Including transactional commit timestamps in the primary keys of relational databases

A computer-implemented method, comprising:

• receiving, at one or more participant servers in a distributed system, a request to commit a transaction;
• obtaining, by each of the participant servers, at least one exclusive lock for a time range starting at a locally chosen starting time to a predetermined upper bound;
• providing, by each of the participant servers to a coordinator server, a notification that the respective participant server is prepared at a respective prepared timestamp chosen within the respective time range;
• committing, at each of the participant servers, the transaction at a commit timestamp that is equal or greater than each of the prepared timestamps;
• and
• releasing, at each of the participant servers, the at least one exclusive lock.

A system, comprising:

• a participant server, the participant server comprising one or more processors configured to:
• determine that a single-site transaction is to be committed at the participant server;
• compute a single-site commit timestamp having a predetermined pattern for the single-site transaction, the predetermined pattern being one that any multi-site transaction cannot choose as its commit timestamp;
• and
• commit the single-site transaction at the single-site commit timestamp.