Document Type



Distributed databases with high performance and availability do not have the traditional ACID properties (Atomicity, Consistency, Isolation and Durability) because long duration locks will reduce the availability and the write performance. The problems of the missing ACID properties may be avoided by using approximated ACID properties, i.e. from an application point of view; the system should function as if all the traditional ACID properties had been implemented. The distributed approximated atomicity property manages the workflow of a transaction in such a way that either all the updates of the global transaction are executed (sooner or later) or all the updates of the global transaction are removed/compensated. In this paper, we will describe a flexible algorithm for implementing distributed approximated atomicity. Frank and Zahle [1] have described how to implement the other global approximated ACID properties.

We will illustrate our algorithm with E-commerce examples. If one of the partaking subsystems fails in a system for E-commerce, the approximated atomicity property will ensure that when an order is accepted, the payment and stock levels are managed automatically in the locations of the partaking banks and product stocks. Even logistics and/or production may be managed by using approximated atomicity.

We have cooperated with one of the major ERP (Enterprise Resource Planning) software companies in designing a distributed version of the ERP system with local autonomous databases in the different sales and stock locations.