What is the benefit of a Key-Value Store over Bigtable?

Question

What is the point of using a dedicated Key-Value Store over Bigtable? My understanding of Bigtable is that it is implemented under the hood with SSTables which are key value based. Given that, then what technical implementation advantages does a dedicated Key-Value Store gain over Bigtable by throwing away the api that Bigtable exposes and keeping to just the minimal put and get api.

Answer 1

(*) I'm not sure which key-value storage you have in mind, allow me to guess.

A feature comparison can be found here: http://vschart.com/compare/dynamo-db/vs/bigtable

As you scroll down, you will see the proverbial devil in details, e.g. integrity model is different.

Answer 2

I think the biggest differeence is whether the keyspace is sequential or not. The other features are kind of implementable on top of either approach.

With a non-sequential keyspace you can get much easier load balancing without having to juggle regions around - since you don't need sequential access to keys, you can just use a hashcode to see which cluster node should keep which values.

Of course, simplicity in the API is also not to be frowned upon.