For the environment with a 20 hour backup window, any maintenance window in excess of 4 hours will require negotiation with the business, as the backup infrastructure will not be available in readiness for the start of the backup window, posing a risk to the business.
Why would such a maintenance window be required? Primarily for an upgrade to the backup application, but also for the upgrade of microcode of the tape library, drives, or SAN switches. Such environments do exist, requiring careful planning and preparation to achieve the required work, usually in a staged approach. This usually causes pressure for project managers, us techies, as well as nervous times for customers.
Now consider backup environments which to all intent and purpose are nearline, where a backup may run as frequently as every hour. While unusual and rare, such environments exist through a mixture of miss management and workarounds for storage issues. An example of such could be a database with a full log file disk. Storage administrators will not allocate more disk, so friendly backup administrators help DBAs by running backups per hour, allowing the administrators to purge backed up log files.
As the result of a favour, the backup environment - of which it's responsibility and requirement is to secure data in the event of corruption, deletion or loss - is all of a sudden providing, in essence, a real time operation to the business. If attributed to a database the backup infrastructure could be directly responsible for generating revenue. As a result, the daily maintenance window is now less than an hour, compounding the ability to perform any required application and hardware upgrades - cutting it fine would be an understatement for such environments. This adds to the pressure which the project managers and us techies are already under, while the customers bite through their nails at a finger munching rate.
It's probably fair to assume SLAs for such backup environments will not have been updated with reference to nearline operation, but I can envisage situations where SLAs provide no defence when the backup environment maintenance of a few hours causes database servers to crash, and the DBA manager is throwing a fit at the CTO and pointing the finger at the backup administrators' actions.
Just make sure it doesn't happen to you!
About the author: Hywel Matthews is a senior consultant at Glasshouse Technologies (UK), a global provider of IT infrastructure services, with over with 12 years experience in the IT Industry and 9 years experience in backup, recovery, disaster recovery (DR), systems and storage.