I think the idea is that you store your data in a relational model and access it using SQL. But for things that need to be looked up a lot, and very quickly, rather than tagging on a KV store in front of MySQL, use the method he describes instead.
This way you don't end up duplicating the cache, it is shared between the two different types of lookups. You also don't end up with data inconsistancies. And yet you get the speed that a dedicated KV store would have got you anyway.
Comments
I think the idea is that you store your data in a relational model and access it using SQL. But for things that need to be looked up a lot, and very quickly, rather than tagging on a KV store in front of MySQL, use the method he describes instead.
This way you don't end up duplicating the cache, it is shared between the two different types of lookups. You also don't end up with data inconsistancies. And yet you get the speed that a dedicated KV store would have got you anyway.