Electronic Commerce using Distributed ERP-Systems with Approximated ACID Properties

In this paper, we will compare and evaluate the properties of some of the most important replication designs with the properties of the no-replication design. The table designs will e.g. be compared with respect to their availability, performance, implementation problems, failure/disaster recovery and ACID properties (Atomicity, Consistency, Isolation and Durability). By using the evaluation, it is possible to design distributed databases/multidatabases that for each table use the most inexpensive table design fulfilling the needs of application. Most of the replication designs that may be used to optimize performance and availability have only approximated or semantic ACID properties, i.e. from an application point of view, the system functions as if all the traditional ACID properties were implemented. But in reality, the designers have used countermeasures against the problems that occur when only local ACID properties are used in the execution of global transactions. 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/or store locations. An E-commerce server can be viewed as a sales location with or without its own store. Therefore, an E-commerce system can be viewed as a special case of a distributed ERP system, and if the Ecommerce server should fail, any of the other sales locations may function as a hot backup.