Convergence and hyper-convergence are two concepts that have proven to be very popular with data centers, allowing them to shorten implementation times and in some cases condense the IT stack to simplify operations and support. But these concepts have mostly focused on the compute, network and primary storage infrastructures. Meanwhile, the backup infrastructure lags behind and remains a thorn in the side of IT. What if the concepts of convergence and hyper-convergence were applied to the data protection process? This is the approach that Rubrik’s founding team, who come from Google and Facebook, have taken with their converged data protection appliance.
The Backup Problem – Too Many Parts
When the cover is pulled back on the data protection process it is a miracle that it works at all. The backup software has to manage not only the backup server but media servers as well. It also has to maintain file history and media indexes. If the software has the ability to deduplicate and/or replicate data it has to manage the indexes associated with those processes. And of course it has to ingest data from a variety of other platforms and make sure it is safely stored.
Even the backup hardware can be complex. Typically a data center needs at least three types; a storage system for backup data, another for archive and still another for off-site storage. Each of these often have their own capabilities, including deduplication, compression and replication, that typically compete with the backup software’s similar offering. In essence the customer is paying for these features at least twice. To make matters worse each of these storage types are often scale-up, meaning that when maximum capacity is reached a new storage system needs to be added to the mix.
The Backup Appliance Problem
Some backup software vendors have tried to solve this problem by delivering a backup appliance that pre-integrates the backup software into a backup server. While this approach may reduce initial installation time it does not resolve the issues with long-term ongoing operations. The organization is still counting on a legacy backup application to protect a thoroughly modern data center. This approach also typically introduces new scaling limits since the backup server compute and storage can not typically be scaled beyond the single server.
The Hyper-Converged Backup Solution
Rubrik is attempting to solve this problem with a hyper-converged approach. It starts with new data protection software built from the ground up for the modern data center. It leverages a web-scale architecture, one where backup server compute and hybrid flash storage capacity can be scaled across multiple nodes. This type of design also allows for fast ingest, and for deduplication and compression efficiencies to scale across nodes. Finally, replication for off-site protection is provided via an integrated connection with Amazon S3.
The Rubrik backup repository is what they call “live storage” which means that if a primary storage system fails the production application can mount its data directly from Rubrik. Because Rubrik has been the first to design a hybrid flash architecture into a backup appliance, the r300 can be used for more than just backup and recovery. For example, dev/ops can leverage the backup copies to conduct their work as well.
Today Rubrik can back up VMware virtual environments and physical Oracle servers via the RMAN API. They will be adding the ability to backup physical MS-SQL servers as well as support the KVM hypervisor in the near future. Once completed that coverage closely matches the needs of the modern data center that is mostly virtualized with the exception of physical database environments.
The number of moving parts and touch points make data protection a very challenging infrastructure to implement and maintain. While backup appliances and backup disk storage simplify these infrastructures, much more progress is needed. A web-scale/hyper-converged concept makes a lot of sense and Rubrik seems well positioned to provide it.