I started having problems recently with CloudWatch returning null sets for the default query and average periods of 60 seconds for the storage metric. This had previously been working, and still works in some cases, but it seems that CloudWatch is perhaps changing its behavior.
Increasing the requested time to 300 seconds and keeping the average at 60 reliably gives metrics to check. The attached patch adds two new options: -t for time period to query (default 5 minutes) and -a for average to request (default 1 minute). It also adds a -d flag for enabling API request debugging via boto.
Experimentation showed that the Amazon API only accepts multiples of 60 for requested averages, so that's why the option units are in minutes rather than seconds.
Thanks Alex for the patch.