3 ms·

The last comment stated that Amazon recently spent a lot of effort ripping out Oracle. I would be interested in knowing how that was done and how much better th

by jamra 6y ago

The last comment stated that Amazon recently spent a lot of effort ripping out Oracle. I would be interested in knowing how that was done and how much better things got afterwards.

blinkingled 6y ago
I don't know how much things got better from performance standpoint - Oracle is pretty good on that front - but they built their own Aurora database adding all the missing features to it and then migrating to it. (Bunch of blog posts on Aurora look it up.) Knowing what I know about Oracle licensing it must have gone pretty well on the money front!
akhilcacharya 6y ago
One way things got better is they just migrated to pre-existing, externally available AWS services for many applications. Being able to sell them to Amazon teams at internal cost, getting performance and scale benefits and being able to use AWS primitives for orchestration is a major win. Here's a PR piece about it: https://aws.amazon.com/blogs/aws/migration-complete-amazons-consumer-business-just-turned-off-its-final-oracle-database/ https://aws.amazon.com/blogs/aws/migration-complete-amazons-...
SmooL 6y ago
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
- kablow 6y ago
  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.
- mxz3000 6y ago
  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.
  - SmooL 6y ago
    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.

by jamra 6y ago

The last comment stated that Amazon recently spent a lot of effort ripping out Oracle. I would be interested in knowing how that was done and how much better things got afterwards.

blinkingled 6y ago

I don't know how much things got better from performance standpoint - Oracle is pretty good on that front - but they built their own Aurora database adding all the missing features to it and then migrating to it. (Bunch of blog posts on Aurora look it up.) Knowing what I know about Oracle licensing it must have gone pretty well on the money front!

akhilcacharya 6y ago

One way things got better is they just migrated to pre-existing, externally available AWS services for many applications. Being able to sell them to Amazon teams at internal cost, getting performance and scale benefits and being able to use AWS primitives for orchestration is a major win. Here's a PR piece about it: https://aws.amazon.com/blogs/aws/migration-complete-amazons-consumer-business-just-turned-off-its-final-oracle-database/ https://aws.amazon.com/blogs/aws/migration-complete-amazons-...

SmooL 6y ago

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

kablow 6y ago
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.
mxz3000 6y ago
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.
- SmooL 6y ago
  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.