Data Egress Cost Analysis

Understand the impact of data transfer and egress costs across Microsoft Azure, Amazon Web Services and Google Cloud Platform.

One of the questions most frequently asked by cloud-savvy, price-aware customers goes something like this:

OK, so we like that your tool makes it easy to integrate our cloud database and storage in our centralized data warehouse, but I know our budget will be scrutinized for total cost of ownership (TCO), including our data egress costs. So how much will it cost to move our data out of our cloud into our warehouse?

What is Data Egress

Data egress, in the context of cloud charges, refers to charged fees when data is exported out of a cloud provider. The opposite of this would be data ingress, when data is imported into a cloud provider. In the case of data ingress, there is typically no charge.

Data Egress Cost Analysis: AWS vs GCP vs Azure

The simple answer to the earlier question is, “Not much," but let’s take a look at the actual egress costs of Amazon, Google and Azure. (These figures were last updated in November 2020.)

Cloud Data Egress Cost Comparison

• Amazon Web Services: $0.09 per GB up to 10 TB/mo, $0.085 per GB for the next 40 TB

• Google Cloud Platform: $0.12 per GB for the first TB, $0.11 per GB up to 10 TB, $0.08 per GB after 10 TB

• Microsoft Azure: $0.087 per GB for up to 10 TB, $0.083 per GB for up to 40 TB

*Respective website defaults are used for the costs above.

Egress charges typically average around a dime per gigabyte (GB), with exponentially scaled pricing that makes it cheaper to export more data per GB after a cloud-specified terabyte (TB) cliff. 

Reduce Data Egress Fees With Fivetran

When any connector is set up in Fivetran, the standard behavior is to do a full import of existing data (historical sync), and then only capture changed data (updates, deletes, inserts and schema changes) going forward. We typically expect a small uptick in your egress costs during the historical sync, depending on how large the data set is. Later incremental syncs will have even lower egress costs because our connectors use cursors to pull only updated data.