TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Ask HN:How to learn Cloud Computing?

4 点作者 bgurupra超过 15 年前
I have been spending some time reading up on Cloud computing and have developed a interest in it.I want to build deeper skills in this area (and eventually try and find a job in that area) and am thinking probably hacking into open source projects in this area is the best bet for building the deep skills.But the problem is let alone a cluster of machines, all I have right now is Ubuntu running on VMWare on Windows.<p>I am sure there must be lot of HN readers who have deep skills in this area, so my question is with the above constraint if it is still a good idea to go ahead and try and understand the open source code ?or that effort is pointless if I can not run that code in the near future?Also what OSS projects would you recommend for this?I was starting to peek into Hadoop, Traffic Server and libCloud for now

1 comment

patio11超过 15 年前
"Cloud computing" is a lot like "Web 2.0" -- it is a buzzword that has an unstable definition and is very, very broad for a professional specialty.<p>Do you mean "cloud computing" like building web applications which store their data on a server rather than storing it on the client like a traditional desktop app? In that case, build web applications. A short, short list of skills to cover includes HTML, CSS, Javascript, SQL, and your server-side language/framework of choice. I like Rails.<p>Do you mean "cloud computing" like running massively parallel calculations across a fleet of machines? In that case, Hadoop is a decent place to cut your teeth. (It is much, much easier to play with than things were a few short years ago when then-employer got the "grid computing" bug. The buzzwords change, the buzzwords remain the same...) As for actually having a cloud, well, you don't really need one. A cloud of 1 instance is enough for you to play with the APIs and concepts. Abstracting away the number of instances you have is a major portion of why the techniques work, after all. Alternatively, you can run multiple VMs on one physical machine. (Or, use Amazon or what have you and rent a half-dozen instances for an hour or two at a time.)<p>Or do you mean "cloud computing" like "on-demand resource scaling"? In that case, I'd pick your favorite vendor (Microsoft is coming out with Azure, Amazon has S3/EC2/etc) and getting familiar with their APIs, limitations, and how you would actually use them to deliver value.<p>Incidentally, that's the big thing to concentrate on. Toy problems are lovely -- for example, you can calculate solutions to n-queens all day long on your cloud, and that is a good exercise for figuring out the nuts and bolts of how to parallelize a calculation. However, sooner or later you want to be able to do something which actually has value to someone. N-queens does not. You'll find that it is <i>much</i> harder to so something valuable than it is to just run a calculation on a fleet of machines. (There is no push-button API for "do stuff customers find useful", but that is eventually where you want to be.)