Increasing the parallelism in transaction processing and maintaining data consistency appear to be two conflicting goals in designing distributed database systems (DDBSs). This problem is especially difficult if the DDBS is serving long-lived transactions (LLTs). A special case of LLTs, called sagas, has been introduced that addresses this problem. A DDBS with sagas provides high parallelism to transactions by allowing sagas to release their locks as early as possible. However, it is also subject to an overhead, due to the efforts needed to restore data consistency in the case of failure. We conduct a series of simulation studies to compare the performance of LLT systems with and without saga implementation in a faulty environment. The studies show that saga systems outperform their nonsaga counterparts under most of conditions, including heavy failure cases. We thus propose an analytical queuing model to investigate the performance behavior of saga systems. The development of this analytical model assists us to quantitatively study the performance penalty of a saga implementation due to the failure recovery overhead. Furthermore, the analytical solution can be used by system administrators to fine-tune the performance of a saga system. This analytical model captures the primary aspects of a saga system, namely data locking, resource contention and failure recovery. Due to the complicated nature of the analytical modeling, we solve the model approximately for various performance metrics using decomposition methods, and validate the accuracy of the analytical results via simulations.
展开▼
