[{"data":1,"prerenderedAt":234},["ShallowReactive",2],{"content-doc-\u002Fzh\u002Fusercase\u002Fimplementing-highly-reliable-system-disk-management":3,"surround-\u002Fzh\u002Fusercase\u002Fimplementing-highly-reliable-system-disk-management":232},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"company":10,"date":11,"logo":12,"logoDark":13,"body":14,"_type":224,"_id":225,"_source":226,"_file":227,"_stem":228,"_extension":229,"coverImage":90,"plainText":230,"authorNames":231},"\u002Fzh\u002Fusercase\u002Fimplementing-highly-reliable-system-disk-management","usercase",false,"","基于 openUBMC 实现服务器系统盘高可靠管理","在 AI 训练、HCI 超融合和分布式存储场景中，系统盘承担操作系统启动、配置与日志记录等核心能力，其可靠性直接关系服务器节点的可用性。传统“一卡两用”方案由同一张 RAID 卡管理系统盘和数据盘，存在以下问题：","百信信息","2026\u002F10\u002F23","\u002Fcategory\u002Fusercase\u002Fimplementing-highly-reliable-system-disk-management\u002Flogo.png","\u002Fcategory\u002Fusercase\u002Fimplementing-highly-reliable-system-disk-management\u002Flogo-dark.png",{"type":15,"children":16,"toc":213},"root",[17,26,31,67,72,78,83,93,100,105,111,116,121,130,136,141,174,180],{"type":18,"tag":19,"props":20,"children":22},"element","h3",{"id":21},"一应用背景",[23],{"type":24,"value":25},"text","一、应用背景",{"type":18,"tag":27,"props":28,"children":29},"p",{},[30],{"type":24,"value":9},{"type":18,"tag":32,"props":33,"children":34},"ul",{},[35,47,57],{"type":18,"tag":36,"props":37,"children":38},"li",{},[39,45],{"type":18,"tag":40,"props":41,"children":42},"strong",{},[43],{"type":24,"value":44},"可靠性风险",{"type":24,"value":46},"：单张 RAID 卡成为系统盘、数据盘共用的单点故障点。一旦控制器故障或数据盘故障触发 I\u002FO 阻塞，会连带造成系统盘不可用，导致节点宕机。",{"type":18,"tag":36,"props":48,"children":49},{},[50,55],{"type":18,"tag":40,"props":51,"children":52},{},[53],{"type":24,"value":54},"性能相互影响",{"type":24,"value":56},"：系统盘的频繁小块写入与数据盘的批量顺序读写共享 RAID 控制器通道和缓存，会抢占资源，拖慢系统盘 IO 响应，影响系统可用性。",{"type":18,"tag":36,"props":58,"children":59},{},[60,65],{"type":18,"tag":40,"props":61,"children":62},{},[63],{"type":24,"value":64},"硬件超配，成本增加",{"type":24,"value":66},"：在分布式存储和 HCI 场景中，数据盘通常只需 JBOD 直通访问，通用 RAID 卡自带的大容量缓存和复杂 RAID 功能存在超配，增加单节点综合成本。",{"type":18,"tag":27,"props":68,"children":69},{},[70],{"type":24,"value":71},"针对上述痛点，百信信息采用系统盘与数据盘独立管理架构，实现故障域隔离，简化运维并降低整机综合成本。",{"type":18,"tag":19,"props":73,"children":75},{"id":74},"二解决方案",[76],{"type":24,"value":77},"二、解决方案",{"type":18,"tag":27,"props":79,"children":80},{},[81],{"type":24,"value":82},"百信信息采用 TrustRAID B310n 国产 NVMe BootRAID 卡，配套 HBA 卡构建“专卡专用”架构：系统盘由 BootRAID 卡提供 RAID1 保护，数据盘由 HBA 卡实现 JBOD 直通访问。整套硬件方案基于 openUBMC 实现统一带外管理。",{"type":18,"tag":27,"props":84,"children":85},{},[86],{"type":18,"tag":87,"props":88,"children":92},"img",{"alt":89,"src":90,"title":91},"一卡两用与专卡专用架构对比","\u002Fcategory\u002Fusercase\u002Fimplementing-highly-reliable-system-disk-management\u002Fimg1.png","图 1 一卡两用（左）与专卡专用（右）架构对比",[],{"type":18,"tag":94,"props":95,"children":97},"h4",{"id":96},"_1-高可靠故障域物理隔离",[98],{"type":24,"value":99},"1. 高可靠：故障域物理隔离",{"type":18,"tag":27,"props":101,"children":102},{},[103],{"type":24,"value":104},"传统“一卡两用”方案中，HDD 数据盘年故障率约 5%，数据盘故障若触发 RAID 卡 I\u002FO 阻塞时，可能连带造成系统盘失效的风险。“专卡专用”方案采用独立控制器和物理通道，隔离两类磁盘的故障域。B310n 通过硬件 RAID1 保护系统盘，并支持 M.2 SSD 热插拔、自动重构及重构中断接续，故障盘更换免停机。",{"type":18,"tag":94,"props":106,"children":108},{"id":107},"_2-高集成简化部署优化-tco",[109],{"type":24,"value":110},"2. 高集成：简化部署，优化 TCO",{"type":18,"tag":27,"props":112,"children":113},{},[114],{"type":24,"value":115},"TrustRAID B310n 将两块 M.2 NVMe SSD 集成在半高半长 PCIe 标卡上，无需额外配置独立系统盘硬盘盒及连接线缆。产品出厂预配置 RAID1，插入标准 PCIe 槽位即可投入使用，无需二次配置。",{"type":18,"tag":27,"props":117,"children":118},{},[119],{"type":24,"value":120},"系统盘采用专用 BootRAID、数据盘采用成本更低的 HBA，减少通用 RAID 卡的能力超配，叠加省去硬盘盒、线缆带来的成本下降，对比传统“一卡两用”方案，整体可实现 20% 以上的成本优化。",{"type":18,"tag":27,"props":122,"children":123},{},[124],{"type":18,"tag":87,"props":125,"children":129},{"alt":126,"src":127,"title":128},"TrustRAID B310n 产品外观","\u002Fcategory\u002Fusercase\u002Fimplementing-highly-reliable-system-disk-management\u002Fimg2.png","图 2 TrustRAID B310n 产品外观",[],{"type":18,"tag":94,"props":131,"children":133},{"id":132},"_3-易管理统一带外管理接口",[134],{"type":24,"value":135},"3. 易管理：统一带外管理接口",{"type":18,"tag":27,"props":137,"children":138},{},[139],{"type":24,"value":140},"依托 openUBMC 存储管理框架完成 B310n 适配，将盘卡一体设备纳入带外统一管理，核心技术创新点如下：",{"type":18,"tag":32,"props":142,"children":143},{},[144,154,164],{"type":18,"tag":36,"props":145,"children":146},{},[147,152],{"type":18,"tag":40,"props":148,"children":149},{},[150],{"type":24,"value":151},"盘卡一体映射",{"type":24,"value":153},"：构建不依赖硬盘背板的物理盘映射机制，实现板载 M.2 盘的统一管理。",{"type":18,"tag":36,"props":155,"children":156},{},[157,162],{"type":18,"tag":40,"props":158,"children":159},{},[160],{"type":24,"value":161},"SDK 差异屏蔽",{"type":24,"value":163},"：基于 SML 封装 Marvell SDK 的接口和数据结构，屏蔽底层差异，复用 openUBMC 的既有存储管理能力。",{"type":18,"tag":36,"props":165,"children":166},{},[167,172],{"type":18,"tag":40,"props":168,"children":169},{},[170],{"type":24,"value":171},"哈希缓存减负",{"type":24,"value":173},"：通过哈希缓存复用查询结果，减少对板卡的重复访问，降低页面高频刷新带来的系统负载。",{"type":18,"tag":19,"props":175,"children":177},{"id":176},"三客户价值",[178],{"type":24,"value":179},"三、客户价值",{"type":18,"tag":32,"props":181,"children":182},{},[183,193,203],{"type":18,"tag":36,"props":184,"children":185},{},[186,191],{"type":18,"tag":40,"props":187,"children":188},{},[189],{"type":24,"value":190},"提升业务连续性",{"type":24,"value":192},"：系统盘与数据盘故障域隔离，消除数据盘故障连锁引发系统盘失效、节点宕机的风险。",{"type":18,"tag":36,"props":194,"children":195},{},[196,201],{"type":18,"tag":40,"props":197,"children":198},{},[199],{"type":24,"value":200},"简化部署与维护",{"type":24,"value":202},"：出厂自动配置 RAID1、盘卡一体化设计、支持硬盘热插拔，大幅减少配置和维护工作，故障盘更换无需停机。",{"type":18,"tag":36,"props":204,"children":205},{},[206,211],{"type":18,"tag":40,"props":207,"children":208},{},[209],{"type":24,"value":210},"降低总体成本",{"type":24,"value":212},"：相比“一卡两用”方案，RAID 卡成本节省 20% 以上，同时减少硬盘盒、线缆及后续维护开销。",{"title":7,"searchDepth":214,"depth":214,"links":215},4,[216,218,223],{"id":21,"depth":217,"text":25},3,{"id":74,"depth":217,"text":77,"children":219},[220,221,222],{"id":96,"depth":214,"text":99},{"id":107,"depth":214,"text":110},{"id":132,"depth":214,"text":135},{"id":176,"depth":217,"text":179},"markdown","content:zh:usercase:implementing-highly-reliable-system-disk-management.md","content","zh\u002Fusercase\u002Fimplementing-highly-reliable-system-disk-management.md","zh\u002Fusercase\u002Fimplementing-highly-reliable-system-disk-management","md","一、应用背景 在 AI 训练、HCI 超融合和分布式存储场景中，系统盘承担操作系统启动、配置与日志记录等核心能力，其可靠性直接关系服务器节点的可用性。传统“一卡两用”方案由同一张 RAID 卡管理系统盘和数据盘，存在以下问题： 可靠性风险 ：单张 RAID 卡成为系统盘、数据盘共用的单点故障点。一旦控制器故障或数据盘故障触发 I\u002FO 阻塞，会连带造成系统盘不可用，导致节点宕机。 性能相互影响 ：系统盘的频繁小块写入与数据盘的批量顺序读写共享 RAID 控制器通道和缓存，会抢占资源，拖慢系统盘 IO 响应，影响系统可用性。 硬件超配，成本增加 ：在分布式存储和 HCI 场景中，数据盘通常只需 JBOD 直通访问，通用 RAID 卡自带的大容量缓存和复杂 RAID 功能存在超配，增加单节点综合成本。 针对上述痛点，百信信息采用系统盘与数据盘独立管理架构，实现故障域隔离，简化运维并降低整机综合成本。 二、解决方案 百信信息采用 TrustRAID B310n 国产 NVMe BootRAID 卡，配套 HBA 卡构建“专卡专用”架构：系统盘由 BootRAID 卡提供 RAID1 保护，数据盘由 HBA 卡实现 JBOD 直通访问。整套硬件方案基于 openUBMC 实现统一带外管理。  1. 高可靠：故障域物理隔离 传统“一卡两用”方案中，HDD 数据盘年故障率约 5%，数据盘故障若触发 RAID 卡 I\u002FO 阻塞时，可能连带造成系统盘失效的风险。“专卡专用”方案采用独立控制器和物理通道，隔离两类磁盘的故障域。B310n 通过硬件 RAID1 保护系统盘，并支持 M.2 SSD 热插拔、自动重构及重构中断接续，故障盘更换免停机。 2. 高集成：简化部署，优化 TCO TrustRAID B310n 将两块 M.2 NVMe SSD 集成在半高半长 PCIe 标卡上，无需额外配置独立系统盘硬盘盒及连接线缆。产品出厂预配置 RAID1，插入标准 PCIe 槽位即可投入使用，无需二次配置。 系统盘采用专用 BootRAID、数据盘采用成本更低的 HBA，减少通用 RAID 卡的能力超配，叠加省去硬盘盒、线缆带来的成本下降，对比传统“一卡两用”方案，整体可实现 20% 以上的成本优化。  3. 易管理：统一带外管理接口 依托 openUBMC 存储管理框架完成 B310n 适配，将盘卡一体设备纳入带外统一管理，核心技术创新点如下： 盘卡一体映射 ：构建不依赖硬盘背板的物理盘映射机制，实现板载 M.2 盘的统一管理。 SDK 差异屏蔽 ：基于 SML 封装 Marvell SDK 的接口和数据结构，屏蔽底层差异，复用 openUBMC 的既有存储管理能力。 哈希缓存减负 ：通过哈希缓存复用查询结果，减少对板卡的重复访问，降低页面高频刷新带来的系统负载。 三、客户价值 提升业务连续性 ：系统盘与数据盘故障域隔离，消除数据盘故障连锁引发系统盘失效、节点宕机的风险。 简化部署与维护 ：出厂自动配置 RAID1、盘卡一体化设计、支持硬盘热插拔，大幅减少配置和维护工作，故障盘更换无需停机。 降低总体成本 ：相比“一卡两用”方案，RAID 卡成本节省 20% 以上，同时减少硬盘盒、线缆及后续维护开销。",[],[233,233],null,1792752793798]