Keep an eye on your I/O usage. If you're using the server for yourself, you're not likely to run up a big I/O usage bill. But if you make your server public, that could change everything -- dramatically.
Figuring out the I/O usage for your instances isn't difficult, but it requires diligence and scrutiny. The EC2 management console provides monitoring tools, although the ones in the free tier aren't as granular as you otherwise get. You can't poll a free instance at more than five-minute intervals, whereas you get one-minute polling with for-pay instances.
You can also poll I/O usage from within the instance itself, using the OS's own tools. Here's one way to do it on Linux. In Windows you can use the Disk Transfers/Sec performance counter.
You can track your service charges through Amazon's reporting system, which also allows you to download the gory details in CSV/XML formats.
Assign an elastic address to save yourself a lot of headache. An elastic address doesn't add a significant amount to your bill, and it allows for easier connections to your system. This goes double for Windows instances, because the Remote Desktop connection tool stores the connection address and password together. Each time your site is provisioned with a new IP address, you have to create an entirely new Remote Desktop connection to reach it.
Back up items in the cloud. You never know when the server you'll be working with may bomb on you or have to be re-initialized. It's better to have any pertinent data already in Amazon's cloud instead of needing to be tediously re-uploaded. An EBS Snapshot is one really convenient way to do this, although you get only 1GB of snapshot storage on the free tier. Alternatively, you can attach an EBS volume and back up files directly to it, the way you'd perform backups from a conventional system to an external drive.
Where from here? Once you've gotten the hang of AWS in the free tier, you'll probably be itching to try something higher up the food chain. The next step up from the micro instances is the M1 Small instance, with twice the memory and a full compute-unit's worth of CPU. Most M1 instances start at around $15 a month.
If you're a penny-pincher who doesn't need a server running 24/7, consider a spot instance. With spot instances, you bid for computing capacity by specifying a maximum price you're willing to pay per hour. If the current price per hour for spot instances goes over that amount (it fluctuates based on supply and demand), your instance will stop running.
Finally, if you want to run something sporadically, such as a backup server, check out the reserved instances. A reserve instance lets you pay a one-time fee for a fixed length of time -- one to three years -- and obtain a significantly discounted hourly usage fee. As of this writing, a single M1 Small Linux instance can be had for as little as $61 a year, plus an hourly rate of 3.4 cents -- or around $354 for the whole year assuming 100 percent utilization.
Sign up for CIO Asia eNewsletters.