A two-part multi-algorithm concurrency control optimization strategy for distributed database systems

Abstract

In this paper, we propose a novel holistic approach to address the issues of concurrency control after an exhaustive examination of the problem and the various forms it can transpire. The proposed strategy was formulated depending on different perspectives that are based on exploring a wide range of algorithms, methods, and strategies proposed in practice and theory that attempted to address the problem and its forms, but only partially succeeded in doing so. Here we proposed a two-part holistic strategy to optimize concurrency control in distributed environments that address a wide range of concurrency control anomalies by taking advantage of several concurrency control algorithms' strengths while minimizing their weaknesses. The novelty of our approach transpires from two interconnected parts that can be applied regardless of the type of distributed database environment. The first is a structured tier-based data classification system based on data sensitivity with respect to serializability requirements and ranges from strict to very relaxed forms of serializability constraints. The second is a concurrency management algorithm that allocates the appropriate concurrency control algorithm to each transaction depending on the type of transaction and/or type of data being accessed from the aforementioned tier-based classification method. Our proposed method also incorporates a priority allocation mechanism within the concurrency management algorithm. Priority is allocated to different tier transactions depending on the tier's level, which in turn reflects data importance and sensitivity. Although our proposed strategy remains an algorithmic approach as we encountered various challenges regarding performance testing of a novel multi-algorithm approach for handling concurrency control in distributed database systems. However, future work involves testing the performance of our proposed strategy either through real-time systems after considerable adjustments or by constructing an appropriate customized simulation framework. Finally, the potentials of the strategy presented here are very promising, hence, we recommend as we are also optimistic that other scholars are encouraged to further exploit the concept of using multiple concurrency control algorithms within the same distributed database environment.