Hacker News new | past | comments | ask | show | jobs | submit login

In my view it depends on what you want.

It's not obvious but in my view, Aurora does not have a great scalability story, it's more of a reliability play currently. It can give you greater throughput due to the internal replication of data, but not faster queries than you can get on your own big server (owned or rented) if you spend the big bucks. When you are getting a big server, it is not particularly cheaper over time, although it does reduce/remove the need for capex and let you spend opex. It is however wonderful if you want to reduce the need for a DBA and/or have really easy to setup and manage replication and resiliency to hardware failures for your database. There aren't great MySQL open source or AWS MPP options out there for scaling-out in my view (MySQL Cluster requires you use NDB tables and MySQL Fabric imposes some significant shard key limitations.)

Postgres-based scale-out options include AWS Redshift (for analytics) or open source Greenplum (for analytics) or, for more regular web/oltp-type traffic the recently open sourced citusdb. The latter two you presumably could run on AWS EC2 instances.

To me the scale-out story for Postgres looks better than MySQL. But if easily-managed reliability is more of a concern, I would probably go Aurora.




Very nice answer, thanks.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: