Factoring Cost Into Your DR Solution
A DR approach that relies on restoration of your SQL Server database from backups or log shipping is going to be the least costly of these approaches to DR—in part because organizations already have a backup strategy in place (as you do, right?). You’ll have to pay for infrastructure at the remote DR site, but you won’t need to spend significantly more to ensure that you have backups and log files ready to go if you need to bring that DR infrastructure online.
Depending on how you configure them, provider-based replication services such as Azure Site Recovery and AWS Disaster Recovery Services can add a significant cost burden. You need to weigh the costs against your RPO and RTO requirements to determine whether this approach is worthwhile.
Finally, implementing a DR solution that relies on asynchronous replication to achieve a near-zero RPO may incur higher costs than any other approach. However, depending on how you implement asynchronous replication, that may not be the case. If you implement using the asynchronous replication services built into SQL Server’s AG feature, you may be required to upgrade to SQL Server Enterprise Edition, which can be very costly if you are accustomed to using SQL Server Standard Edition.
Alternately, if you implement asynchronous replication using a third-party SANless clustering solution you’ll incur the cost of licensing the SANless clustering software itself, but you will not need to upgrade your SQL Server infrastructure to SQL Server Enterprise Edition. That difference alone can drastically reduce the price of implementing asynchronous replication to your DR site. While the total cost of ownership (TCO) of a DR solution based on SANless clustering may still be higher than the TCO of a DR solution that relies on backups or log shipping, a DR solution based on SANless clustering can provide you with a near-zero RPO and an RTO that can be measured in seconds. If your operational needs require that kind of responsiveness from a DR solution, a DR solution based on SANless clustering and asynchronous replication may be ideal.