大规模采用Arm：过渡到多架构环境
This is the second blog in a two-part series that describes how Uber adopted Arm at scale. In the first part, we described the foundational work of introducing Arm-based hosts into an extensive x86 infrastructure. We covered how we untangled multiple layers of infrastructure tailored to a single architecture environment and reached the initial milestone of building and deploying a simple service to Arm-based hosts through our deployment platform.
这是描述Uber如何大规模采用Arm的两部分系列中的第二篇博客。在第一部分中，我们描述了将基于Arm的主机引入广泛的x86基础设施的基础工作。我们讨论了如何解开针对单一架构环境量身定制的多个基础设施层，并通过我们的部署平台达成了将简单服务构建和部署到基于Arm的主机的初始里程碑。
In this blog, we describe the journey from a single service running on Arm-based hosts to scaling the adoption of thousands of services. Each service brought its own dependencies, build nuances, and performance considerations. We had to address hidden issues that hadn’t surfaced in years and refine our operational strategies to ensure that we could run both x86 and Arm with confidence.
在这篇博客中，我们描述了从在基于Arm的主机上运行单个服务到扩展数千个服务采用的旅程。每个服务都有其自身的依赖关系、构建细节和性能考虑。我们必须解决多年来未曾显现的隐藏问题，并完善我们的操作策略，以确保我们能够自信地同时运行x86和Arm。
A new mindset emerged that viewed multi-architecture as an integral aspect of our production environment. As we dig deeper, you’ll see how we prepared our codebases, revised our container images, and gradually introduced more services into this multi-architecture environment.
一种新的思维方式出现了，将多架构视为我们生产环境的一个 integral 方面。当我们深入挖掘时，你会看到我们是如何准备我们的代码库，修订我们的容器镜像，并逐渐将更多服务引入这个多架构环境的。
After proving a handful of services could run successfully on Arm-based hosts, we faced a new challenge of extending the adoption of our entire platform, which includes more than 5,000 services. Achieving this would mean revisiting and refining multiple code repositories, untangling years of legacy dependencies, and ensuring nothing would break when we switched architectures at scale.
在证明少数服务可以成功运行在基于 Arm 的主机上后，我们面临着一个新的挑战，即扩展我们整个平台的采用，这包括超过 5,000 个服务。实现这一目标意味着需要重新审视和完善多个代码库，理顺多年的遗留依赖关系，并确保在大规模切换架构时不会出现故障。
Uber’s services are mostly built from three language-specifi...