Design and Evaluation of a Transaction Model with Multiple Consistency Levels for Replicated Data

Abstract

We present here a transaction model which simultaneously supports different consistency levels, which include serializable transactions for strong consistency, and weaker consistency models such as causal snapshot isolation (CSI), CSI with commutative updates, and CSI with asynchronous concurrent updates. This model can be useful in managing replicated data with different consistency guarantees to make suitable tradeoffs between data availability, performance, and consistency of different data items. Data and the associated transactions are organized in a hierarchy which is based on consistency levels. Certain rules are imposed on transactions to constrain information flow across data at different levels in this hierarchy to ensure the required consistency guarantees. The building block for this transaction model is the snapshot isolation model. We present an example of an e-commerce application structured with data items and transactions defined at different consistency levels. We have implemented a testbed system for replicated data management based on the proposed multilevel consistency model. We present here the results of our experiments with this ecommerce application to demonstrate the benefits of this model.