Everyone agrees that backups must be sent off site in order to protect your data from large disasters such as fire, earthquake, tornado, hurricane or flood.\nIf you store your backups in the same place you create your data, your backups will get destroyed along with your data. If you put them somewhere else, you at least have a chance at restoring data when the worst happens. The method you choose to get the data to another location will determine the likelihood and cost of such an event. Here are the various ways that people get data their off-site.\n\nOriginal tape given to a man in a van\nThe oldest method is to hand to a \u201cman in a van\u201d the original tapes that contain your backup, but it comes with a number of downsides. Because you are using your original backup tape, you have two choices: use the latest tape or an older tape. If you use the latest tape, you will be able to restore from your most recent backups if you have a disaster. This would allow you to meet a lower recovery point objective, or RPO.\nOn the other hand, you want to use an older tape because sending last night's backup off-site would make it unavailable for day-to-day operational restores. Since operational restores are the ones you do most often, this is the most significant downside of this method. Any recovery time objective (RTO) for operation recoveries would have to account for the amount of time it takes to get the tape back on site.\nAnother downside is the risk of physically shipping media. Unless all of your backups are encrypted as they are being written to tape, the loss of any of your backups might need to be reported, especially if you are subject to data-privacy laws or regulations. Any time you hand tapes to a man in a van, you are opening yourself up to risk.\nThe main advantages of this method are simplicity and a reduction in media costs. The Catch-22 decision of having to decide between a good disastar-recovery RPO and a good operational recovery RTO is a significant one, as is the risk of lost data by physically shifting media \u2013 especially important because the tapes in question are the only copy of your backup.\nCopy of backup to a man in a van\nThe next method is very similar to the first, but instead of sending your original off-site, you send a copy. This addresses many of the previously mentioned downsides, as it allows you to send the latest backup off-site without affecting the RTO of operational recoveries or placing the only copy of your latest backup at risk. You do still have the risk of lost data unless you encrypt your backups, though.\nThe on-site backup used for operational restores could be disk or tape. You would first send your backups to disk, then copy them to tape, then hand the copied tape to the man in a van.\nDisk and replication\nReplication of backups is possible if backups are deduplicated, and you have sufficient bandwidth. Deduplication can reduce by two orders of magnitude the size of data that needs to be replicated.\nBefore discussing this topic, it\u2019s important to understand the difference between target deduplication and source deduplication. Target deduplication is where you use a traditional backup software to send full and incremental backups to an appliance that deduplicates them at the target \u2013 where the data is being sent. Source duplication is a more modern method that eliminates duplicate data before sending backups across the network \u2013 eliminating the duplicate data at the source. This allows backups over much smaller connections.\nBackup directly off-site using source duplication\nIf you can eliminate duplication before ever sending the backups, it gives you an option not possible with any other method \u2013 send your backups directly to wherever they need to go. You can send them to your source deduplication system stored in a different data center, or directly to a cloud-based backup system that supports source deduplication.\nBacking up directly to another location can be very efficient and cost-effective; however, any restores will need to come across the WAN connection, so make sure you are aware of how long a given restore will take. If the restore time from a remote backup meets your RTO requirements, this method may be very appropriate for operational recoveries. However, if a restore across the land will not meet your RTO, you're going to want to look at the next option.\nBackup to a dedupe system on-site, replicate off-site\nYou could backup to a system on-site that replicates your backups to another system off-site. This method will work with either source or target deduplication. The idea is that once the backups have been deduplicated and stored to disk, they can more easily be replicated off-site.\nThis approach has the most advantages and relatively few disadvantages. It gets your backups offsite without using a human being, and it leaves a copy on site for operational recoveries without having to perform those recoveries over the WAN.\nThe main downside of this system is typically cost. Deduplication products often charge a premium, and buying two of them might cost quite a bit. Like the previous method, it also requires significant bandwidth.\nMake sure you use one of these methods\nLeaving the only copy of your backups on-site where your servers are is a really bad idea. Make sure you use one of these methods to put some distance between the data you are protecting and the backup you are protecting them with. Then, of course, make sure you test how well that copy works. But that's a topic for another article.