ABSTRACT Dynamic workloads on distributed systems and deterministic flow‐based models have comparable challenges. The fixed execution techniques tend to fail as contention or skewed access patterns arise. This causes greater conflicts, latency, and inefficiency of the system. To solve this problem, this article introduces the proposal of the Timestamp‐Aware Deterministic Concurrency Control (TADCC) framework. TADCC is a mix of a timestamp‐based ordering and adaptive execution control. It maintains a constant check‐up on the operation status like the density of conflicts and skews in the workload. On this, it dynamically scales the serialization windows, scheduling strategies, and recovery mechanisms. Incremental Conflict Reordering is also used to manage conflicts in a way that does not involve re‐execution in the framework. Through experiments, it can be observed that TADCC yields better throughput, less latency, and lower conflict rates than current deterministic and adaptive algorithms. It works steadily with high and low contention. The findings show that deterministic execution is adaptive and stable in response to real‐time system signals.
Kanimozhi et al. (Thu,) studied this question.