I'm looking to gather some insights from those working with large OLTP databases like DynamoDB, Aurora, RDS, or even self-hosted options. Specifically, I'd love to know your monthly expenditures, the types of databases you're running, how your costs break down by category, and if you'd be willing to trade off some performance (in terms of read/write latencies—let's say both 100 milliseconds and 1 second) for cost savings. Any details would be greatly appreciated!
3 Answers
Currently, I’m running ClickHouse and OpenSearch for a few hundred bucks a month, but it’s mostly for development and testing right now.
How's your experience with ClickHouse on AWS? I'm asking for a customer considering a migration.
In my previous job, we used RDS MySQL (db.t3.xlarge) with one primary and two replicas. Our monthly cost was around $1.2k in us-east-1. Here’s how the costs broke down:
- Compute (3 × db.t3.xlarge): ~60%
- Storage (2TB gp3 @ 6k IOPS): ~20%
- Backups: 7-8%
- Cross-AZ traffic: ~6%
- RDS Proxy: ~5%
For performance, writes were about 5-20ms, reads around 5-50ms depending on replica lag. We could live with around 100ms p95 on replicas to save on IOPS, but anything above 1 second would have caused issues. We saved money by using gp3 instead of io1 and adjusting our replicas seasonally.
Thanks for sharing that detailed breakdown! It really helps to understand the costs involved.
It's tough to compare costs since I've worked with companies spending anywhere from a few hundred to millions on AWS. The expenses vary based on database types—SQL Server, Oracle, MySQL, or Aurora—all at different price points. The spending really depends on scale!
Exactly! Small workloads on a 10 MB database vs. a 10 PB database will have wildly different costs. Sometimes 100ms latencies are just unacceptable for serious applications.
Right, AWS can hit those high margins, and I'm curious how flexible companies are when performance isn't a top concern.
So what happens when you push it to production? And why did you choose ClickHouse?