It’s really groovy how time flies when you’re having fun. I’ve now been at Rubrik for a little over a year! You may have noticed that I rarely write about them on this blog – other than my work with PowerShell – because I like writing about general interests in technology and going deep into the technical aspects of cloudy things. In this case, I felt the two worlds converged enough that I wanted to review a few things about the brand new Rubrik Firefly early access announcement. Also, I will take the blame (or credit) for the naming of our Firefly release. I’m sure Joss Whedon would be proud. 😉
With a brand new investment round under our belts, the team now has an additional 61 million lead by Khosla Ventures to put towards growing the company further and making the solution even more appealing. I personally think this is cool because the market has been tough for everyone and it shows a lot of confidence in Rubrik. I think this stems from the fact that our customers really like the solution and see actual value from deployment (such as the one fellow who can finally start taking vacations). Thumbs up there. I’ve seen a lot of solutions that sound great on paper but fall on their face (or severely disappoint) when implemented.
The idea behind Rubrik is to abstract away physical topologies and legacy, imperative thinking. Instead, we focus on wrapping data with declarative policies and allowing a smart fabric ensure that the requirements (such as RPO, RTO, availability, etc.) are met. This can all be done via a RESTful set of APIs, which are my favorite way to work with the platform. If you’re looking to build a cloud of some shape and form, this really is a clutch solution that makes it possible to easily offer data services without spending much time managing and owning that piece of the data center.
If you’re keen on cloud, there’s a snazzy panel discussion entitled How to Make Cloud Work for You happening on August 16th at 10:00 AM Pacific. You can register here. It features luminaries Tim Crawford (AVOA), Andi Mann (Splunk), Edgar Magana (Workday), and Krish Subramanian (CloudMunch).
From a technology perspective, Rubrik Firefly is taking an already heady suite of powerful APIs and making them even more awesome. This release includes a plethora of new endpoints, resources, parameters, deeper versioning support, and much more that I’ll cover in the next section. I’ve been taking a lot of enhancement requests from the community and engaging with the API team to see them baked into the product. It’s been really cool to be on the “other side” of the fence and work directly with a product team to build something.
I harp on APIs a lot on this blog, and that’s because I firmly believe that they are the data center glue that everyone should learn. It’s also a language-agnostic process for sending and receiving data from one solution to the next. In fact, my long time friend Eric Shanks and I have put together a little video (below) showing the integration between VMware’s vRealize Automation and Orchestrator products with Rubrik’s Cloud Data Management stack. Suffice to say that there’s not a whole lot of sweat effort to expend on this, and the vRO plugin is freely available and open sourced. This stands in contrast with some other stacks I’ve had the pleasure of trying to integrate that require far too much labor to get working (and keep working).
The video really highlights how simple it is to provide data protection for pets and cattle in the data center. Also, one thing that stands out in my mind as an architect – there’s no need to go back and tell Rubrik that a workload has disappeared. In essence: no retirement / de-provisioning process. Because policies are the abstraction point, the lifecycle of a workload / application is handled intrinsically by the platform. I think that’s pretty groovy.
One of the great things about being an enterprise software company is the ability to add value with code. The type and quantity of the underlying hardware used for on-premises data doesn’t ultimately hold back the value that can be added to the platform. To this end, Rubrik Firefly has added four! new major features in addition to dozens of other enhancements.
The neat thing about Physical SQL and Physical Linux support (which is more aptly just saying native Microsoft SQL and Linux OS integration) is that the same policy-driven SLA Domains will dictate the protection values. This is pretty huge, and was something that was brought up as a concern by our early access customers. They told me that they wanted the product to halt on new features because of concerns over becoming a big, clunky machine (like what they were using before). Rest assured, this is something we take really seriously. Simplicity is king. Making the complex simple is the hard part. 🙂
The Virtual Appliance, called Rubrik Edge, is a nice option for those with satellite or remote offices with a small quantity of VMs providing services to the local staff. I ran a ton of these at a few of my past work environments, and it typically meant relying on someone local to swap out tapes from a desktop with a tape drive. Which ultimately meant that it wasn’t getting done! With Rubrik Edge, you can use policies to bring data back into the primary location while also having a local cache for instant recoveries of files, folders, or entire VMs. Snazzy.
And last, but not least, Erasure Coding. This one is going to put a smile on a lot of faces, because it effectively doubles our capacity and performance on every Brik (the 2U / 4 node enclosure). The math on this is fairly simple. Today, every block that comes into the system is triple mirrored. One data block will always have two mirror blocks somewhere else. Thus, for every 6 blocks written, 2 of them are data and 4 are mirrors. This limits the logical capacity to 33% of the physical capacity. With erasure coding, there are 4 data blocks to every 2 parity blocks resulting in 66% logical space available. And because we leverage the hardware capabilities offered by Intel to handle the coding, there’s no meaningful speed penalty.
If you’re wondering how this affects performance, look at the data block counts again. With twice as many blocks being data blocks (and not mirror blocks), devices within the cluster are spending twice as much time writing actual data. Write performance goes up holistically and scales with the cluster. This can also have an affect on reads since we leverage the read-ahead cache and each node can now contain much more data than before. Either way, performance is going up and not down with nothing more than a code change. 🙂
There you have it. The early access program for Rubrik Firefly has been a hit and I’ve been enjoying the time spent “nerding out” with folks on our new features, workflows, and use cases. There’s something really great and fun about using declarative policies and programmatic workflows to recover, manage, and secure data spanning on-prem and cloud deployments. I hope everyone gets to try it out at some point! 🙂
If you want to talk about Rubrik, the Firefly release, or just technology in general, you can find me at both of the 2016 VMworld conferences and this list of sessions and at Sips and Stogies. Rubrik is at booth 1131 (right by VMware) in the US conference.