MinIO
2022-03-03 10:03:30 13 举报AI智能生成
MinIO 对象存储
作者其他创作
大纲/内容
特性
Efficient S3 compatibility
Erasure Coding 纠删码
BitRot Protection
数据治愈
权限管理 IAM
Identity and Access Management
Encryption
Data Lifecycle & Tiering
Immutability(Locking)
Versioning
Notification
Scalability
Replication
Metadata
MinIO has no separate metadata store
object level, strong-consistency
Monitoring, Logging and Alerting
Limits
Maximum object size
5 TiB
Maximum length for bucket names
63
Maximum length for object names
1024
Maximum length for '/' separated object name segment
255
Recover
磁盘损坏
更换、重新mount磁盘即可
node 故障
重新启一个节点即可
是什么？
用于构建高性能对象存储服务 的 软件系统
兼容 AWS S3
设计目标
高性能
可扩展
简约
组件
MinIO Server
MinIO Client
MinIO Operator
Cluster Installation & Management
MinIO Console
GUI Dashboard
MinIO SDK
Java/Go/Python/Js
部署模式
快速体验
单节点单磁盘
Gateway
nfs
Distributed(drives >= 4)
ES(erasure-coding sets)
= Erasure Code Stripe Size
= Erasure Code Stripe Size
4 to 16 drives per set. (单节点多磁盘也可以启用纠删码)
Each object is written to a single EC set
Minio uses the largest possible EC set size
18 drives: 9 * 2
24 drives: 12 * 2
The drives should all be of approximately the same size.
Erasure Code Parity (EC:N)
N 最大取 ES中磁盘数的一半 (即 最小利用率为 50%)
Nodes
Each node shoud have the same number of drives
NTP: 节点间时间差应小于 15min
N nodes & M drives/node
minio server http://host{1...n}/export{1...m}
Cluster Expanding
minio server http://host{1...n}/export{1...m} http://host{o...z}/export{1...m}
pool 1: host{1...n}
pool 2: host{o...z}
New object upload requests automatically start using the least used cluster.
Each pool should have the same erasure coding parity configuration
drives
Direct-Attached Storage (DAS) with XFS for best performance.
0 条评论
下一页
