Data center backup simply wasn’t built for cloud applications like Cassandra and MongoDB. That was Shalabh Goyal’s main point when I interviewed him last week as the Director of Products for Datos IO, and I couldn’t help but agree with him. As you can see in the below video there are a number of reasons why this is true.
I’d say the overall reason that this is true is that traditional backup software was designed with a traditional data center in mind. It was made to backup a series of servers to a series of tape or disk drives via a network of media servers. The classic backup software product that was originally designed for physical servers has adapted over time to the idea that some – if not most – of the servers might be virtual. It has also added functionality for really large servers to be able to backup directly to the backup device without going through a media server. And it has also adapted its support for the device itself, with the most significant advancement being deduplication. One might make the argument that there are newer backup products that were designed to backup virtual machines directly to deduplicated disk, and these products should not be considered traditional backup products.
But when we think about modern cloud applications such as ultra-scalable databases that run on multiple nodes, we are faced with the reality that even modern backup software packages that were designed for virtualization still weren’t designed with much of the cloud in mind. All traditional and most modern backup software packages are still based around the idea of backing up a physical machine or virtual machine to some type of device, be it disk or tape.
The problem with this design is that there are many applications in the cloud that are either not running in a VM or certainly not running in a single VM. There are applications whose data is distributed across dozens to hundreds of nodes, and no single node has access to all the data. How do you backup something like that when you only know how to talk to a VM? Of course there are also applications that are not installed on any VM’s, but are instead running in containers. How do you backup a VM that does not exist? To learn how make sure you watch the video.