There are three pillars to the software-defined data centre (SDDC): software-defined compute, software-defined storage, and software-defined networking. Without any one of these three, the whole edifice of the data centre falls down. We build all three to be resilient, “designed for failure,” and robust. Each can be built and rebuilt from scripts that are stored in distributed version control systems. But at the bottom of every application stack in our SDDC, there is a database or file store that cannot—by definition—be re-created from scripts. This is the core data that we mine and make profit from. What happens if (or when) the edifice collapses? How is that core data protected, and is traditional backup up to the task?
There are few fundamental rules to backup, but the one everyone has agreed on for the last thirty years is the “rule of threes.” Three copies of the data: online, offline, and offsite. Without the first, we have no working system. Without the second, errors can creep into our copies, rendering them useless. As nice as instant snapshots are, corruption of uncopied blocks renders all snapshots based off that block useless. Modern Trojans use encryption to keep our data from us in an attempt at blackmail. Finally, offsite protects us from the catastrophic physical failures: fire, theft, and flood.
Online and offsite are covered in most SDDC designs. It’s only the very smallest of data centres these days that are totally contained within one site, and it’s arguable whether those sites should be looking to Infrastructure as a Service (IaaS) to fix this issue. Two online copies on different sites is certainly better than one. Our issue comes when we look at the offline copy and, specifically, at how we make copies in a way that is as agile as the rest of the data centre. We not only need some way of copying the data offline, but we also need to do it in a robust, automated way, through virtualization-aware logic or, preferably, an API.
Traditionally, backup has run through agents running on servers, backing up to a central device, usually a tape autoloader. As systems have evolved, backup systems have grown to incorporate hypervisor awareness. Veeam entered the market with a new agentless way of looking at virtualized systems, backing up the whole VM as an object, rather than just the data within. VMware and Microsoft both have basic backup systems that can be programmatically controlled (through vCenter or Microsoft System Center). Many other vendors have entered or adapted to this space. This is fine for on-premises systems or hybrid systems for which a copy of the data is onsite with the backup system. But what about pure cloud systems? If your entire data centre is in AWS or Azure, what do you do?
Amazon seems to be leading in this space with its Glacier product. This is an archival-type system with vaults of data that cannot be updated in real time. Glacier acts just like an offline file store, and since it can take any type of data in large volumes, it works perfectly for an offline copy. Like the rest of AWS, it uses a feature-rich API. Microsoft sells a backup service as part of Azure that looks very much like a traditional backup solution. It involves seeding data into the cloud and taking incremental backups from there. This has the advantage that restore is quick, and it works for all infrastructures (public, private, and hybrid cloud, as well as traditional data centres). However, it is less flexible than Glacier. There are pros and cons to each system.
In the traditional space, Asigra is moving in with its software-defined data protection (SDDP) solution based on commodity hardware, with the aim of reducing the costs of traditional backup. Silverstring, founded in 2002, provides Backup as a Service (BaaS) and Disaster Recovery as a Service (DRaaS) in a system similar to that offered by Azure Backup, but with options for onsite appliances, too.
Finally, there is the new kid on the block: Rubrik. Rubrik is producing an appliance that combines backup hardware with software to back up, dedupe, and restore. With the ability to plug-and-play more appliances into one coherent system, Rubrik will appeal to the on-premises data centre owners. The ability to then roll in cloud storage for longer-term archival gives Rubrik its edge.
The backup space for the SDDC is a complex affair, with options covering all aspects of a backup and DR. Whether in the cloud or on-premises or hybrid, there is an option out there for backup and for DR. This leads to a huge and fragmented market and much to consider. The traditional vendors are by no means being left behind, but the upcoming competition is interesting, and it will throw a spanner in the works of any company that rests on its laurels.