The idea does sound interesting - so it looks like they are trying to reduce the amount of network handshaking by imposing a stricter isolation.
I am not sure I am convinced by their results though. They say their deterministic system seems viable when comparing its performance to traditional systems under short in-memory transactions. That is a special case that is clearly in their favor though. In that situation, the amount of time spent in processing data is greatly reduced so the network overhead becomes much more significant - so the system that does less network communications will obviously win...
I guess it may potentially be good for in-memory database systems for stuff like OLTP apps (e.g. VoltDB and TimesTen), but then I think most OLTP apps are okay with a more relaxed isolation...
Daniel Abadi is one of the authors of the HStore paper that VoltDB is based on. It looks like the deterministic order system is using similar requirements and assumptions.
I am not sure I am convinced by their results though. They say their deterministic system seems viable when comparing its performance to traditional systems under short in-memory transactions. That is a special case that is clearly in their favor though. In that situation, the amount of time spent in processing data is greatly reduced so the network overhead becomes much more significant - so the system that does less network communications will obviously win...
I guess it may potentially be good for in-memory database systems for stuff like OLTP apps (e.g. VoltDB and TimesTen), but then I think most OLTP apps are okay with a more relaxed isolation...