Amazon Glacier Pricing – a master class in confusion
I am fascinated by the Amazon Glacier product and its potential as the cloud backup store of choice, largely because of it its extremely low price point at around 1 cent per 1 GB per month – or around 30 dollars for a years’ worth of 250 GB quality storage. Also, transferring data into Amazon Glacier is free.
So far so good, right? Well, there is an additional charge for the number of requests or API calls for a lack of a better description – but they should be relatively cheap as well. This charge isn’t very predictable, but depending on your backup software it might be a slightly larger number than the amount of files you are backing up. Anyway, there is really no way to know for sure.
In short, we have an affordable way to back up our data to the cloud. Well, I guess it is time for the confusing part. You had a massive system failure or an accidental deletion, and it is time to get your data back. The process itself is relatively simple; you just fire up your backup software and kick of a restore operation. The problem is figuring out how much it will cost you – and these charges may be substantial.
While it is cheap to get data into the Glacier cloud, it may be relatively expensive to get it back again. Even worse, you are entirely at the mercy of your backup software. Before we dig deeper, you must keep in mind that Glacier reaches its low storage price by being optimized for getting data in, not for changing it or getting it back again.
There are three main price components involved when retrieving data from Glacier, a data transfer or download fee, a requests or API call fee, and most importantly a peak hourly usage fee. Just by looking at the name you can guess that it is the peek hourly usage fee that is the confusing part.
I will write a more detailed blog post about Peak Hourly Usage, but I will give you a short introduction. Within a month or rather a billing period, they will find the hour where you retrieved the most amount of data, and charge as it that was the rate you retrieved data at for every single hour within that month. So retrieve a lot of data in a few hours, and you will pay for it through the entire month. Also, just to add an additional level of confusion to the mix – they don’t count the amount of data downloaded, but the amount of data retrieved or requested. This puts you entirely at the mercy of the backup software you use, if it starts a restore by requesting all your data in one big operation you may were well be looking at a substantial bill at the end of the month. The fact that you may spend weeks downloading the requested data is irrelevant. I won’t give you any numbers right now, as there are additional parameters involved that would typically make it somewhat cheaper – but as mentioned earlier I will cover this in a later post.
Another thing to note is that they will also charge you extra if you delete data that is less than 3 months old. This is only around 3 cents per GB, but it is important to keep in mind. Once again, Glacier is optimized for uploads, not changes or downloads.
And a final and very important note – this is the way I understand the Glacier pricing model. I may very well be confused…