The 3-2-1 backup strategy is recommended almost everywhere: keep three copies of your data, use two different storage types, and keep one copy offsite. I understand why it is considered reliable, but I'm curious about other approaches to data preservation. Are there established variations or alternative strategies, and how do organizations choose between them based on cost, recovery time, and the importance of the data?
4 Answers
A practical alternative is a tiered retention plan. For example, keep daily incremental backups for several weeks, weekly full backups for a few months, and monthly full backups for years. Tape is still useful for long-term, offline, and offsite retention. Whatever schedule you choose, regularly perform test restores; retention rules do not help if the backup media or backup software cannot actually recover the data.
There isn’t really a universally safer replacement for 3-2-1. Most alternatives are customized versions of it. For example, an organization might use two local copies plus an immutable offsite copy, or keep additional offsite copies for different retention periods. A smaller setup may use one local backup and one cloud backup, accepting higher recovery time and more risk in exchange for lower cost.
The right design depends on the value of the data and the required recovery objectives. Some environments prioritize availability with replicated storage across multiple sites, while others prioritize protection from disasters and ransomware with offline or write-once storage. RAID, replication, snapshots, and cloud redundancy can improve uptime or make restores faster, but none of them automatically replaces an independent backup. A replicated deletion or ransomware event can otherwise spread everywhere.
Some organizations use schemes such as 4-3-2-1: four total copies including the live data, three separate networks or locations, two offsite copies, and one immutable copy. The exact numbers are less important than covering different failure modes—disk failure, operator error, malware, site loss, provider failure, and corrupted backups. If the data is disposable, a simpler plan may be reasonable; if losing it would be catastrophic, adding independent and immutable copies is usually worth the expense.
That makes sense. Testing restores seems like the part people are most likely to skip, even though it is what proves the backups are usable.