Q: So you went to the darkside, are there really cookies?
A: Not only are there cookies, but also plenty of bacon! I will say that having spent time on the customer and partner sides of the triangle has given me a true appreciation for the contrasts of working at a vendor (and a startup, no less).
Q: Besides the obvious, what enticed you to work for Rubrik as a technical evangelist?
A: Rubrik’s executive team spent time listening to my personal and career goals to best understand how we could make the business relationship thrive. My deeper research revealed a dazzling team of supremely talented folks coupled with a glowing strategic roadmap, thus creating an opportunity that demanded my attention. It also triggered my passion to try something new and challenging while still remaining tightly bound to the community at large. In short, it was a role that I couldn’t turn down.
Q: Why doesn’t Rubrik want to be known as hyperconverged? It’s a buzzword we all know and love.
A: Ah, yes, that’s definitely a hot button of mine! Rubrik is certainly a converged data management platform due to coupling scale-out storage across on-site and cloud resources while also allowing users to directly mount workloads on our appliances. Hyperconverged platforms (otherwise referred to as Server SAN by Wikibon) pool both storage and compute resources for workloads to run directly on the nodes. It’s a subtle yet important difference.
Q: A big hurdle with backing up VMs if they are large, the snapshots can run long and can be stunned for too long. Rubrik says they have a solution for it, how do you solve this problem?
A: That’s a great question. Stunning is a common issue when using vSphere APIs for Data Protection (VADP) with standard VMware snapshots against chatty virtual machines, even if they are relatively small in size. There are two ways that we combat the stun duration for successful backups. First, the Rubrik engineering team wrote a Volume Shadow Copy Service (VSS) Provider that works directly with the guest operating system to efficiently quiesce IO. It also grants much greater granularity than standard API usage. Second, all of the data sent to the Rubrik fabric is ingested by flash, thus reducing the amount of time we have to hold open a snapshot to grab modified data listed by Change Block Tracking (CBT). The muscle power in flash also allows a multitude of backup jobs to flow into the fabric simultaneously while also providing a vSphere datastore for Instant Mount workloads. This mix of reads and writes is the infamous “IO blender” concern. Contrast this approach to legacy backup storage systems that are built of capacity disk and can only eek out a few hundred or thousand IOPS across a cabinet or two while assuming long, sequential writes or the occasional restore.
Q: How do you think Rubrik will change the backup mindset that we as engineers have had for so many years. Data protection that doesn’t do restorations? You know heads are exploding, right?
A: Haha, well, I think we’re making everyone’s life better in a multitude of ways. Another way of looking at change is to label it opportunity, grab on with both hands, and see where it leads. I’d like to imagine that in a few years, the idea of having instantaneous access to data across on-site and cloud resources using a Google-like search engine will be the new normal. A bit like when the iPhone came out and made the Blackberry and RAZR look obsolete overnight. Engineers that can realize the potential of a platform that gives them the ultimate control over data will be of incredible value to their organizations. It’s also much simpler to express data protection in terms of an SLA, and then let Rubrik handle the daily minutia of jobs and tasks, versus wasting time fine tuning a set of backup schedules.
Q: Bipul Sinha, really emphasizes that the file system and software is inspired by modern web architecture, is this the true magic behind Rubrik?
A: Other than the vendors, no one enjoys head swaps or having to buy storage controllers with enough headroom to last 3, 5, or 7 years down the road. Not only is that nearly impossible to do, but it encourages wasted CapEx on depreciating assets and prevents organizations from iteratively improving their data center. The ability to create a shared-nothing file system that also provides instant restore capabilities is the key to scale – both up and down. I’d also give a nod to the massive amount of Rubrik engineers that worked at Google, Facebook, VMware, and so forth – they’re not just talking the talk, they’ve already walked the walk.
Q: This leads me to the scale-out dedupe or using a dedupe domain to cover many nodes. That is pretty slick.
A: I’d agree! Global data efficiencies are especially important as the data capacity required to run an organization increases, and many environments I worked with as a consultant have an “infinite retention” policy. The more data you put into a globally efficient the system, the better the resulting space saving ratios will become. Additionally, the data efficiencies scale-out into the cloud, further reducing the cost of off-premises storage while still being searchable in real-time and supporting single-file restores. It’s pretty groovy tech. 🙂
Q: What’s the biggest game changer that Rubrik brings to the world of backups?
A: My mind is tied between two features: real-time search and a rich set of RESTful APIs. After all, Amy, you hit the nail on the head earlier – it’s all about getting your data back! Predictive search across all data points with full-VM or single-file restore blows away having to keep catalogs on relational databases or tape vaults, especially when you’re the engineer that is dealing with bringing back lost or corrupt data. When added to a full suite of native APIs, it gives our customers and the entire ecosystem the ability to customize or tweak the platform to their unique tastes. In fact, our UI uses the same API that is made available to our customers – everyone is a first class citizen.
Q: Do you miss articulation, your minifigure does not have knees but, on the plus side, a sweet leisure suit!
A: I’m just happy to know that he’s found a good home. And lack of knees means he can cut a rug on the dance floor like no one’s business, right?
Chris viewing Mt. Maunganul, New Zealand courtesy of Alastair Cooke
Watch the technical deep dive from Virtualization Field Day 5.
Arvind Nithrakashyap takes a deep look at the technology behind Rubrik. Recorded at Virtualization Field Day 5 on June 25, 2015. For more information, please visit http://www.rubrik.com or http://TechFieldDay.com/event/vfd5