服务架构有时是两极分化的。我们半数的行业正在尝试使用用户级TCP堆栈来降低每一盎司的延迟（提升网络性能），而另一半人则使用Ruby＆Kubernetes和覆盖网络(overlay network)来降低网络性能并将CPU使用率提高到10年前。
Service architecture is so polarized at times. Half of our industry is trying to use userland TCP stacks to reduce every ounce of latency and the other half is using Ruby & Kubernetes and overlay networks that reduce our network perf and increases our CPU usage to 10 years ago.