72,"parentId":5,"siteId":6,"languageId":6,"createdAt":73,"updatedAt":72,"siteName":7,"children":74},106,"AGIOne  模型AI服务+算力服务","\u002Fproducts\u002Fagione",40,"2026-08-17 17:22:51","2026-09-10 11:28:17","2025-10-20 18:02:36",[75,83],{"tenantId":5,"id":76,"name":77,"type":6,"url":78,"sortIndex":79,"isHome":22,"status":6,"navVisible":6,"publishedAt":80,"contentUpdatedAt":81,"parentId":67,"siteId":6,"languageId":6,"createdAt":80,"updatedAt":81,"siteName":7,"children":82},109,"模型 AI 服务","\u002Fproducts\u002Fagione-modelone",90,"2026-08-14 11:02:03","2026-08-17 15:01:34",[],{"tenantId":5,"id":84,"name":85,"type":6,"url":86,"sortIndex":79,"isHome":22,"status":6,"navVisible":6,"publishedAt":87,"contentUpdatedAt":88,"parentId":67,"siteId":6,"languageId":6,"createdAt":87,"updatedAt":88,"siteName":7,"children":89},110,"算力服务","\u002Fproducts\u002Fagione-powerone","2026-08-14 11:02:19","2026-09-09 17:56:21",[],{"tenantId":5,"id":91,"name":92,"type":6,"url":93,"sortIndex":94,"isHome":22,"status":6,"navVisible":6,"publishedAt":10,"contentUpdatedAt":10,"parentId":5,"siteId":6,"languageId":6,"createdAt":95,"updatedAt":96,"siteName":7,"children":97},6,"EasyBill账单易","\u002Fproducts\u002Feasybill",50,"2024-03-25 17:35:59","2026-04-17 10:42:16",[],{"tenantId":5,"id":99,"name":100,"type":5,"url":10,"sortIndex":101,"isHome":22,"status":6,"navVisible":6,"publishedAt":10,"contentUpdatedAt":10,"parentId":10,"siteId":6,"languageId":6,"createdAt":102,"updatedAt":103,"siteName":7,"children":104},9,"解决方案",13,"2024-03-25 17:36:26","2024-04-08 10:58:05",[105,117,137],{"tenantId":5,"id":106,"name":107,"type":5,"url":10,"sortIndex":61,"isHome":22,"status":6,"navVisible":6,"publishedAt":10,"contentUpdatedAt":10,"parentId":99,"siteId":6,"languageId":6,"createdAt":108,"updatedAt":109,"siteName":7,"children":110},71,"技术场景方案","2024-07-18 10:37:50","2024-11-19 23:25:16",[111],{"tenantId":5,"id":30,"name":112,"type":6,"url":113,"sortIndex":30,"isHome":22,"status":6,"navVisible":6,"publishedAt":114,"contentUpdatedAt":114,"parentId":106,"siteId":6,"languageId":6,"createdAt":115,"updatedAt":114,"siteName":7,"children":116},"从VMware到云","\u002Fsolutions\u002Fvmware-to-cloud","2026-09-09 17:29:10","2024-03-25 17:39:30",[],{"tenantId":5,"id":118,"name":119,"type":5,"url":10,"sortIndex":61,"isHome":22,"status":6,"navVisible":6,"publishedAt":10,"contentUpdatedAt":10,"parentId":99,"siteId":6,"languageId":6,"createdAt":120,"updatedAt":121,"siteName":7,"children":122},72,"行业场景方案","2024-07-18 10:38:34","2024-11-19 23:25:48",[123,129],{"tenantId":5,"id":37,"name":124,"type":6,"url":125,"sortIndex":30,"isHome":22,"status":6,"navVisible":6,"publishedAt":10,"contentUpdatedAt":10,"parentId":118,"siteId":6,"languageId":6,"createdAt":126,"updatedAt":127,"siteName":7,"children":128},"制造业","\u002Fsolutions\u002Fdr-manufacture","2024-03-25 17:39:42","2026-04-17 10:42:43",[],{"tenantId":5,"id":130,"name":131,"type":6,"url":132,"sortIndex":53,"isHome":22,"status":6,"navVisible":6,"publishedAt":133,"contentUpdatedAt":134,"parentId":118,"siteId":6,"languageId":6,"createdAt":135,"updatedAt":134,"siteName":7,"children":136},42,"金融行业","\u002Fsolutions\u002Fdr-finance","2026-09-14 14:33:34","2026-09-14 14:47:59","2024-04-08 10:58:45",[],{"tenantId":5,"id":138,"name":139,"type":5,"url":10,"sortIndex":94,"isHome":22,"status":6,"navVisible":6,"publishedAt":10,"contentUpdatedAt":10,"parentId":99,"siteId":6,"languageId":6,"createdAt":140,"updatedAt":141,"siteName":7,"children":142},69,"应用场景","2024-07-12 17:38:19","2024-07-18 10:39:07",[143,150],{"tenantId":5,"id":144,"name":145,"type":6,"url":146,"sortIndex":53,"isHome":22,"status":6,"navVisible":6,"publishedAt":147,"contentUpdatedAt":147,"parentId":138,"siteId":6,"languageId":6,"createdAt":148,"updatedAt":147,"siteName":7,"children":149},65,"数据流转","\u002Fsolutions\u002Fdata-mobility","2026-09-14 14:19:04","2024-05-11 11:33:03",[],{"tenantId":5,"id":151,"name":152,"type":6,"url":153,"sortIndex":70,"isHome":22,"status":6,"navVisible":6,"publishedAt":154,"contentUpdatedAt":154,"parentId":138,"siteId":6,"languageId":6,"createdAt":155,"updatedAt":154,"siteName":7,"children":156},66,"云上保护","\u002Fsolutions\u002Fcloud-protection","2026-09-10 11:31:27","2024-05-11 11:49:22",[],{"tenantId":5,"id":158,"name":159,"type":5,"url":10,"sortIndex":160,"isHome":22,"status":6,"navVisible":6,"publishedAt":10,"contentUpdatedAt":10,"parentId":10,"siteId":6,"languageId":6,"createdAt":161,"updatedAt":162,"siteName":7,"children":163},12,"文档中心",14,"2024-03-25 17:40:31","2024-04-08 10:57:56",[164,171,177],{"tenantId":5,"id":101,"name":165,"type":42,"url":166,"sortIndex":167,"isHome":22,"status":6,"navVisible":6,"publishedAt":10,"contentUpdatedAt":10,"parentId":158,"siteId":6,"languageId":6,"createdAt":168,"updatedAt":169,"siteName":7,"children":170},"产品使用手册","https:\u002F\u002Fdocs.oneprocloud.com\u002Fzh\u002F",140,"2024-03-25 17:40:51","2025-09-04 17:37:06",[],{"tenantId":5,"id":160,"name":172,"type":6,"url":173,"sortIndex":167,"isHome":22,"status":6,"navVisible":6,"publishedAt":10,"contentUpdatedAt":10,"parentId":158,"siteId":6,"languageId":6,"createdAt":174,"updatedAt":175,"siteName":7,"children":176},"产品白皮书","\u002Fdocumentation\u002Fwhite-papers","2024-03-25 17:41:01","2026-04-17 10:43:05",[],{"tenantId":5,"id":178,"name":179,"type":6,"url":180,"sortIndex":167,"isHome":22,"status":6,"navVisible":6,"publishedAt":10,"contentUpdatedAt":10,"parentId":158,"siteId":6,"languageId":6,"createdAt":181,"updatedAt":182,"siteName":7,"children":183},15,"客户案例","\u002Fdocumentation\u002Fcase-studies","2024-03-25 17:41:10","2026-04-17 10:43:11",[],{"tenantId":5,"id":185,"name":186,"type":6,"url":187,"sortIndex":178,"isHome":22,"status":6,"navVisible":6,"publishedAt":188,"contentUpdatedAt":189,"parentId":10,"siteId":6,"languageId":6,"createdAt":190,"updatedAt":189,"siteName":7,"children":191},16,"博客","\u002Fblog","2026-09-20 11:38:58","2026-09-20 13:57:10","2024-03-25 17:41:18",[],{"tenantId":5,"id":193,"name":194,"type":5,"url":10,"sortIndex":185,"isHome":22,"status":6,"navVisible":6,"publishedAt":10,"contentUpdatedAt":10,"parentId":10,"siteId":6,"languageId":6,"createdAt":195,"updatedAt":39,"siteName":7,"children":196},17,"资源中心","2024-03-25 17:41:29",[197,204,211,218,225],{"tenantId":5,"id":198,"name":199,"type":6,"url":200,"sortIndex":30,"isHome":22,"status":6,"navVisible":6,"publishedAt":10,"contentUpdatedAt":10,"parentId":193,"siteId":6,"languageId":6,"createdAt":201,"updatedAt":202,"siteName":7,"children":203},18,"关于我们","\u002Fresource\u002Fabout-us","2024-03-25 17:41:44","2026-04-17 10:43:22",[],{"tenantId":5,"id":205,"name":206,"type":6,"url":207,"sortIndex":53,"isHome":22,"status":6,"navVisible":6,"publishedAt":10,"contentUpdatedAt":10,"parentId":193,"siteId":6,"languageId":6,"createdAt":208,"updatedAt":209,"siteName":7,"children":210},19,"培训认证","\u002Fresources\u002Ftraining","2024-03-25 17:41:55","2026-04-17 10:43:26",[],{"tenantId":5,"id":53,"name":212,"type":6,"url":213,"sortIndex":214,"isHome":22,"status":6,"navVisible":6,"publishedAt":10,"contentUpdatedAt":10,"parentId":193,"siteId":6,"languageId":6,"createdAt":215,"updatedAt":216,"siteName":7,"children":217},"获得帮助","\u002Fresources\u002Fsupport",30,"2024-03-25 17:42:11","2026-04-17 10:43:30",[],{"tenantId":5,"id":219,"name":220,"type":6,"url":221,"sortIndex":70,"isHome":22,"status":6,"navVisible":5,"publishedAt":222,"contentUpdatedAt":10,"parentId":193,"siteId":6,"languageId":6,"createdAt":223,"updatedAt":222,"siteName":7,"children":224},101,"隐私政策","\u002Fprivacy","2026-09-09 15:19:42","2024-11-19 15:38:33",[],{"tenantId":5,"id":226,"name":227,"type":6,"url":228,"sortIndex":70,"isHome":22,"status":6,"navVisible":5,"publishedAt":229,"contentUpdatedAt":10,"parentId":193,"siteId":6,"languageId":6,"createdAt":230,"updatedAt":229,"siteName":7,"children":231},104,"Cookies条款","\u002Fcookies","2026-09-09 15:19:46","2024-12-03 14:23:44",[],1790563554950,{"data":234,"status":23,"timestamp":240},[235],{"tenantId":5,"id":6,"name":236,"type":5,"content":237,"userId":5,"siteId":6,"status":6,"languageId":6,"createdAt":238,"updatedAt":239},"百度统计oneprocloud.com.cn","\u003Cscript>\nvar _hmt = _hmt || [];\n(function() {\n  var hm = document.createElement(\"script\");\n  hm.src = \"https:\u002F\u002Fhm.baidu.com\u002Fhm.js?3057ee8f05e696c5ac041ba3428c0183\";\n  var s = document.getElementsByTagName(\"script\")[0]; \n  s.parentNode.insertBefore(hm, s);\n})();\n\u003C\u002Fscript>\n","2024-12-30 11:15:49","2025-02-19 10:32:23",1790563555016,{"data":242,"status":23,"timestamp":260},{"tenantId":5,"id":243,"typeId":244,"productId":53,"languageId":6,"siteId":6,"content":245,"html":245,"contentType":246,"desc":247,"cover":248,"slug":249,"coverBig":10,"isTop":22,"topType":10,"title":250,"tags":251,"seoTitle":252,"keywords":253,"description":254,"status":6,"date":255,"publishedAt":256,"contentUpdatedAt":256,"sortIndex":10,"createdAt":256,"updatedAt":256,"productName":257,"productCode":257,"productIcon":258,"blogTypeName":259,"languageName":16},225,25,"\u003Carticle class=\"agione-article\" style=\"color:rgb(48, 50, 58);font-family:-apple-system, BlinkMacSystemFont, "Segoe UI", "PingFang SC", "Microsoft YaHei", sans-serif;font-size:16px;letter-spacing:0.01em;line-height:1.85;margin:0px auto;max-width:900px;\">\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">在AI驱动的商业浪潮中，算力已成为企业最核心的生产力之一。然而，许多企业正面临一个共同的困境：斥巨资采购的GPU服务器，利用率却长期在低位徘徊，不同团队、不同项目间的算力资源如同孤岛，无法共享与高效流转。这不仅造成了巨大的资本浪费，更拖慢了AI应用的迭代速度。解决这一问题的关键，在于实施一套系统性的\u003Cstrong style=\"color:rgb(36, 38, 45);\">GPU资源池化\u003C\u002Fstrong>策略。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">本文将从一个资深基础架构师的视角，为您拆解从零到一构建高效GPU资源池的五个关键步骤。我们不会空谈概念，而是聚焦于可落地的技术选型、管理策略与优化手段，旨在帮助CTO、CIO、企业架构师以及平台工程团队，将分散、异构的算力资源转化为标准化、可运营的算力服务，从而\u003Cstrong style=\"color:rgb(36, 38, 45);\">显著提升AI计算效率\u003C\u002Fstrong>。\u003C\u002Fp>\u003Ch2 style=\"border-left:4px solid rgb(146, 79, 210);color:rgb(36, 38, 45);font-size:20px;line-height:1.5;margin:42px 0px 20px;padding-left:14px;\">\u003Cstrong>第一步：资源评估与规划——从“有多少”到“怎么用”\u003C\u002Fstrong>\u003C\u002Fh2>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">在开始任何技术部署之前，一次彻底的资源评估与规划是成功的基石。这一步的目标是摸清家底，明确需求，并制定出符合业务发展的资源蓝图。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">\u003Cstrong style=\"color:rgb(36, 38, 45);\">1. 存量盘点与异构性分析\u003C\u002Fstrong> 首先，你需要对现有的GPU资源进行一次全面盘点。这不仅仅是统计有多少张卡，更重要的是分析其异构性：\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">芯片架构与型号\u003C\u002Fstrong>：你的集群中是否混合了NVIDIA A100、H100、L40S，华为昇腾芯片？不同芯片在算力（TFLOPS）、显存（VRAM）、互联带宽（NVLink）上差异巨大。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">服务器配置\u003C\u002Fstrong>：CPU、内存、本地存储（NVMe SSD）与网络（InfiniBand\u002FRoCE）的配置是否均衡？是否存在因CPU或内存瓶颈导致GPU无法满载的情况？\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">地理位置与网络\u003C\u002Fstrong>：资源是集中在一个数据中心，还是分布在多个地域或机房？跨地域的网络延迟和带宽将成为资源池化的关键约束。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">\u003Cstrong style=\"color:rgb(36, 38, 45);\">2. 工作负载画像与需求预测\u003C\u002Fstrong> 接下来，深入分析你的AI工作负载。不同的任务对资源的需求模式截然不同：\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">训练任务\u003C\u002Fstrong>：通常是“计算密集型”和“数据密集型”的，需要长时间独占多卡（甚至多机）资源，对GPU算力、显存和节点间通信带宽极为敏感。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">推理任务\u003C\u002Fstrong>：通常是“吞吐量敏感型”或“延迟敏感型”的，可能要求高并发、低延迟，对资源的弹性伸缩能力要求更高。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">开发与调试\u003C\u002Fstrong>：属于“交互式”或“批处理式”任务，需要快速创建和销毁环境，对资源的敏捷交付要求高。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">基于历史数据和业务规划，预测未来6-12个月的算力需求峰值、谷值及增长趋势。这将直接决定你资源池的规模与弹性策略。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">\u003Cstrong style=\"color:rgb(36, 38, 45);\">3. 制定池化目标与KPI\u003C\u002Fstrong> 在评估的基础上，设定清晰的、可量化的池化目标。例如：\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">整体GPU利用率\u003C\u002Fstrong>：从当前的20%提升至60%以上。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">任务排队平均等待时间\u003C\u002Fstrong>：缩短50%。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">资源交付时间\u003C\u002Fstrong>：从以天\u002F周计缩短到分钟级。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">多租户资源隔离与成本分摊\u003C\u002Fstrong>：实现项目\u002F团队级别的精准计量。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">\u003Cstrong style=\"color:rgb(36, 38, 45);\">最佳实践建议一：建立“算力规格”目录\u003C\u002Fstrong> 不要直接暴露裸金属服务器。像\u003Ca style=\"color:rgb(146, 79, 210);text-underline-offset:3px;\" href=\"https:\u002F\u002Fwww.oneprocloud.com.cn\u002Fproducts\u002Fagione\">\u003Cstrong style=\"color:rgb(36, 38, 45);\">\u003Cu>AGIOne\u003C\u002Fu>\u003C\u002Fstrong>\u003C\u002Fa>\u003Cstrong style=\"color:rgb(36, 38, 45);\"> \u003C\u002Fstrong>\u003Ca style=\"color:rgb(146, 79, 210);text-underline-offset:3px;\" href=\"https:\u002F\u002Fwww.oneprocloud.com.cn\u002Fproducts\u002Fagione\">\u003Cstrong style=\"color:rgb(36, 38, 45);\">\u003Cu>PowerOne\u003C\u002Fu>\u003C\u002Fstrong>\u003C\u002Fa>\u003Cstrong style=\"color:rgb(36, 38, 45);\">（算模方）\u003C\u002Fstrong>这类平台所倡导的理念一样，根据芯片型号、卡数、显存大小、性能等级等维度，定义标准化的“算力规格”（如 \u003Ccode style=\"background-color:rgb(244, 241, 248);color:rgb(93, 43, 115);font-family:Consolas, SFMono-Regular, monospace;font-size:14px;padding:2px 5px;\">8卡-A100-80GB-训练型\u003C\u002Fcode>）。这能将底层异构复杂性向上层用户屏蔽，实现算力服务的标准化交付，是高效管理的第一步。\u003C\u002Fp>\u003Ch2 style=\"border-left:4px solid rgb(146, 79, 210);color:rgb(36, 38, 45);font-size:20px;line-height:1.5;margin:42px 0px 20px;padding-left:14px;\">\u003Cstrong>第二步：池化技术选型与部署——构建统一的管理平面\u003C\u002Fstrong>\u003C\u002Fh2>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">完成规划后，你需要选择一个核心平台或技术栈，作为整个资源池的“大脑”和“调度中心”。这个平台负责将物理的GPU资源抽象化、虚拟化，并按需分配给上层应用。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">\u003Cstrong style=\"color:rgb(36, 38, 45);\">1. 核心能力选型标准\u003C\u002Fstrong> 一个合格的异构智算纳管平台应具备以下核心能力：\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">异构统一纳管\u003C\u002Fstrong>：能够无缝接入并管理来自不同厂商、不同架构的智算芯片，而不仅限于某一种品牌。这是应对未来技术迭代和供应链风险的关键。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">以容器化为核心的调度\u003C\u002Fstrong>：现代AI工作负载的最佳载体是容器。平台应基于Kubernetes等容器编排系统，实现对GPU资源的细粒度调度（如指定卡型号、共享策略等），而非简单的虚拟机分配。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">算力服务化交付\u003C\u002Fstrong>：平台不应只是一个监控面板，而应能直接交付多种形态的算力服务实例，例如：为训练任务创建专属的容器实例、为模型部署推理服务环境、或快速提供集成了JupyterLab的AI开发环境。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">内置AI运行环境\u003C\u002Fstrong>：提供预置主流AI框架（PyTorch, TensorFlow等）和优化依赖库的标准化镜像，极大降低用户环境配置的复杂度，实现“开箱即用”。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">\u003Cstrong style=\"color:rgb(36, 38, 45);\">2. 部署模式考量\u003C\u002Fstrong> 根据企业自身的技术能力和运维体系，选择适合的部署模式：\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">私有化部署\u003C\u002Fstrong>：将平台软件部署在自有或托管的数据中心内，实现数据的完全自主可控。适用于对数据安全、合规性要求极高，且拥有专业运维团队的大型企业或智算中心。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">SaaS化订阅\u003C\u002Fstrong>：直接使用服务商提供的在线平台。这种方式免去了复杂的安装、运维和升级过程，可以快速启动资源池化管理。特别适合希望快速验证模式、或专注于AI应用开发而非底层基础设施运维的团队。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">\u003Cstrong style=\"color:rgb(36, 38, 45);\">最佳实践建议二：采用“物理服务器纳管”模式，而非虚拟化层抽象\u003C\u002Fstrong> 对于高性能AI计算，应优先考虑直接纳管物理GPU服务器，并通过容器进行隔离和调度。避免引入额外的虚拟化层（如vGPU），这通常会带来不可忽视的性能开销（通常为5%-15%），并增加管理复杂性。\u003Cstrong style=\"color:rgb(36, 38, 45);\">AGIOne PowerOne（算模方）\u003C\u002Fstrong>等平台即采用此模式，聚焦于将物理算力高效、无损地转化为容器化服务。\u003C\u002Fp>\u003Ch2 style=\"border-left:4px solid rgb(146, 79, 210);color:rgb(36, 38, 45);font-size:20px;line-height:1.5;margin:42px 0px 20px;padding-left:14px;\">\u003Cstrong>第三步：配额、队列与调度策略设置——实现公平与效率的平衡\u003C\u002Fstrong>\u003C\u002Fh2>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">资源池建好后，如果没有良好的治理策略，很快就会陷入“公地悲剧”——少数任务占用大量资源，其他任务无休止等待。因此，必须建立清晰的资源配额、队列和调度规则。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">\u003Cstrong style=\"color:rgb(36, 38, 45);\">1. 多级配额管理体系\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">组织\u002F项目级配额\u003C\u002Fstrong>：为每个部门或AI项目设置总的GPU资源上限（如每月10000卡时），从宏观上控制成本。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">用户\u002F团队级配额\u003C\u002Fstrong>：在项目内部，进一步为不同团队或个人设置并发资源使用上限（如最多同时使用4张A100），防止单个用户垄断资源。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">队列（Queue）管理\u003C\u002Fstrong>：创建不同优先级的队列（如 \u003Ccode style=\"background-color:rgb(244, 241, 248);color:rgb(93, 43, 115);font-family:Consolas, SFMono-Regular, monospace;font-size:14px;padding:2px 5px;\">high-priority\u003C\u002Fcode>， \u003Ccode style=\"background-color:rgb(244, 241, 248);color:rgb(93, 43, 115);font-family:Consolas, SFMono-Regular, monospace;font-size:14px;padding:2px 5px;\">normal\u003C\u002Fcode>， \u003Ccode style=\"background-color:rgb(244, 241, 248);color:rgb(93, 43, 115);font-family:Consolas, SFMono-Regular, monospace;font-size:14px;padding:2px 5px;\">low-priority\u003C\u002Fcode>）。紧急的线上推理任务或高管关注的模型训练可以提交至高优先级队列，确保其快速获得资源；而研发测试任务则可以进入低优先级队列，利用空闲资源运行。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">\u003Cstrong style=\"color:rgb(36, 38, 45);\">2. 智能调度策略\u003C\u002Fstrong> 调度器是资源池的“交通警察”，其策略直接影响效率：\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">优先级调度\u003C\u002Fstrong>：基于队列优先级和任务提交时间进行调度。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">抢占式调度\u003C\u002Fstrong>：允许高优先级任务抢占低优先级任务占用的资源（前提是平台支持任务的检查点保存与恢复），最大化关键任务的资源保障。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">亲和性\u002F反亲和性调度\u003C\u002Fstrong>：将需要频繁通信的多个任务（如分布式训练）调度到同一台服务器或高速互联的服务器组内；或将负载均衡的服务实例分散到不同故障域。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">基于资源规格的调度\u003C\u002Fstrong>：用户提交任务时指定所需的“算力规格”（如 \u003Ccode style=\"background-color:rgb(244, 241, 248);color:rgb(93, 43, 115);font-family:Consolas, SFMono-Regular, monospace;font-size:14px;padding:2px 5px;\">4卡-H100\u003C\u002Fcode>），调度器自动寻找匹配的节点，实现需求与供给的精准匹配。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">\u003Cstrong style=\"color:rgb(36, 38, 45);\">最佳实践建议三：实施“分层计量”与“内部结算”\u003C\u002Fstrong> 效仿云服务商的成熟做法，建立清晰的计量计费体系。平台应能按\u003Cstrong style=\"color:rgb(36, 38, 45);\">算力规格\u003C\u002Fstrong>、\u003Cstrong style=\"color:rgb(36, 38, 45);\">使用时长\u003C\u002Fstrong>等维度，自动统计每个用户、每个团队、每个项目的详细用量。这些数据不仅能用于成本分摊和“内部结算”，培养团队的资源成本意识，更是后续进行资源扩容、优化采购决策的核心依据。\u003C\u002Fp>\u003Ch2 style=\"border-left:4px solid rgb(146, 79, 210);color:rgb(36, 38, 45);font-size:20px;line-height:1.5;margin:42px 0px 20px;padding-left:14px;\">\u003Cstrong>第四步：全面监控、性能优化与自动化——从“可用”到“高效”\u003C\u002Fstrong>\u003C\u002Fh2>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">一个稳定的资源池只是及格线，我们追求的是高性能、高可用的资源池。这需要建立覆盖基础设施、平台、应用三层的立体监控体系，并基于数据驱动进行持续优化。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">\u003Cstrong style=\"color:rgb(36, 38, 45);\">1. 构建全方位监控仪表板\u003C\u002Fstrong> 监控指标应至少包括：\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">基础设施层\u003C\u002Fstrong>：GPU利用率、显存使用率、GPU温度、功耗；节点CPU\u002F内存\u002F磁盘IO\u002F网络带宽使用率。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">平台调度层\u003C\u002Fstrong>：集群总资源容量\u002F已分配\u002F剩余量；各队列排队任务数、平均等待时间；调度成功率\u002F失败率。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">任务应用层\u003C\u002Fstrong>：单个训练任务的迭代速度（iterations\u002Fsec）、损失曲线；推理服务的请求延迟（P99 Latency）、吞吐量（QPS）、错误率。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">\u003Cstrong style=\"color:rgb(36, 38, 45);\">2. 性能优化关键点\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">消除瓶颈\u003C\u002Fstrong>：通过监控数据，识别常见瓶颈。例如，发现GPU利用率低但CPU使用率100%，可能是数据预处理环节成为瓶颈，需要考虑优化数据加载或使用GPU加速的数据加载库。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">存储与网络优化\u003C\u002Fstrong>：确保训练数据存储在高速分布式存储（如CephFS， Lustre）或每个节点的本地NVMe SSD上。对于多机训练，必须使用高性能网络（如InfiniBand）并启用NCCL的优化配置。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">模型与算力适配\u003C\u002Fstrong>：这是提升效率的“黑科技”。优秀的平台会提供预置的 \u003Cstrong style=\"color:rgb(36, 38, 45);\">“模型×算力硬件”部署模板\u003C\u002Fstrong>。例如，针对某款特定型号的GPU和某个流行的视觉大模型，平台已预先完成了深度学习框架版本、算子库、并行策略等的最佳实践调优。用户直接使用该模板部署，即可获得稳定且接近硬件峰值的性能，避免了繁琐的调优过程。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">\u003Cstrong style=\"color:rgb(36, 38, 45);\">最佳实践建议四：建立“健康度巡检”与“自动化修复”机制\u003C\u002Fstrong> 制定GPU节点的健康度标准（如温度阈值、ECC错误数）。通过监控系统定期巡检，对异常节点自动标记并隔离，防止有故障的节点影响任务运行。同时，可以设置自动化脚本，对常见的软件层问题（如驱动异常、容器运行时僵死）进行重启修复，提升集群的整体可用性。\u003C\u002Fp>\u003Ch2 style=\"border-left:4px solid rgb(146, 79, 210);color:rgb(36, 38, 45);font-size:20px;line-height:1.5;margin:42px 0px 20px;padding-left:14px;\">\u003Cstrong>第五步：持续改进、弹性扩展与生态集成——面向未来的资源池\u003C\u002Fstrong>\u003C\u002Fh2>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">技术和业务都在不断演进，资源池也必须是一个能够持续迭代和扩展的有机体。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">\u003Cstrong style=\"color:rgb(36, 38, 45);\">1. 基于数据的持续改进\u003C\u002Fstrong> 定期（如每季度）回顾在第一步中设定的KPI。分析目标未达成的原因：是调度策略问题？还是资源规格定义不合理？或是某些类型的工作负载暴增？利用计量和监控数据，驱动资源配额调整、队列策略优化，甚至硬件采购计划的修订。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">\u003Cstrong style=\"color:rgb(36, 38, 45);\">2. 弹性扩展能力\u003C\u002Fstrong> 资源池应能平滑地接纳新购的服务器，无论是同构扩容还是引入新一代的异构芯片。平台应支持新资源节点的快速注册、纳管，并自动纳入调度范围。对于存在明显波峰波谷的业务，可以考虑与公有云GPU服务集成，在私有资源池满载时，将低优先级任务或突发任务“溢出”到公有云，实现混合云的弹性伸缩。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">\u003Cstrong style=\"color:rgb(36, 38, 45);\">3. 向上层模型服务闭环\u003C\u002Fstrong> 资源池化的终极价值不仅是管理GPU，更是为了高效地运行AI模型。因此，平台需要具备与上层模型服务生态集成的能力。例如，平台在完成一个模型的推理服务部署后，应能将其\u003Cstrong style=\"color:rgb(36, 38, 45);\">发布为标准化的API服务\u003C\u002Fstrong>，并自动生成调用端点。这个端点可以被上层的模型服务平台（MaaS）、API网关或业务应用直接调用，从而形成从\u003Cstrong style=\"color:rgb(36, 38, 45);\">算力纳管 -> 模型部署 -> 服务发布 -> 业务调用\u003C\u002Fstrong>的完整闭环，真正打通AI生产力链条。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">\u003Cstrong style=\"color:rgb(36, 38, 45);\">最佳实践建议五：规划“算力运营”路线图\u003C\u002Fstrong> 将GPU资源池视为一个内部的产品进行“运营”。除了技术功能，还要考虑：\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">用户门户与体验\u003C\u002Fstrong>：为数据科学家提供简洁易用的自助服务门户，让他们能一键申请资源、部署环境。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">文档与知识库\u003C\u002Fstrong>：建立完善的内部文档，包括平台使用指南、最佳实践案例、故障排查手册。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">· \u003Cstrong style=\"color:rgb(36, 38, 45);\">社区与反馈\u003C\u002Fstrong>：建立用户反馈渠道，定期收集需求，让资源池的演进始终贴近实际业务。\u003C\u002Fp>\u003Ch2 class=\"special-title\" style=\"border-left:0px;color:rgb(146, 79, 210);font-size:20px;font-weight:400;line-height:1.5;margin:42px 0px 20px;padding-left:0px;\">结语\u003C\u002Fh2>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">GPU资源池化并非一蹴而就的简单项目，而是一项需要精心设计、分步实施的系统工程。通过上述\u003Cstrong style=\"color:rgb(36, 38, 45);\">资源评估、技术选型、配额治理、监控优化、持续演进\u003C\u002Fstrong>五个步骤，企业可以系统地构建一个高效、公平、可运营的异构智算资源池。\u003C\u002Fp>\u003Cp style=\"font-size:16px;line-height:1.85;margin:0px 0px 18px;\">其核心价值在于，将昂贵的GPU硬件从静态、专属的固定资产，转变为动态、共享的关键生产要素。这不仅直接\u003Cstrong style=\"color:rgb(36, 38, 45);\">提升了AI计算效率\u003C\u002Fstrong>，降低了单位计算成本，更通过标准化的服务交付和敏捷的资源供给，加速了AI从实验到生产的全过程，为企业在AI时代的竞争构筑了坚实的算力基石。\u003C\u002Fp>\u003C\u002Farticle>\u003Cdiv style=\"-webkit-text-stroke-width:0px;color:rgb(0, 0, 0);font-family:"Noto Sans SC";font-size:medium;font-style:normal;font-variant-caps:normal;font-variant-ligatures:normal;font-weight:400;inset:0px;letter-spacing:normal;orphans:2;pointer-events:none;position:fixed;text-align:start;text-decoration-color:initial;text-decoration-style:initial;text-decoration-thickness:initial;text-indent:0px;text-transform:none;white-space:normal;widows:2;word-spacing:0px;z-index:2147483647;\" id=\"codex-browser-sidebar-comments-root\">\u003Cdiv> \u003C\u002Fdiv>\u003C\u002Fdiv>\u003Cp>\u003Cbr> \u003C\u002Fp>","html","本文从资源评估、平台选型、配额调度、监控优化和持续演进五个阶段，梳理企业建设高效、共享、可运营 GPU 资源池的关键方法。","\u002Ffile\u002F2\u002F2026\u002F9\u002F7316c13994aa.png?i=9263&name=GPU池化.png","gpu_resource_pooling_best_practices","GPU资源池化最佳实践：提升企业AI计算效率的5个步骤","GPU资源池化,异构智算,算力调度,AI基础设施,AGIOne,PowerOne","GPU资源池化最佳实践：提升AI计算效率的5个步骤","GPU资源池化,AI计算效率,异构智算纳管,算力调度,GPU资源管理,AI基础设施,容器化GPU,AGIOne,PowerOne","从资源评估、技术选型、配额调度、监控优化和持续演进五个步骤，梳理企业 GPU 资源池化与异构智算管理的落地方法。","2026-09-28","2026-09-28 10:39:29","AGIOne","\u002Ffile\u002F2\u002F2026\u002F6\u002F291951eaf479.png?i=9162&name=AGIOne_反白.png","行业科普",1790563555078,{"data":262,"status":23,"timestamp":269},{"previous":263,"next":10},{"id":264,"title":265,"cover":266,"slug":267,"productName":257,"blogTypeName":259,"blogTypeCode":268},224,"模型服务乱局：为何通用平台不如专业路由？AGIOne ModelOne 揭秘企业多模型治理之道","\u002Ffile\u002F2\u002F2026\u002F9\u002F42f1df6a573a.png?i=9262&name=揭秘企业多模型治理.png","enterprise_multi_model_governance_routing","popular-science",1790563555120,{"data":271,"status":23,"timestamp":366},[272,289,306,324,339,353],{"tenantId":5,"id":273,"title":274,"desc":275,"intro":276,"industryId":277,"slug":278,"productId":53,"languageId":6,"tags":279,"siteId":6,"cover":280,"coverBig":281,"seoTitle":282,"keywords":283,"description":284,"pdfurl":285,"isTop":22,"status":6,"publishedAt":286,"contentUpdatedAt":286,"sortIndex":10,"platform":10,"createdAt":287,"updatedAt":286,"languageName":16,"productName":257,"productIcon":258,"productCode":257,"industryName":288},59,"墨西哥医疗机构：AGIOne统一AI效能层驱动GPU利用率增长","墨西哥一家国家级医疗机构面临GPU资源碎片化、利用率不足20%的困境。通过部署AGIOne统一AI效能层，实现异构算力池化、多模型统一服务与混合云弹性调度。GPU利用率实现翻倍以上增长，新AI服务部署周期从\"月\"压缩到\"天\"，成功将AI从成本项目转化为支撑核心诊断业务的战略能力。","背景：该墨西哥医疗机构率先部署多个AI试点项目用于影像诊断、放射分析及患者服务，但各应用独占硬件导致GPU利用率低于20%，扩容依赖采购、CapEx持续飙升，AI规模化陷入瓶颈。\n\n方案：采用AGIOne构建统一AI效能层，实现四大突破：GPU资源池化整合碎片化算力、异构纳管跨型号GPU统一调度、多模型聚合实现LLM统一抽象服务、混合云弹性架构本地资源触顶时自动扩展至云端。同步引入全生命周期运营方法论，基于延迟、成本、数据敏感度三维智能路由自动决策工作负载去向。\n\n收益：GPU利用率从不足20%实现翻倍以上增长，硬件投资回报率显著改善；新AI服务上线及运营变更周期从\"月\"压缩至\"天\"；混合云弹性架构使扩容不再受制于硬件采购，业务全程不中断；集中化控制台强化数据治理与合规管控，AI从成本负担转变为支撑核心诊断业务的战略护城河。\n\n",5,"mexico_healthcare_gpu_utilization_ai","异构GPU资源池化$$GPU利用率翻倍增长,混合云弹性调度$$本地云端无缝扩展,多模型统一治理$$异构LLM聚合智能路由","\u002Ffile\u002F2\u002F2026\u002F6\u002F9b010d5fc872.png?i=7884&name=预览图.png","\u002Ffile\u002F2\u002F2026\u002F6\u002F236f21262dc6.png?i=7885&name=官网Bannner.png","","医疗AI，AI效能层，GPU资源调度，模型聚合平台，混合云AI，智能路由，AGIOne，异构算力纳管，AI规模化运营，医疗影像诊断，AI生命周期管理，企业AI基础设施，智算资源调度，AI运营平台，算力池化","墨西哥国家级医疗机构在推进多个AI试点项目后，遭遇GPU资源碎片化、利用率低于20%、扩容严重依赖硬件采购等困局。该机构采用AGIOne统一AI效能层解决方案，通过资源池化、异构纳管、模型聚合与混合云弹性四大能力，构建全新的AI运营基础设施。项目落地后GPU利用率实现翻倍以上增长，新AI服务部署周期从\"月\"压缩到\"天\"，为医疗行业AI规模化运营提供了标杆实践。","\u002Ffile\u002Fcase\u002F59\u002F墨西哥医疗机构：AGIOne统一AI效能层驱动GPU利用率增长.pdf","2026-09-09 17:22:19","2026-06-16 12:15:15","卫生医疗",{"tenantId":5,"id":290,"title":291,"desc":292,"intro":293,"industryId":6,"slug":294,"productId":50,"languageId":6,"tags":295,"siteId":6,"cover":296,"coverBig":297,"seoTitle":10,"keywords":298,"description":299,"pdfurl":300,"isTop":22,"status":6,"publishedAt":10,"contentUpdatedAt":10,"sortIndex":10,"platform":6,"createdAt":301,"updatedAt":302,"languageName":16,"productName":43,"productIcon":303,"productCode":304,"industryName":305},8,"江苏农信：跨地域，多网点，大数据量迁移上金融云","江苏省农村信用社联合社作为我国发达地区的农村金融机构，通过打造行业云平台，联结辖内省联社机构，重塑省联社价值。\n\n","挑战：数据中心亟需上云，缺少成熟工具保障迁移安全与业务连续性，传统方式风险高、周期长。\n方案：采用HyperMotion迁移工具，实现对银行VMware集群的全量无中断迁移至云环境。\n收益：完成多业务系统稳定上云，迁移过程零中断，节省70%人工投入，保障业务高可用。\n","hypermotion-case-studies-migration-jiangsu-rural-bank","跨地域$$多地数据中心统一上云,多网点$$56个地市网点,大数据量$$1156TB","\u002Ffile\u002F2\u002F2025\u002F3\u002F099eb15481a7.jpg?i=8592&name=los-angeles-498285_1280.jpg","\u002Ffile\u002F2\u002F2025\u002F4\u002F97e353d2048d.png?i=8659&name=江苏农信.png","江苏农信,农商行迁移,数据迁移,金融数据迁移,无代理迁移,热迁移,业务不中断迁移,迁移成本","HyperMotion助力江苏省农村信用社联合社在短时间内成功将56家农商行的1285套系统、1156TB数据迁移至金融云平台，采用无代理热迁移和MotionGear设备，确保业务零中断，节省70%人力成本，荣获银保监会科技风险管理成果奖。","\u002Ffile\u002Fcase\u002F8\u002F江苏农信：跨地域，多网点，大数据量迁移上金融云.pdf","2021-12-13 14:59:44","2026-08-02 15:21:09","\u002Ffile\u002F2\u002F2025\u002F4\u002Fe4aee932c781.png?i=8652&name=云迁移横版文字反白.png","hypermotion","金融",{"tenantId":5,"id":307,"title":308,"desc":309,"intro":310,"industryId":311,"slug":312,"productId":160,"languageId":6,"tags":313,"siteId":6,"cover":314,"coverBig":315,"seoTitle":10,"keywords":316,"description":317,"pdfurl":318,"isTop":22,"status":6,"publishedAt":10,"contentUpdatedAt":10,"sortIndex":10,"platform":10,"createdAt":319,"updatedAt":320,"languageName":16,"productName":51,"productIcon":321,"productCode":322,"industryName":323},44,"某城开集团：用云容灾成功抵御病毒勒索攻击","国内某城开企业业务系统运行于华为 HCS8.1 私有云，选用 HyperBDR 容灾工具，采用云容灾架构实现差异化容灾策略，提升业务韧性。项目实施后，企业业务连续性提升，成功应对一次病毒勒索事件。","背景：某农担金融企业 8 套业务系统运行于本地华为 HCS8.1 私有云，原备份机制有缺陷，传统容灾成本高，亟需优化。\n方案：选用万博智云 HyperBDR，采用华为公有云对象存储，依业务定容灾策略，分阶段交付测试。\n收益：业务连续性提升，成本降低，管理更高效，面对勒索病毒等风险时响应快速，恢复能力强。",22,"chengkai-group-successfully-resists--ransomware-attacks-with-cloud-disaster-recovery","混合云容灾$$HCS到华为云,定制同步策略$$结合业务定制容灾SLA,弹性架构$$容灾资源按量计费","\u002Ffile\u002F2\u002F2025\u002F4\u002Fe07eacec88c6.png?i=8747&name=城开集团.png","\u002Ffile\u002F2\u002F2025\u002F4\u002F0550fc6e1110.png?i=8665&name=地产.png","云容灾,混合云容灾,公有云容灾,对象存储容灾,抵御勒索病毒,业务连续性保护,HyperBDR","国内某城开企业 8 套业务系统运行于本地 IDC 华为 HCS8.1 私有云，原备份方式缺陷明显。企业与万博智云合作，选用 HyperBDR 容灾工具，采用云容灾架构。该方案实现差异化容灾策略，优化资源配置。项目实施后，企业业务连续性提升，成本降低，管理效率提高，应对风险能力增强。","\u002Ffile\u002Fcase\u002F44\u002F某城开集团：用云容灾成功抵御病毒勒索攻击.pdf","2025-04-29 17:31:13","2026-08-03 12:09:05","\u002Ffile\u002F2\u002F2025\u002F4\u002F6f0a8978d60a.png?i=8654&name=云容灾横版文字反白.png","hyperbdr","地产",{"tenantId":5,"id":325,"title":326,"desc":327,"intro":328,"industryId":42,"slug":329,"productId":160,"languageId":6,"tags":330,"siteId":6,"cover":331,"coverBig":332,"seoTitle":10,"keywords":333,"description":334,"pdfurl":335,"isTop":22,"status":6,"publishedAt":10,"contentUpdatedAt":10,"sortIndex":10,"platform":10,"createdAt":336,"updatedAt":337,"languageName":16,"productName":51,"productIcon":321,"productCode":322,"industryName":338},52,"某头部制造企业：轻松构建自动化、弹性扩展的跨平台容灾","广东某电气制造企业基于HyperBDR与国产超融合平台构建混合异构的跨平台容灾方案，1个月完成部署、同步与演练验证，分钟级恢复、一键演练、兼容原系统与国产平台，整体成本节省90%，为制造业容灾建设提供高性能、可扩展的实用参考范式。","背景：某电气制造企业正加速智能工厂建设，需保障核心业务系统在复杂混合架构下实现跨地域业务连续，原有Hyper-V平台仍承担重要业务，容灾建设需兼容原系统并支持国产替代演进。\n\n方案：企业选用 HyperBDR与国产超融合平台组成一体化联合容灾方案，实现数据统一备份、自动化编排恢复与一键演练，1个月内成功完成跨城市容灾部署及验证。\n\n收益：容灾平台支持分钟级恢复，演练操作简便，运维效率提升60%+，无须新增硬件资源，整体成本节省超90%，真正实现从“能部署”到“敢用、会用”。\n\n","leading-electrical-enterprise-builds-scalable-high-performance-hybrid-dr","混合异构容灾$$轻松实现,分钟级恢复$$高性能,弹性资源调度$$可扩展","\u002Ffile\u002F2\u002F2025\u002F7\u002Fa4684cbd4dfa.png?i=8836&name=电气制造案例图片.png","\u002Ffile\u002F2\u002F2025\u002F4\u002F5a4062831c0d.png?i=8676&name=BDR制造.png","制造业容灾,混合架构容灾,Hyper-V容灾,超融合容灾,异地容灾上线,HyperBDR,自动化容灾,一键演练,业务连续性保障","广东某电气制造企业采用HyperBDR联合超融合平台，构建异地混合架构容灾体系，1个月完成部署与演练验证，兼容原系统并支撑国产信创平台演进，分钟级恢复与自动化演练全面提升容灾实用性与运维效率，优化成本。","\u002Ffile\u002Fcase\u002F52\u002F某头部制造企业：轻松构建自动化、弹性扩展的跨平台容灾.pdf","2025-07-10 10:57:59","2026-08-03 08:33:13","制造",{"tenantId":5,"id":185,"title":340,"desc":341,"intro":342,"industryId":50,"slug":343,"productId":50,"languageId":6,"tags":344,"siteId":6,"cover":345,"coverBig":346,"seoTitle":10,"keywords":347,"description":348,"pdfurl":349,"isTop":22,"status":6,"publishedAt":10,"contentUpdatedAt":10,"sortIndex":10,"platform":6,"createdAt":350,"updatedAt":351,"languageName":16,"productName":43,"productIcon":303,"productCode":304,"industryName":352},"北明数科：交付团队采用HyperMotion加速政务云迁移上云进程","某市政府想要进一步提升政务系统的信息化水平，将下辖各单位的资源和数据统一集中管理，计划将各单位的业务数据和系统迁移上云，同时完善等级保护服务体系。","挑战：某市政府老机房即将到期，仅有 15 天迁移时间；源端环境复杂，涉及多种平台；迁移后需更改 IP 并演练；业务关联性复杂，各单位迁移时间难协调。\n方案：采用整机热迁移和块级别差量复制；利用 API 智能适配实现高度自动化迁移；进行灾备演练式渐进迁移；提供整体解决方案。\n收益：15 天内完成近百台主机、56T 数据迁移，获客户高度肯定，降低迁移成本与运维压力。","hypermotion-case-studies-migration-government-beimin-shuke","混合云迁移$$物理机\u002FKVM\u002FXEN虚拟化,政务系统上云$$多部门多平台环境复杂,自动化迁移$$支持T-stack自动调度编排","\u002Ffile\u002F2\u002F2024\u002F11\u002Fb44b65b5f50e.jpg?i=87&name=移动通信.jpg","\u002Ffile\u002F2\u002F2025\u002F4\u002F9bc27e3628a1.png?i=8670&name=政府.png","HyperMotion云迁移,政务云,腾讯云T-stack,KVM迁移,虚拟机迁移,Xen迁移","北明数科采用HyperMotion在15天内完成某市政府91套系统、56TB数据的无中断上云迁移，支持多平台环境和IP变更演练，自动化程度高，显著降低运维压力，助力政务系统高效集约化部署。","\u002Ffile\u002Fcase\u002F16\u002F北明数科：交付团队采用HyperMotion加速政务上云进程.pdf","2022-04-13 09:52:16","2026-08-03 09:42:00","政府",{"tenantId":5,"id":94,"title":354,"desc":355,"intro":356,"industryId":6,"slug":357,"productId":50,"languageId":6,"tags":358,"siteId":6,"cover":359,"coverBig":360,"seoTitle":10,"keywords":361,"description":362,"pdfurl":363,"isTop":22,"status":6,"publishedAt":10,"contentUpdatedAt":10,"sortIndex":10,"platform":10,"createdAt":364,"updatedAt":365,"languageName":16,"productName":43,"productIcon":303,"productCode":304,"industryName":305},"某农信：VMware替代，迁移万台虚机，Apasara Stack成核心承载平台","为加快信创建设与IT架构统一，某农信启动全省范围的云迁移工程，计划2年内完成37个分行、约一万台虚拟机的上云任务。项目引入高自动化迁移工具HyperMotion，通过总行统筹与基层IT赋能相结合的方式，分批次、有节奏地推进大规模异构环境迁移，显著提升效率与自主运维能力。","背景：某农信统一部署Apasara Stack等云平台，计划分步将各地分行的物理机、VMware虚拟机系统迁移上云，但面临环境差异大、时间紧迁移任务重、业务不中断等挑战。\n\n方案：由总行技术团队统一培训和平台部署，将迁移能力快速下沉至各行社IT团队。各分行采用HyperMotion自主推进迁移任务，迁移支持小组提供全程远程协助和技术响应，确保迁移顺畅高效。\n\n收益： 多个地市已完成设备批量迁移，整体效率提升超3倍；基层IT人员掌握操作流程，降低外部依赖；统一云平台纳管提高后续运维能力，全省迁移工程按“两年万台”目标有序推进。","rural-credit-apasara-stack-platform-hosts-migration-and-replacement-of-10000-vmware-vim","10000台虚机$$迁移规模庞大,85个行社$$全省各地跨架构迁移,3倍效率提升$$批量迁移、流程自动化","\u002Ffile\u002F2\u002F2025\u002F7\u002F8490a84e18bd.png?i=8820&name=浙江农信封面图.png","\u002Ffile\u002F2\u002F2025\u002F7\u002Fbf3ac7737da6.png?i=8821&name=浙江农信-案例下载页图.png","农信上云,VMware替代方案,跨平台虚机迁移,阿里Apasara迁移,阿里私有云,省级金融迁移,信创云迁移工具,金融行业云迁移,自主迁移工具,自动化迁移工具,混合云IT架构,HyperMotion案例","农信全面推进信创建设，依托Apasara Stack平台替代VMware环境，计划2年内完成全省超过30家分行、约1万台虚拟机的自主迁移。采用HyperMotion工具，IT团队分批推进，高效实现跨平台云化，构建统一、安全、可控的金融云基础架构。","\u002Ffile\u002Fcase\u002F50\u002F浙江农信：VMware替代，迁移万台虚机，Apasara Stack成核心承载平台.pdf","2025-07-01 16:24:40","2026-07-31 16:34:23",1790563555159,{"data":368,"status":23,"timestamp":463},[369,377,385,395,403,413,421,430,438,445,451,457],{"id":370,"title":371,"cover":372,"desc":373,"siteId":6,"slug":374,"typeId":244,"productId":53,"languageId":6,"status":6,"date":375,"createdAt":376,"productName":257,"productCode":257,"productIcon":258,"blogTypeName":259,"blogTypeCode":268,"languageName":16},209,"Hybrid AI 是什么？公有模型、私有模型与本地算力如何协同","\u002Ffile\u002F2\u002F2026\u002F8\u002F3abea5c50832.png?i=9225&name=what-is-hybrid-ai-public-private-models-local-compute.png","解释 Hybrid AI 的定义，以及公有模型、私有模型、本地与云上算力如何通过服务目录、身份授权、部署、计量和运营规则协同。","what-is-hybrid-ai-public-private-models-local-compute","2026-08-25","2026-08-25 09:33:46",{"id":378,"title":379,"cover":380,"desc":381,"siteId":6,"slug":382,"typeId":244,"productId":53,"languageId":6,"status":6,"date":383,"createdAt":384,"productName":257,"productCode":257,"productIcon":258,"blogTypeName":259,"blogTypeCode":268,"languageName":16},177,"企业CIO必读：2026年模型聚合服务五大趋势与AGIONE解决方案","\u002Ffile\u002F2\u002F2026\u002F4\u002Ff31a70d35b9d.png?i=9149&name=配图.png","对于任何一位企业CIO、CTO或技术决策者而言，2026年的AI应用格局已不再是“是否使用大模型”，而是“如何高效、经济、安全地驾驭一个由数十甚至上百个模型构成的复杂生态系统”。","model_aggregation_service_trends_2026","2026-04-23","2026-04-23 11:15:20",{"id":386,"title":387,"cover":388,"desc":389,"siteId":6,"slug":390,"typeId":391,"productId":53,"languageId":6,"status":6,"date":392,"createdAt":393,"productName":257,"productCode":257,"productIcon":258,"blogTypeName":100,"blogTypeCode":394,"languageName":16},176,"连Token 都有了中文名，企业 CIO 却还是最后一个知道公司在AI上花了多少钱？","\u002Ffile\u002F2\u002F2026\u002F3\u002Fbaf17694f843.jpg?i=9146&name=未标题-1.jpg","企业 AI 规模化应用下，影子 AI 扩散、成本失控、技术债高筑等问题凸显，单一模型时代已落幕。ModelOne 企业级多模型管理平台，以统一 API、全链路成本观测、智能调度等能力，让 AI 从粗放走向可控，助力企业低成本、安全地实现 AI 规模化落地。","even-token-have-chinese-name-enterprise-cio-last-know-ai-spending",21,"2026-03-27","2026-03-26 17:30:05","solutions",{"id":396,"title":397,"cover":398,"desc":399,"siteId":6,"slug":400,"typeId":391,"productId":53,"languageId":6,"status":6,"date":401,"createdAt":402,"productName":257,"productCode":257,"productIcon":258,"blogTypeName":100,"blogTypeCode":394,"languageName":16},215,"政务AI平台建设：私有化、审计与多租户治理","\u002Ffile\u002F2\u002F2026\u002F8\u002Fb65686bb4fa1.png?i=9235&name=government-ai-platform-private-audit-multitenancy.png","政务AI平台需要统一服务目录、分级账号权限、私有化部署、审计、资源配额与计量。本文提供多租户治理结构和分阶段建设路径。","government-ai-platform-private-audit-multitenancy","2026-08-26","2026-08-26 18:05:00",{"id":404,"title":405,"cover":406,"desc":407,"siteId":6,"slug":408,"typeId":290,"productId":53,"languageId":6,"status":6,"date":409,"createdAt":410,"productName":257,"productCode":257,"productIcon":258,"blogTypeName":411,"blogTypeCode":412,"languageName":16},178,"华为联合万博智云发布AGIOne AI基础设施编排联合方案，加速金融行业迈向智能体时代","\u002Ffile\u002F2\u002F2026\u002F5\u002F94b5e399bc94.png?i=9152&name=封面图-官网.png","华为携手万博智云在HiFS 2026峰会发布基于AGIOne的AI基础设施编排联合方案，助力金融机构构建智能体银行底座。","agione_ai_infrastructure_orchestration_financial_industry","2026-05-22","2026-05-22 14:25:56","新闻","news",{"id":414,"title":415,"cover":416,"desc":417,"siteId":6,"slug":418,"typeId":244,"productId":53,"languageId":6,"status":6,"date":419,"createdAt":420,"productName":257,"productCode":257,"productIcon":258,"blogTypeName":259,"blogTypeCode":268,"languageName":16},188,"AI Agent 从 Demo 到业务流程：企业为什么需要统一运行底座","\u002Ffile\u002F2\u002F2026\u002F8\u002F7d684b5ba2b7.png?i=9213&name=封面图.png","本文分析企业 AI Agent 从 Demo 进入业务流程时面临的模型服务、权限、算力、计量和运营问题，并说明统一运行底座的作用、建设路径与能力边界。","ai_agent_runtime_foundation","2026-08-20","2026-08-20 17:56:23",{"id":422,"title":423,"cover":424,"desc":425,"siteId":6,"slug":426,"typeId":427,"productId":53,"languageId":6,"status":6,"date":375,"createdAt":376,"productName":257,"productCode":257,"productIcon":258,"blogTypeName":428,"blogTypeCode":429,"languageName":16},201,"GPU与NPU如何统一组织？多芯片资源池规划方法","\u002Ffile\u002F2\u002F2026\u002F8\u002F4a7a87573ba3.png?i=9220&name=gpu-npu-unified-multi-chip-resource-pool.png","GPU与NPU统一组织需要分层管理资产、资源池、规格、环境、调度、配额和计量，并保留芯片架构差异。本文提供规划方法与验收清单。","gpu-npu-unified-multi-chip-resource-pool",24,"最佳实践","best-practice",{"id":431,"title":432,"cover":433,"desc":434,"siteId":6,"slug":435,"typeId":244,"productId":53,"languageId":6,"status":6,"date":436,"createdAt":437,"productName":257,"productCode":257,"productIcon":258,"blogTypeName":259,"blogTypeCode":268,"languageName":16},175,"2026 年模型聚合服务趋势：从多模型部署到智能治理","\u002Ffile\u002F2\u002F2026\u002F3\u002Fe0ebbe8cec10.png?i=9144&name=111.png","2023 年生成式 AI 爆发推动企业多模型部署普及，却引发接入碎片化、运维复杂、成本失控等问题，催生出模型聚合服务平台。该平台以统一 API 网关为基础，通过策略化路由、模型可观测性实现多模型智能调度与治理，还从计量、路由优化、模算联动层面实现 Token 成本精细化管控。其正从基础聚合向策略智能化、治理即代码、融合 AIOps 演进，未来将成为连接多方的模型生态操作系统。企业落地需明确层级边界、渐进式实施策略，做好成本与可观测性管理，这一从部署到治理的升级，是企业实现 AI 能力高效转化的关","2026-model-aggregation-service-trend-from-multi-model-deployment-to-intelligent-governance","2026-03-20","2026-03-20 16:01:42",{"id":439,"title":440,"cover":441,"desc":442,"siteId":6,"slug":443,"typeId":427,"productId":53,"languageId":6,"status":6,"date":375,"createdAt":444,"productName":257,"productCode":257,"productIcon":258,"blogTypeName":428,"blogTypeCode":429,"languageName":16},197,"企业 AI 平台如何选型？自建、拼装还是统一平台","\u002Ffile\u002F2\u002F2026\u002F9\u002F3966164996d1.png?i=9259&name=企业AI平台选型指南.png","比较企业 AI 平台自建、工具拼装和统一平台三种路线，从适配度、维护、治理、集成和扩展成本建立选型矩阵与 PoC 验证清单。","enterprise-ai-platform-selection-guide","2026-08-25 09:32:03",{"id":446,"title":447,"cover":448,"desc":449,"siteId":6,"slug":450,"typeId":427,"productId":53,"languageId":6,"status":6,"date":375,"createdAt":376,"productName":257,"productCode":257,"productIcon":258,"blogTypeName":428,"blogTypeCode":429,"languageName":16},199,"企业模型服务目录怎么建设？分类、发布、权限与生命周期","\u002Ffile\u002F2\u002F2026\u002F8\u002F5f2cd05189c9.png?i=9245&name=enterprise-model-service-catalog.png","模型服务目录不只是模型清单。本文提供对象划分、分类方法、服务卡字段、发布审核、权限、计量和下线流程，帮助企业建设可运营的模型服务入口。","enterprise-model-service-catalog",{"id":452,"title":453,"cover":454,"desc":455,"siteId":6,"slug":456,"typeId":427,"productId":53,"languageId":6,"status":6,"date":375,"createdAt":376,"productName":257,"productCode":257,"productIcon":258,"blogTypeName":428,"blogTypeCode":429,"languageName":16},206,"模型服务生命周期管理：从接入、评测到下线","\u002Ffile\u002F2\u002F2026\u002F8\u002Fdd24300f8bf8.png?i=9218&name=model-service-lifecycle-management.png","模型服务生命周期覆盖接入、评测、部署、发布、授权、计量、变更、弃用与下线。本文提供八阶段框架、必备记录和成熟度检查表。","model-service-lifecycle-management",{"id":458,"title":459,"cover":460,"desc":461,"siteId":6,"slug":462,"typeId":427,"productId":53,"languageId":6,"status":6,"date":375,"createdAt":444,"productName":257,"productCode":257,"productIcon":258,"blogTypeName":428,"blogTypeCode":429,"languageName":16},189,"AI 成本分摊怎么做？按部门、项目、租户和模型归集","\u002Ffile\u002F2\u002F2026\u002F8\u002Ffa931a3dfd30.png?i=9229&name=ai-cost-allocation-by-department-project-tenant-model.png","说明 AI 模型调用与算力成本如何按部门、项目、租户、应用和模型归集，并提供直接成本、共享成本、Showback 和 Chargeback 实施步骤。","ai-cost-allocation-by-department-project-tenant-model",1790563555199,1790563554882]