Overview of Disaster Recovery
A Backup Article Contributed by Andrew Whitehead
The Need for a Disaster Recovery Plan
In a world of large enterprises, with global operations, operating continuously, and business continuity becomes ever more critical, making a disaster recovery plan becomes ever more necessary. In the ideal disaster recovery plan, the recovery will be completely automatic, with absolutely no loss of data, cost nothing, and happen instantly with no effect on business operations. It would, in fact, be invisible to the business clients. While this is impossible, it makes sense to get as close to the ideal as possible.
Disaster Recovery Issues
The two accepted criteria of a disaster recovery plan are the Recovery Time Objective and the Recovery Point Objective. Recovery Time Objective is the time in which normal business must be restored, this naturally wants to be as short as possible. Recovery Point Objective is the time to which data must be restored to successfully resume processing, commonly the last backup point.
Not all of the data held by a business is critical to basic operations, but deciding what is and what isn't critical can itself become a big undertaking, and actually segregating it even more so. For this reason, many businesses choose not to take this approach and instead replicate everything they have. For businesses with a fairly local site for replication and a direct link, this is a very attractive option.
If protection from regional disasters is necessary, requiring the use of a telecommunications link to transfer data, the cost of regularly replicating everything can be extremely high and it may be necessary to either prioritize data or use less frequent copies. This has an impact on setting the Recovery Time Objective (how much data can be transferred? ) and the Recovery Point Objective (how frequently is data transferred? ).
Rolling Disasters and Disaster Recovery
Like any other event, disasters have a beginning and an end. The time in between is called the "rolling disaster".
Any disaster recovery solution has to provide an image or copy of data, as it existed prior to the disaster, to a secondary location. While any image or copy of data at any time before the disaster can be considered reliable, the reliability of any copies made during the rolling disaster cannot be guaranteed. This is unlikely to be a problem for a short duration of rolling disaster, but it is during an extended disaster. This is becomes especially relevant if a continuous availability disaster recovery solution, in which data is continuously copied, is being employed.
Geographical Aspects of Disaster Recovery
Directly connecting primary and alternate sites via ESCON, with repeaters, sets a maximum geographic separation limit of 43km. Using more modern dark fiber and dense wave division multiplexor (DWDM) technology, sites can be directly linked up to 90km apart. This is enough to connect two metro data centers, and gives greater protection from a metropolitan disaster while allowing the higher bandwidths that this technology offers.
If there is a requirement to give regional disaster protection, by separating primary and alternate sites by a distance greater than 90km, the only means of data replication is over telecommunication lines. As distance, bandwidth requirements, and data amounts increase, this can become a very expensive choice.
