Hacker News Reader

Fairly certain they're alluding to Amazon's company wide effort to get off of oracle databases, with most teams opting to use dynamodb instead. AFAIK it took about ~4 years.

Source: interned / worked at amazon for a bit a few years back

View on HN

In at least a few orgs I've seen the switch was to RDS (usually postgres) instead of dynamodb because they couldn't justify an entire redesign of the schema at the same time as the switch. It was a huge effort with a lot of late nights especially at the very end of the cutover.

They definitely didn't replace oracle with dynamo, those are very different beasts and would require complete redesigns of the migrated service. RDS sounds much more appropriate/realistic.

Part of the reason for opting towards dynamodb was it's essentially unlimited horizontal scaling. I know at least 1 team had scale that was just too large for traditional relational DBs, so dynamodb it was. The other part was a large amount of sql usage was essentially just as a key/value store anyway, and dynamodb was vastly cheaper. I agree that RDS is much more akin to oracle, but in my (extremely) limited exposure to teams doing the migration, the majority we're going to dynamodb. As another poster replied, it was clearly a different story in different orgs.

Reply on news.ycombinator.com