Lots of people have heard about backing up to the cloud. Many are aware of amazon S3 and similar. Here are some better ways to plan it.

For Safer you need to look toward encryption and multiple copies.

For Durable you need either multiple copies, or other strategies.

For Responsive you need to think about recovery time.

Responsive

Firstly, you need to consider recovery time. If it takes many hours a week to upload the backup, you dont wanting to be waiting for days or weeks to recover your entire data set. Some provider offer courier services. You should test a full recover once or twice a year. You should also ongoing test the accuracy of the data, md5 checksums are good for this.

Security

Secondly, you need to consider security. Do you need to encrypt data, or are you happy with the provider’s security. In some instances, even having a whole dataset encrypted at one cloud provider is not safe. You need to consider jurisdiction, and a raid from authorities or a court order. If the data is only partial and encrypted, then it much safer.

Copies

Having 1 backup in your office, and 1 cloud provider is not really durable. There is a chance you screw both up. It is unlikely the cloud provider will corrupt or delete your data, but there may be other issues, like contractual or missed payments, or your uploads start to fail or get behind.

Another issue is sync vs copy vs snapshot. Lots of solutions involving syncing, and there is only 1 copy of data backed up. This is fairly dangerous. You need to be in a situation of syncing plus snapshots, or replicating snapshots, or copied incremental backups. One solution that is really smart is always archive. So you upload 1 copy of each change file, and keep it forever. It isn’t appropriate for all data, but document based data, the extra size isn’t an issue and it makes the backup much more robust. At a minimum keep multiple copies at your cloud provider, so when you are updating 1 there is still another to recover from.

Backup Format

You need to be careful not to lock up your data, so that it is too difficult to retrieve.

If it is in some proprietary format, or backup up to the cloud in a way that isnt quickly recovered, you may struggle.

In a disaster scenario you need options and speed, not a cumbersome and slow bottleneck.

Multiple Cloud Suppliers

This is where a better solution is born. 2 or more cloud suppliers. Given it is fairly cheap, why not back up to 2 providers. Often they give different guarantees, and different recovery strategies, which gives you some diversity.

You can even have a different strategy for each supplier, like 1 supplier only has the most recent data, and the other has a deeper archive set.

The 50+ year durable strategy

One strategy is to RAID your data across cloud suppliers.

Here is an example. Using something like RAID6, and 4 backup providers. Say onsite in your office is 1, 1 geographically local supplier, and 2 international suppliers.

The data is striped into 4 sets using RAID6. Each of the 4 backup sites get exactly 50% of the data set, or equivalent checksum data.

No one site has a full copy to be compromised. You can encrypt as well if you like.

You need any 2 sites to be available to recover. So your site plus 1 other. So the geo local site can be used to courier the data perhaps the same day.

If you are just reading RAID6 data it is fairly fast to recover.

Now you can loose a cloud provider and replace them, and not have lost the ability to recover.

You can achieve this strategy with as little as 3 suppliers.

Or you can use more suppliers, with a higher redundancy factor. Like requiring data from any 2 suppliers from selection of 6 to be able to recover. Coupled with the write once deep archive, this is where you get the 50+ years of durabilty.