Using Multi-Version Concurrency Control

The idea is that in a well-designed database system, 99% of the time transactions succeed. Or you use optimistic concurrency control and check at the end if anything bad happened. You discussed partitioning the data by core. How does a system like this handle transaction including data that is pinned to different cores? Right. That was the scheme Volt used to use. They found out that they have a much more sophisticated scheme now that doesn't require you to synchronize clocks across multiple systems. And so that requires a blackboard for me to tell you exactly how it works.