I did exactly the same thing. Why, with just a few dozen hours of hard work, I could probably slice 10% off my $150 a month server bill! And it would make pretty graphs and take cool algorithms and be so much cooler than A/B testing button colors even if that would be about 20 times more effective per unit of effort.
My job as a business owner is to ruthlessly soulcrush the engineer in me some days.
Yes there is: if you have work to be done with a low priority and no deadline, it's perfect.
For example, I bet the EC2 bill would be cheaper than my electric bill if I wanted to run SETI@home on a bunch of instances when it dropped below a certain mark.
The tech didn't exist 2 weeks ago, but now that it does there is a whole class of problems that can be solved more easily.
I'll echo akamaka -- I like your graphs far more than timf's flash-based ones.
You might like to consider adding some indication (dotted lines, perhaps) of what the hourly rate for reserved instances is, since this should be approximately a floor on the hourly spot rate.
EDIT: Also, please adjust the vertical scaling based on the set of instances shown -- if I'm only looking at the price history for m1.small, it's silly to have it squished into the bottom 5% of the graph.
Suggestion -- use small multiples showing the data center prices on one screen, probably by making the graphs 1/4 size. Place your explanation in the space left empty since there's only (currently) 3 centers.
"A small multiple design consists of a single design repeated several times within the eyespan, each example showing a different value of the independent variable(s)."
With us-east-1.linux.m1.small prices averaging around $0.03/hour, it might make sense to just set up an instance to continuously run on spot prices instead of paying the regular $0.085/hour. It won't ever be more than $0.085/hour (otherwise users would simply switch to the standard price) and it will probably be a lot less.
Unless, of course, the spot price instances are somehow different due to their transient nature, but I don't see anything to that effect on the site. Is there some reason why this would not work?
That's what I would figure, but it also makes no sense why it would exceed the on-demand price for the same instance, except perhaps in moments of sudden demand. Maybe Amazon will intentionally make it difficult to move an instance from on-demand to spot and back to discourage this?
Spot pricing does look good when you don't need extra EC2 instances on a regular basis and when you can write restartable applications.
That said, spot prices are fairly much the same as reserved instance pricing, except with reserve instances you have an extra up front fee. I (and some of my customers) simply use reserve instances.
I also started thinking of ways to use this, but reminded myself to not get to distracted from current work.
1. Put checkboxes for every type and then combine the chart for selected types. It will be great to compare m1.small for all regions, or most cheap types in us-east-1
2. Put a chart with calculated savings: us-east-1 m1.small cost $10 but currently is trading for $5, so the chart will show you are saving %50, etc.
There is no rational reason for this.