7 Comments
User's avatar
Oleg Kuralenko's avatar

I'd say a lot of details are missing that might make this article a lot more informative:

- how does it work in multi-region environment?

- trillion messages per hour / day / week?

- what's used as a schema registry?

- how do they handle "exactly once" delivery for emails sending?

- do they have any patches on top of the mainstream version?

- is replication used at all?

- how many boxes are used and what configurations do they find optimal?

Etc etc

iamtodor's avatar

- What were the reasons for choosing Protobuf over Avro, since the latter also provides schema, schema versioning, etc?

Ravi's avatar

How should we read trillion message count? Is it a rate of message ingestion? Or is it total messages that are stored in Kafka?

Rahul's avatar

Very informative. Thank you so much for sharing this.

Anubhav Jain's avatar

Awesome explanation man !!

Meng Li's avatar

Great architecture sharing