Opening the Door to 200TB in the Cloud
How much data is too much for cloud? Well, the answer is going to be different for everyone – and one of the biggest variables is speed. As it happens, speed is a subject near and dear to our hearts.
We recently asked third-party tester Mediatronics to benchmark our latest release’s speed performance against a leading appliance. Zetta Data Protection outperformed the appliance on every front – including making local backup copies.
While the results are obviously relevant to the “cloud backup versus appliance” debate, they also have implications for large datasets in the cloud, period.
Backup Windows: The Real Constraint on Dataset Size
When people think of backing up large datasets, they often think of the initial backup – “I’ve got this many terabytes, it’ll take forever to back that up to the cloud!” While cloud can be much faster than people assume, it’s also important to keep in mind that the initial backup is a one-time thing. The real governing factor is the size of the incremental backups – and the amount of time allotted for transferring them.
After the data is in the cloud, subsequent backups will only transfer data that’s changed since the previous backup (there are several approaches to how to do this – we use reverse incremental backups). So when you’re making nightly backups, you’re not backing up the entire dataset each night – just changed data.
Which brings us to our test results.
The initial backup in our test was 490GB, which took 2 hours, 56 minutes to complete transfer to the cloud. The incremental backup used a 5% change rate and completed in 1 hour, 7 minutes. What’s more, while we used a 1Gbit connection in the test, the incremental completed while using only 1/20 of the available pipe, the equivalent of a 50Mbit connection speed. 1Gbit connections, admittedly, are not super common (yet), but 50Mbit connections are increasingly popping among businesses.
So to recap, Zetta Data Protection finished an incremental backup for a 490GB dataset with a 5% change rate in a little over an hour, using the equivalent of a 50Mbit connection. That’s great news for people with 490GB datasets, 5% daily change rates and 50Mbit connections, but what does it mean beyond that?
Squeezing More Data Through the Window
To figure out the bigger picture, we’re going to need to extrapolate a bit from our test numbers.
While daily backups can technically take up to 24 hours (any more than that is, by definition, not daily), most companies do their backups at night. For the sake of argument, we’ll define “nightly backup window” as a period of 12 hours (since, say, 7pm to 7am are not traditionally hoppin’ times on most corporate servers).
So, the total viable dataset is going to be determined by how big of an incremental backup can fit in that 12-hour window. Once you’ve determined your maximum nightly incremental size, you can work backwards to the maximum dataset size. At the speed achieved in the test, it would be possible to transfer 250GB in 12 hours over a 50Mbit connection. Assuming the same 5% change rate, that means a total dataset size of 5TB (since 250GB is 5% of 5TB).
That’s at a 50Mbit connection. But what if you ARE one of those lucky people with a 1Gbit pipe? Well, that’s where things get interesting.
Say you have that 5TB server, using a fraction of your pipe to transfer a 5% incremental backup at the equivalent of a 50Mbit connection. If the pipe you have is actually 1Gbit, you can fit 17 of those 5TB servers, all transferring their incremental backups at the same time. All of a sudden you have a much bigger prospective total dataset.
At the same time, a 5% change rate is actually pretty aggressive. Most companies’ daily change rate is closer to 2%. At a 2% change rate, our 12-hour backup window can now hold 40 5TB servers making nightly incremental backups, for a total protected dataset of 200TB.
So there you have it – 200TB in the cloud, closer than you’d think. Want to see how we can be that fast? Check out our free trial and try it out for yourself.