Deadline-Driven Transactions for any Application Pattern and RTOS
Why deadlines matter
eXtremeDB/rt transaction managers adapt seamlessly to the underlying RTOS schedulers and application-specific data flows, without making assumptions about transaction patterns. Because real-time scheduling priorities differ by application, eXtremeDB/rt can maximize weighted success, honor priority preemption, or follow earliest-deadline-first policies.
How deadlines are guaranteed
To support this flexibility, eXtremeDB/rt provides two alternative transaction managers based on its proven MURSIW (Multiple Read Single Write) policy. The High Priority Earliest Deadline First (EDF) transaction manager maintains a kernel-level queue, sorting requests first by priority and then by deadline to ensure the most urgent work is executed on time.
The Priority Inheritance (PI) transaction manager enables the RTOS to actively manage scheduling by leveraging its synchronization primitives, temporarily boosting the priority of threads blocking critical work so deadlines can still be met. In both approaches, real-time transactions can be preempted when necessary, always under defined rules that preserve consistency and predictability.
The preemption policies balance priorities, deadlines, and rollback costs. While the exact rules differ depending on the scheduling algorithm, in both cases they ensure high-priority work is not delayed by lower-priority tasks.
Click to enlarge
EDF-Based vs. PI-based transaction managers — which is better and when?
The key difference is execution control: the EDF manager orders transactions inside the database kernel, while the PI manager relies on the RTOS scheduler and task priorities. Each approach is better suited to different application patterns, depending on whether tighter kernel-level control or closer RTOS integration is required.
| Aspect | EDF Transaction Manager | PI Transaction Manager |
| Control | Database kernel–driven, ordered by priority + deadline | RTOS-driven, guided by task priorities |
| Strength | Predictable deadlines with kernel-level control | Seamless RTOS integration with priority inheritance |
| Best When | You need strict deadline enforcement in the database kernel | You want system-wide scheduling harmony |