CFD3 Prep Post: Rubrik

This is part of my series of posts on Cloud Field Day 3.

I’ve covered Rubrik a bunch of times, both here on the blog, over at Forbes, and at CRN. My considered opinion is that they have a solid product and have been completely nailing the marketing side of things.

Rubrik started out as a hyperconverged backup appliance, based on the same technology as Nutanix was back in the day. Prior to Rubrik, you basically had client-server software that wrote out data to tape (or virtual tape) via a media manager. It was a pain to set up, and usually quite brittle. Managing the scheduling of backup jobs was tedious, and looking after the backups was a full time job for a specialised team at large enterprises. I worked on quite a few projects dealing with backups back in the pre-DataDomain days.

The timing was also excellent, about ten years since the last major change to how backups were done (DataDomain) and nothing much had happened since, and everyone still hated backup. We had a lovely confluence of different technologies: fast compute, cheaper RAM and flash storage, and much denser harddrives. You could throw all of these into a single box of about 4RU and have a basically self-contained backup infrastructure where once it’d take several servers you had to buy yourself.

The software had improved as well, because deduplication and erasure coding was far more useful with the much faster x86 chips that were also vastly more affordable. The client software was generally set up with some kind of backup/restore capability as well, thanks to baking in stuff like VSS and changed-block tracking instead of having to load in separate client software. Multiple vendors took advantage of this “agentless” approach.

Back To The Future?

As I look at the current product range from Rubrik, I see that it has followed basically every other VC-backed tech company into repositioning as a software company. VC people hate hardware, mostly because software has better margins and ‘scale’, as far as I can tell. It’s amusing to me that this is how backup was done for, well, decades, before the new people started delivering their software pre-built into appliances. Everything old is new again, so now we see Rubrik offering software-only versions of their stuff.

Of course, it’s totally different this time around because now the software runs in cloud and not on servers you buy yourself. This is different because…

Now you rent the servers instead of buying them? I guess?

Why Go Software?

Software is easier to sell as an appliance because you have tight control over the hardware platform you allow the software to run on. Look at the Apple ecosystem, for example. When you’re a startup with limited funds, you need to focus on getting just enough stuff working—and working well—that people can start giving you money. You don’t try to support backing up Commodore64 or MicroBEE computers. Startups that try to do too much too early run out of cash and die.

An appliance is also easier to sell, partly because it’s tangible, and partly because you wheel it in and start solving a problem straight away. You can do this with some kinds of software (which is why SaaS is popular) but with something like backup, there’s a local hardware component you can’t get away from in a lot of cases. The cost of large bandwidth WAN links is fairly high compared to 10gigE inside your datacentre, and bandwidth particularly matters for restores.

Rubrik now also has a cloud archive function, which pushes deduped and compressed backup data up to the cloud. This reduces the bandwidth needed (because deduped and compressed) but it’s still further away and over lower bandwidth links than inside a datacentre, so restores back to the datacentre are slower. Restores into the cloud where the data is now can be quite fast. Data has mass.

However, large organisations the likes of the enterprise customers that Rubrik is focused on (for the most part) also tend to have good deals with major server and storage vendors because they get volume discounts. They also like to try to have an homogeneous ecosystem; they’ll try to go all-in on Dell, or HPE, or Cisco, or whatever their server vendor happens to be. (M&A can break this, but that’s a digression for another time.)

This creates pressure for startup vendors to start supporting these enterprise-deal structures, both as a technology, and the ways these companies like to buy IT. As a startup, it can be easier to get brought in on a deal if you partner with a large incumbent vendor and get the benefits of their massive sales investments instead of getting your own (relatively) smaller sales team to go after the big fix on their own.

What’s Next?

I’m sure Rubrik will have some fun stuff to say about what they’re going to go after next. I look forward to seeing what they’ve chosen as the direction to expand.