Don't plan on an instant failover in the event of a crisis. By getting an accurate picture of how long each phase of recovery will take, you can predict downtime and minimise disruption to your systems.
Disaster recovery (DR) plans are often based on the principle of: "I will fail over instantaneously in the event of emergency," when frankly that's not usually an option. How fast you can fail over is a vital part of the process of planning for DR, and cannot be overlooked without consequence when an emergency hits. There are three main points of time consumption to consider. First is the time-out before you take action at all, next is the amount of time it takes to recover the data, and finally the amount of time it takes to recover the system so users can get back online.
Time to fail over generally will be decided by a combination of technical and non-technical factors. The "time until declaration" is the length of time it will take your technology staff to confirm that there is an emergency in progress and that you must initiate DR procedures. This can be a few seconds for local, automated failover systems, but could be hours for remote or tape-restore solutions. Once you declare the disaster, you must also factor in time for all of the steps required to get ready to fail over. This could include travel to a DR site, recovery of tapes from an off-site storage provider, etc.
Data recovery may very well be the largest time factor to consider if you have to restore from tape. Most drives can take up to several hours to restore a server, and you can't do much else until the data is back on a spinning disk. You can minimise the hit by pre-configuring hardware to immediately accept the restored data, but you still have to copy it from the tape to the system. Replication tools are more expensive to put in place, but save a tremendous amount of time by having the data already on disk when you need to perform a failover process. Either way, you may need to mount volumes, restore snapshots, or perform other tasks that can add significantly to your recovery time.
Finally, once you're ready to restore services, and the data is back on the disks, you still have to take at least some time for last-minute configuration and restoration of data systems. This can include reinstallation of applications, re-routing of DNS systems to allow end-users to connect, and other such tasks. Failover solutions can speed up this process, but you may still need to tweak them, especially for more complex solution sets and software packages.
All told, instant failover is seldom a realistic option for DR planning. There will always be some time spent during the process, but with the right planning and tools you can minimise the delay. Find the combination of features and functions that fits both your budget and your tolerance of downtime, and you won't be surprised if and when you find you have to fail over quickly.
