Amazon EC2 Billing Discrepancy
I just found out the hard way that EC2 does not charge by the number of instance minutes used, but instead by the number of portions of an hour that individual instances use. Sound the same? Well, it depends on how you are using it 
I wrote a simple shell script that automates the process of creating an ec2 instance. Not particularly complicated, but I had to run it a number of times until I got it to behave exactly the way I wanted to. During this time, I was creating instances and shutting them down a few minutes later, so that I could test the program execution and flow. I needed to do this a number of times, as I was trying to get everything just right. I tested quite a bit with flat files, although it isn’t quite the same as a live run.
Unfortunately, this is means that for each of these attempts I was being charged for an hour’s worth of time, even though I each time I used less than 5 minutes worth of time. This is a bit bewildering to me, as in my day job I work in Data Warehousing and I know that it is a relatively simple procedure to summarize the number of instance minutes used before billing the credit card. So, I feel that I was overcharged by a significant amount, due to a lack of effort. There have been billing disputes with other internet companies for similar practices, and they ended up in litigation.
To be fair:
- My entire bill so far has been $4.20, not incredibly high. Still, I hardly think this hardly fair, considering I probably used less than $0.25 worth of actual time.
- This policy is spelled out in the AWS terms of service, once you dig in and look for it. However, billing this way is not intuitive and I think hidden on purpose.
Regardless of legalities, I don’t think this is the right way to do things. I hope that the folks at AWS will fix this before the beta ends, as it isn’t going to keep customers happy ::sigh::
October 16th, 2006 at 9:12 pm
I have to admit upfront that my company, 3tera, equips competitors to EC2. Still, I think you’re taking an unfair shot at Amazon. Its just my opinion, but they probably assumed the system would be used more for production runs which, of course, last longer than a few minutes.
I can say this with a little confidence because that’s exactly what my engineers assumed as well when designing our metering system. In fact, I had the exact conversation with them that you spelled out above, and for exactly the same reason. (fwiw, my team chose 30 minute windows, but we eventually compromised on 12 minutes before release.)
IMHO Amazon took an important step in cutting the commitment from a month to an hour. However, we’re in the infancy of utility computing and the billing models are going to evolve and mature at a rapid pace for the next few years.
October 17th, 2006 at 8:20 am
Bert,
Thanks for your feedback. I wanted to be fair to Amazon, as I understand they are running a business and have to make some baseline assumptions. Overall, I still use the service and think it is a good thing.
However, my point is that if we can break the model down to a per hour charge we should be able to divide that by 60. If there are additional costs incurred due to shorter utilization times, then the customer should bear that cost. Still, I think it would make more sense to build it into the per minute/hour charges or provide some sort of discount for longer utilization periods.
My $0.02 …
October 30th, 2006 at 8:55 pm
[…] Every time you run this script you will be charged for at least one hour’s worth of time by Amazon, even if you shut things down immediately. These aren’t my rules, I have complained about them previously. […]