知识卡片
数据可靠性不变量
为及时发现系统未达标,应明确“关于数据永远应为真”的不变量:①耐久性(durability,数据不丢失/损坏);②一致性(consistency,所有副本相同);③版本控制(version control,定义与数据随变更被追踪);④性能(performance,读写足够快);⑤可用性(availability,读时数据在位)。声明这些不变量不是要保证万无一失,而是让偏差可被察觉、进而改进。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 数据可靠性不变量
原书出处 · 1 处
- Data Management Principles · 查看原书依据
The intent of articulating a set of invariants that should always be true about our data is that it permits us all to notice when they are not true, or when a system cannot properly guarantee that they will be true. This permits us to take action to do better in the future.
参考来源
- 位置:《Reliable Machine Learning》第02章《Data Management Principles》(源文件:extracted/Reliable Machine Learning/_epub-src/OEBPS/ch02.xhtml)
- 结论依据:原文以“invariants that should always be true”归纳五类数据可靠性属性。
- 原始内容:The intent of articulating a set of invariants that should always be true about our data is that it permits us all to notice when they are not true, or when a system cannot properly guarantee that they will be true. This permits us to take action to do better in the future.