and the resulting costs mean that keeping our DR in house is the cheapest option. We operate out of two data centers in the UK, one in London the other 150 miles away in Wales and have excellent band width between the sites. Our storage is virtualised across both sites, with storage-layer replication that mirrors a given LUN in both locations. Our server layer is also virtualised using Hyper-V, and we can run any given server image at either site. We also replicate just about all data to a third site 250 miles away in Scotland that just has a file storage capability.

We have chosen to use SQL 2014 with clustering to provide our HA and DR. The underlying infrastructure allows us to run one or both nodes at either site, but typically we run only one node at a given site. Failover then becomes the normal failover time for a cluster, maybe a minute or two for the busiest systems.

We did consider using Availability Groups, but found this very difficult to set up at the infrastructure level. However, we do plan to try this again in a few months.