云中如何实现容错？
我正在开展一个项目，旨在通过弹性 IP 寻址和负载平衡实现容错云。最初，我选择了 Windows Azure，但它通过其门户提供自动故障处理，并且用户无法控制对云上实例的访问。我读过有关亚马逊 EC2 的内容，但我不确定它是否能让我自由地以编程方式处理实例，而不是通过亚马逊门户手动处理实例。我是这个领域的新手。因此，我需要有关如何继续我的项目的帮助。我的算法涉及直接访问运行 Web 应用程序或存储数据的实例。谢谢。
I am working on a project which aims at achieving fault tolerant cloud through elastic IP addressing and load balancing. Initially, I opted for Windows Azure but it provides automatic fault handling through its portal and user cannot control the access to the instances on cloud. I have read about amazon EC2 but I'm not sure whether it gives me freedom to handle the instances programmatically and not manually through Amazon's portal. I'm new in this domain. So I need help on how do I proceed with my project. My algorithm involves direct access to the instances on which my web application runs or my data is stored. Thank you.
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（1）
简短的回答是肯定的。
我在云端设置了网络，从启动新服务器、安装软件到监控，我做的所有事情都没有使用 Amazon 的 GUI。他们的 API 工具可以完成 GUI 可以做的所有事情，甚至更多。
结合使用 Elastic Load Balancing、Auto Scaling Groups 和 CloudWatch，您可以轻松创建一个容错能力很强的环境。
我不明白为什么你需要弹性IP。从本质上讲，它们不具有容错能力，如果分配 ip 的机器崩溃了，您需要手动将 ip 分配给另一台机器。这个过程并不需要很长时间，但它的可用性或容错性显然不高。
亚马逊的弹性负载均衡器无疑是一款容错能力很强的产品，负载均衡实例可以跨区域分布，这意味着您最多可以有 5-6 个实例，每个实例位于一组单独的硬件后面。如果亚马逊在某个地区遇到停机，您的应用程序仍然在线。
弹性负载均衡器还将监视每个实例上的端口，以检查实例上运行的应用程序的运行状况，除了对应用程序进行运行状况检查之外，亚马逊还会对硬件进行运行状况检查，并在出现故障时替换您的实例。
The short answer is yes.
I have my network set up in the cloud and I do everything from starting a new server, installing software, and monitoring without ever using the GUI from Amazon. Their API tools do everything that their GUI can do and more.
Using a combination of their Elastic Load Balancing, Auto Scaling Groups and CloudWatch you can create a very fault tolerant environment with very little effort.
I don't understand why you would need Elastic IP's though. By nature they're not fault tolerant, if the machine that the ip is assigned to crashed you need to manually assign the ip to another machine. This process doesn't take long but its defiantly not highly available or fault tolerant.
Amazons Elastic Load Balancer is defiantly is a very fault tolerant product, load balanced instances can be spread across regions, which means you can have up to 5-6 instances each behind a separate set of hardware. If somehow amazon experiences an outage in one of the regions you still have your application online.
The Elastic Load Balancer will also monitor a port on each of the instances to check the health of the application running on the instance, in addition to health checks on the application amazon also does a health check of the hardware and will replace your instance on failure.