How to estimate S3 costs?

I’m having a hard time arriving at a good object storage cost estimate; any suggestions?

First be sure you take a look at the CloudBank cost estimation calculator!

Next you will want a month-by-month estimate of the total volume of data you want to keep on the cloud. I quote AWS numbers here but costs are comparable across cloud platforms so these will apply, approximately, for any cloud.

Importing data to cloud object storage is free.

Storing 1 TB of data on the cloud for one month costs $23. This is fast-access object storage. There are archival modes that cost significantly less. They are more appropriate for data you want to keep on hand but will not be accessing frequently.

Transferring data from the cloud back to the internet costs $0.09 per GB. For this reason we discourage researchers from Downloading Tons Of Data From The Cloud. If the plan is to make data available to other researchers there are three things to do to minimize cost.

  1. Fragment the data into small files (sensibly). This way when a colleague wants to download some of that data: They get just what they need, keeping the download volume (and cost) low.
  2. Make the data available for copying or access from your cloud account to someone else’s cloud account. In this way the data never leaves that cloud and you have essentially Zero export (or ‘egress’) cost to pay.
  3. Make the data available to the internet on a Requestor Pays basis. This allows someone else to copy your data to the internet; but they pay for it.

In principle we would like to reduce the egress cost to Researchers to be Zero. At the moment this is not the case. There is a fee as noted above; so we look at ways to mitigate it at this time.