Amazon S3 Glacier Instant Retrieval 128KB issue; how much cost savings over S3 could be tested.

Hello, my name is Agata.

The other day, we backed up all of our office data and tested the price difference between storing it as-is in S3 and storing it in Glacier Instant Retrieval.

S3 is an object storage system that you are probably familiar with, but Glacier Instant Retrieval is an object storage system that, like S3, can retrieve data in milliseconds, and the price is very similar to Glacier, but at a much lower cost. The retrieval fee is higher than S3, so it is basically a storage for backups.

The reason I decided to do this cost comparison this time is because Glacier Instant Retrieval has the following slightly complicated limitations.

S3 Glacier Instant Retrieval has a minimum billable object size of 128 KB. Smaller objects may be stored but will be charged for 128 KB of storage at the appropriate storage class rate.

https://aws.amazon.com/s3/pricing/?nc1=h_ls

In short, if you save an object under 128KB, it’s going to be the price of S3.

And actual office data are often small files, aren’t they?
If you’re going to back up office data, wouldn’t it be better to store it as it is in S3, because the retrieval fee is cheaper? That’s what we started talking about.

Intuitively, I thought it wouldn’t have much of an impact. But I couldn’t be convinced unless I could actually show the figures, couldn’t I?

I could have made a model data and made an estimate, but I thought it was more proof than argument, so I actually tried it with our office data first.
The capacity of the data is about 15 GB.
There are a lot of quite detailed files such as text data, and the ratios are under the following conditions.

I thought it was quite demanding, but here are the results. (Amounts are per day)

It’s obvious at a glance.

Incidentally, the price in the Tokyo Region is 0.025USD/GB for Standard and 0.005USD/GB for Instant Retrieval, which is almost exactly the ratio.

Regardless of the poor conditions, it was within the margin of error for the impact of files less than 128KB.
In the end, if you have a slightly larger file, it doesn’t seem to matter how many 128KB files you have.

I could do a more proper calculation, but as for me, I’m satisfied with this because I found a trend with the actual data.