What it is
Data egress is traffic leaving the cloud provider's network toward the internet, another region, or another provider. AWS and Azure both meter this. They are not shy about it.
Inbound traffic is often free or low cost. Outbound traffic is not. The more you pull out (backups to another vendor, analytics to an on-prem tool, users downloading large files) the more the line item grows.
Architecture that stores everything "in the cloud" without asking where it must travel later is how you invent a monthly toll booth.
Why it matters in the meeting
Decisions that look cheap on day one can lock in a tax on every byte that leaves. Moving a dataset later is not only engineering work. It is a bill. JJ discovers this when networking jumps and nobody changed headcount.
When finance asks why networking jumped, egress is often the quiet answer. The loud answer is usually "the cloud is expensive." Bart will ask which job started shipping terabytes at 2am.
Real world
Teams discover egress when a reporting job starts shipping terabytes nightly to another cloud, or when a partner integration pulls the full object store every hour. Store was easy. Retrieval was not priced into the plan.
Same pattern across AWS and Azure: free to arrive, paid to leave. Multi-cloud "for resilience" without an egress plan is how resilience shows up as a line item named Data Transfer.
Ingress is the welcome mat. Egress is the toll booth. Design for where data needs to live, not only where it is cheapest to land.
What to ask
- Where does production data need to go after it lands, and how often?
- Which jobs pull the largest volumes out of the cloud each month?
- If we moved this dataset to another region or provider, what would egress cost?
- Are we caching or processing closer to the data, or shipping the whole pile for every report?
You just knew a little more Jack than you did five minutes ago.
All concepts