What are the things to remember when building a website on Amazon AWS?

Source: http://gigaom.com/2012/02/11/which-is-less-expensive-amazon-or-self-hosted/comment-page-2/ Note that is relatively trivial for anyone good to build a CDN using Squid, maybe Varnish, some spinning rust and a few SSDs.

Perhaps an even more important question to ask is "Should I use AWS at all?". At my day job we do an annual review of our infrastructure to make sure AWS is still the best fit for us. After the most recent round of this, I hammered out a blog post that shows the cases where I think AWS is and isn't a good fit. http://gc-taylor.com/blog/2013/12/03/when-and-when-not-to-use-ec2/ Especially if your budget is tight, I'd encourage you to look elsewhere if you can't afford the necessary redundancy (multi-AZ at minimum). AWS does have an excellent portfolio of other services, so that can factor into the decision as well (it did for us).

#1 Assume any component (instance) that is part of your app can fail at any time. Instances go down for many reasons, and there's no HA. You must write code to monitor your instances and that code itself has to be redundant. #2 Remember local storage is ephemeral. If your instance dies, all local data disappears. Most people use EBS or S3 for persistent storage, though it is possible to avoid these by using multiple redundant copies of a database stored locally (this is what Netflix does) #3 Don't make your site's scale dependent on EBS performance. It's slow and highly variable performance will eventually be a problem. #4 You will need an ops team to monitor and manage your AWS instances 24x7. Do not assume you can do it all with software, or that you won't need to do it.

Read the Netflix tech blog for answers to all of your questions. Avoid high disk IO on AWS - it is extremely expensive and will performance poorly.

Many points regarding AWS services, the what to use and not to use points were already raised here a couple of times by other responders, but what falls short is the lack of direction you have for best practices in the architecture design, implementation, and continuous management of your solution to be hosted in AWS. Of course, there are many points you need to address to make sure your site/service remains functional. In general terms, you should make sure the supporting infrastructure resources you do choose to leverage for this should all come together to provide you a robust amount of resources which can handle traffic, mitigate single points of failure, and scale properly. I work for RightScale, where we have helped many organizations simplify the obstacles of cloud infrastructure management and at the same time help you focus on the main priority you indicated in your question: "Coding the website so that hosting would be simple". We have built very high performance reference architectures which are made to be fault tolerant while giving you total control over how you allocate those resources. I encourage you to contact us. We can take a look at what your needs are to better understand and propose a solution that is easy to implement and wont be too taxing on your time while you focus on your bread-and-butter.