|
IT organizations consider several criteria when evaluating remote data replication architectures during Disaster Recovery (DR) planning efforts. These include consequent application performance, usability, reliability, effectiveness with respect to Recovery Time Objective (RTO) and Recovery Point Objective (RPO) criteria, cost, etc. However, such analysis is unnecessarily complicated because DR objectives are constrained by a variety of direct and indirect influences such as available network bandwidth, application write patterns, network behavior (stability, protocol, reliability, latency, etc), processing resources and storage subsystem performance.
When evaluating long-distance, remote data replication strategies, most IT organizations quickly discover a primary challenge is to introduce the new remote data replication capability unobtrusively -- in a way that is transparent to application users and ongoing operations. Unfortunately, as IT organizations attempt to extend their existing synchronous data replication approaches over long distances, application performance is adversely affected.
A long-standing, popular alternative technique to achieve long-distance, remote replication is known as asynchronous data replication. Asynchronous data replication's advantage is that it unobtrusively provides long-distance, remote data replication, thereby preserving application performance as it provides disaster recovery data protection.
One asynchronous data replication factor of particular importance is the often-overlooked asynchronous replication buffer where data waits at a primary site pending replication to a secondary remote location. While the concept of a replication buffer is common to all asynchronous or periodic replication technologies, the mechanism for achieving such is not. How a solution handles the asynchronous replication buffer has significant ramifications on a number of critical DR measures.
This paper presents the benefits of a logical volume-based asynchronous replication buffer in the context of performance, reliability, recovery objectives and cost.
To read the complete article, please download the PDF. |
