This paper aims to use a case study approach to compare the performance of traditional models and an in‐house and outsourced solution and assess the importance of adequate network bandwidth in the remote replication process.
Using the authors’ autonomous computer system, the cost/benefits of five different disaster recovery (DR) models are determined and reported in a table. Further, experimental data gleaned from a series of data updates on remote replicas connected via the internet are reported.
The results indicate that the traditional models lack performance flexibility. The virtualized replica model and the outsourced solution are similar in terms of investment cost, but the former offers attractive performance and sound reliability characteristics, and provided adequate network bandwidth. The minimum practical speed to support replication updates in the authors’ autonomous system was 10 Mbs.
The results herein are only directly applicable to a system with characteristics similar to the authors’ autonomous system. However, the results do provide a general trend one would expect to carry across a wide variety of autonomous systems.
The primary value of this paper is the evaluation of several DR models in relationship to a live data centre. Each model is evaluated with regard to its cost, complexity, recovery time capabilities, and replication characteristics. While the models themselves are not new, the data puts “a face” on them and provides the reader with the degree of difference among the models.
