| Concurrency control of buffer caches among instances |
| Maintain independent, yet consistent cache of blocks for all instances |
| Allow multiple nodes to simultaneous read/write to database (disks) |
| Maximizes concurrent data access and minimizes (disk) I/O activity |
| Block-level PCM locks, row-level locking preserved |
| A fault-tolerant DLM: if a node goes down, the DLMs will automatically recover and reconstruct the lock database on behalf of the failed DLM |
