Last weekend I spent 1.5 days at Railsconf. It was a 6,000 mile round trip squeezed into a weekend between busy working weeks, and interrupted by the datacenter explosion that took down our site. So, I didn't get to see many technical talks. I did get a feeling for what's hot: cloud computing.

The vendor booths were dominated cloud computing infrastructure providers. Amazon was there representing EC2, the key enabling infrastructure. Heroku showed rails development and deployment on EC2, backed by their recently announced VC funding. Morph showed a more enterprise-oriented service for hosting and scaling Rails and java applications on EC2. Rightscale showed a polished version of their more general-purpose EC2 cluster management tools, backed by their recent VC investment. Engineyard, which runs their own full-service, fully staffed virtual server cloud, was doing a booming business, backed by their recent VC investment. They talked about releasing some of their management tools as open source projects. Clearly the Silicon Valley money is going into infrastructure at this stage. Sun was showing off an Assembla-like workspace that will showcase their ability to host team applications on their cloud computing infrastructure. Google was the big player that wasn't there, as their recently announced Google App Engine runs only Python and faces off against the Ruby-centric Amazon cloud.

I'm a convert to cloud infrastructure. We have started moving our clients onto Amazon servers. Assembla.com was hurt by our dependence on a more traditional server topology, and we'll be moving it into a virtual server cloud over the next month.

There are number of implications for software developers:

• You can get instant staging servers. Start releasing even faster. We find this very useful, and we will build it into Assembla workspaces.

• You can put high-volume fragments of your app on high-volume hosts. For example, the big selling point of Google App Engine is that it can handle a very large number of HTTP requests and survive a slashdotting or a facebook-scale rampup. If you are considering adding a facebook front end, or a component like polling chat that is very request intensive, you have an incentive to throw that onto Google App Engine and link it to the rest of your systems with web services.

• You need to think about file management. Files in a system like Amazon EC2 don't just live on a permanent, homogeneous SAN. You need to back them up to a storage cloud (S3) and restore them when you move to new virtual servers. Amazon and Google offer BigDB and SimpleDB, which aren't exactly the same as the databases and persistent stores you are used to.

• We have to move beyond HTTP request based applications to new mesh-computing architectures that send messages in real time. I saw some interesting projects based on the XMPP jabber protocol, Erlang, and the eJabberd server, which is a jabber server written in Erlang. To back up, XMPP is a protocol for sending messages to jabber instant messengers, but it can also be used to send messages between servers, or to route messages through a high-volume cluster to handle Twitter style applications. Erlang is a computer language that naturally distributes processing to multiple servers. We are seeing these tools deployed to manage servers in the cloud. However, we are also seeing applications in which browser ajax panels connect directly to XMPP servers for real-time chat or dashboards.

• Having a professional-quality repository with a build, stage, release process is important for cloud applications, because your team and your code are likely to be distributed in the cloud just as your servers are.