知识卡片
小步改造先验证收益
改造既有服务应先选一个明确问题,用可独立发布的小改动观察可衡量收益,再决定后续投入。把很多关键修改合进一次发布,既增加故障范围,也使效果难以归因。性能、稳定性或更新成本的改善需要实际观察,不能仅因接口更像某种风格就认定改造成功。
这张卡片看不懂?
这张卡片的反馈已满 10 条,暂时无法继续提交。
已有反馈 · 0 / 10 条
还没有人提交反馈。
在整书笔记中的位置
原书出处 · 1 处
- 结语:按收益渐进采用配方 · 查看原书依据
Like any transformation work, it only pays off if there are clear advantages to changing an already existing system. The best strategy is to make a number of independent, small changes over time instead of trying to package up a lot of important modifications in a single production release. You can use the recipes in this book to identify a single problem (e.g., “How can we improve the performance of our data queries?”), and then work through one or more changes to your services and/or interface APIs that will result in measurable improvements. Focusing on making changes that result in measurable improvements has a kind of “flywheel” effect in the company. As each change brings better performance, a more stable runtime experience, and less resource-intensive updates, this adds momentum to your work and makes it easier to get support throughout the organization for other improvement initiatives in the future.