 |
Business disruption is a real risk that all organizations face, whether from data corruption, human error, hardware failure, or natural disasters. And the downtime and potential data loss caused by these disruptions become very costly from the standpoint of lost opportunity, wasted productivity, and potentially breaches in compliance. For this reason business continuity has become a hot topic for most organizations.
Business continuity from an IT standpoint is the assurance that data is manageable, accessible, and in the case of unforeseen disasters, recoverable. This paper provides a comparison of three common strategies used by organizations to achieve business continuity: snapshot, data replication, and continuous data protection. Snapshot
The simplest and most straightforward definition of snapshot is the following:
Vendor definitions of snapshot aren't always consistent, but the most common understanding of snapshot is a series of logical point-in-time images of data created by logging the incremental changes from an original (or base) copy of the data. Administrators can typically set the frequency of the point-in-time images to meet their business needs (higher frequencies provide more protection from data loss, but require more space and I/O overhead). Snapshot is a popular method for business continuity, largely due to its space efficiency where only the incremental changes in the data need to be stored. The trade-offs are that it provides no protection against physical media failures (because it requires a valid original copy), and recovery is more complicated because there is no live volume to restore from-the volume must first be created by updating the original file with the changes for a given point-in-time snapshot. Snapshot is best used for cost-conscious situations where users are willing to spend more time and effort in the recovery process and forego physical data protection. Data Replication
Data replication is the process of copying data from one environment to another and keeping the subsequent copies in sync, thereby maintaining more than one live, up-to-date version of the data. Backup is different from replication in that it saves a copy of data unchanged for a long period of time. There are two versions of data replication:
The most time / space-efficient form of either synchronous or asynchronous data replication is delta replication - a modified version where replication can be paused at will, including under application control (quite common with relational databases). When the replication process is resumed, only the data that has changed since the time of the previous pause will flow through to the replication target. Compared to the original form of complete data replication, delta replication saves time and bandwidth, and allows replicated volumes or files to be checkpointed, or frozen at a particular point in time. Delta replication is best used for business continuity situations that require improved protection from logical and physical failures with a faster recovery. However, it generally does require more disk space compared to snapshot. Continuous Data Protection
Continuous data protection (CDP) tracks, logs and journals every change made to data, allowing the administrator to restore data at any point in time (as opposed to the set intervals of a scheduled snapshot or delta replication). When data is written to disk, all changes to data are asynchronously recorded to a second location. Continuous data protection is generally easier to configure (doesn't require setting up application interaction, baseline policies, and IT-determined schedules) and requires less time and intervention when data needs to be restored. And because it provides the finest granularity possible, it is ideal for applications that require crash-consistent images to logical objects such as files, mail boxes, messages, and logs. In addition, continuous data protection is as space efficient as either delta replication or snapshot. Continuous data protection is an excellent all-around choice for business continuity, and is particularly important for those applications where crash-consistent and timely recoveries are important.
|
