Super interesting. Wonder why anyone would pick something like DynamoDB over this moving forward? Seems like you lose serverless pricing, easy scaling for free, and global replication but you gain performance and don't get locked into a proprietary database.<p>EDIT: I guess you get secondary indexes and other sorting/querying options beyond what the Redis API offers. Also looking at current instance pricing for MemoryDB, looks like this won't work for large datasets. You'd need to have a data retention policy in place to archive "cold" data.
Will it handle clustering and failovers better than elasticache?<p>I'd hope so, given its claims about durability, and this would make it more useful than redis/elasticache.