[gtowey]

This article is pretty thurough on the subject of replication. It goes through pretty much all the options and trade-offs I've learned over my career in database operations.

The takeaways, I hope people will get from this article is that there is no one solution which will work best in all situations. It is a design decision about what advantages your application needs to have and what sacrifices you are able to make. The more limited and strict you can be about how your application reads and writes data, the more options you have to make your database backend more robust. Applications that "try to have it all" usually just end up doing everything poorly.

The other takeaway is the the physical limits of the universe are the greatest barrier database systems are working against. You simply cannot get around them. When you end up working with a team who is unwilling to acknowledge this, run as fast as you can.