I recently moved my Joomla backed consulting website completely to Amazon S3, and have been very happy with the results. I would like to do something similar for my personal blog site at jonzobrist.com, however I would like it to be more dynamic, or at least easily update-able.

For my Joomla site, I did a complete mirror to static html and then uploaded all of that to S3, in a bucket with the same name as the site’s (www.bluesun.net), and changed DNS to point to the CNAME for that bucket’s HTTP address. This involved running wget -r -k -E -p -U Mozilla http://www.bluesun.net, editing the files wget copied to all point at the right places for things like menus, etc, and then uploading the files to Amazon S3.

My goal here is to recreate that in a more automated way, so that I can have a main site that is dynamic, but most, if not all, of the content is served from a static repository on S3. The expected outcome I think will be to take a site that costs around $15-20/month and make it cost < $1 /month. And, if I get some huge surge of traffic, to handle the load gracefully, and scale into the many terabytes of serving up data affordably.

A few quick thoughts/notes;

First, if you don’t change permission on newly uploaded items on S3 they default to your default, which is usually no public access. However, if you upload a new version of a file, it keeps the permissions the previous version had.

Second, you cannot host a naked domain (in this case http://bluesun.net) on Amazon S3. This is more a limitation of the the standards that say you shouldn’t. It means that you need something to redirect your naked domain to your web server. A lot of people don’t do this at all, but I think it’s a good thing to do. I think the details of this limitation will actually come in handy in my hybrid dynamic/static WordPress site.

Third, it makes a lot of sense to compress objects, and setting the right headers on the object will, I believe, get S3 to automatically server it up in a way a browser can understand. Most of the things that make up web pages (HTML and javascript) are text based and compress very well. On the other side images used on the web are generally already very compressed.

Fourth, having a hybrid site means you will still have some dynamic objects and this will mean manually processing (or manually setting up automated processing) html files to separate dynamic from static content.

Fifth, I’m a huge fan of things like Google Analytics, which are hosted by Google, and only included in my site as a static snippet of code that pulls more code direct from their servers. I would love to have something similar for comments and other user generated content that messes up the static website paradigm. I think technologies like AJAX can really shine here.

Brief background, my site (jonzobrist.com) is a standard WordPress install, currently running on an EC2 Micro Instance running Ubuntu 10.04 with Apache/PHP/MySQL all running on one machine. It’s an EBS backed instance, and I snapshot the root volume. I don’t really make updates more than once or twice a week, and none of my content needs to be pushed live in any kind of urgent manner. That said, I use WP to Twitter to auto tweet new posts, so I need to be able to force an update, or handle not having new content on S3 gracefully. I don’t get a particularly large number of visitors, lately about 1,000 a month. My main motivation for doing this is to see if it can be done, so I can do it for other sites I support.

Here is a graphical representation of what I think it will look like when done.

Then I just need to push all the very static content to Cloud Front for CDN!

What do you think?