It was a follow-on system to BigTable. A BigTable setup typically was just in one datacenter, but we had built some replication support on top that supported weak consistency. Spanner, though was meant to be a system that spanned all our datacenters with a single storage system.