Nov
02

Rubrik: Not Hyper-converged, just easy.

Rubrik is a backup appliance that uses the next wave of technology from the web world to allow an scale out, pay as grow infrastructure for your backup environment. While Rubik is simple and easy to get up and going and has a similar architecture to Nutanix it is not hyper-converged. It’s not hyper-converged because it does not run applications. Rubrik being backed by scale is allowing hard problems to solved and giving simplicity to its customers.

Rubrik Architecture
rubrik

The Interface
The interface layer allows Rubrik to be agnostic in the future but today they only backup VMware environments today. Like Nutanix, anything in the UI is done with RESTAPI call which aloows customer to interact with the system in variety of ways. Cloud Connect allows them to plug in multiple vendors even though they only support AWS today. I know the work required to connect to different cloud providers is no easy task but it’s great to see that they have planned for it up front. The interesting feature that they is the ability to restore at the file level, even if the vmdk is stored in AWS. If you have a 500 GB VMDK in AWS and you only want to restore a 20 GB file, you have the ability to do so at a fraction of the cost.

The Logic
The data management layer handles the whole life cycle of your data, ingest, maintaining snapshots based on splices and archiving to the cloud based on polices. The search indexes all of the data coming into the system so you can quickly grab any data you need.

The Core

The system is built to scale and is a master-less system so any node can be lost and services will continue to run. The metadata layer resides on flash so all metadata lookups are fast. All of the backup catalog data lives inside of the metadata layer, so as you scale out and select different restore points and everything remains fast. The distributed task framework uses the concept of data locality in scheduling jobs so data is getting shuffled around needlessly when doing data processing.

Rubrik also has its only distributed file system to ensure both storing the data in the most compact way but also lay out the data to allow for very quick restores for any version on demand. Data is stored using RF3 but it is configurable. Being a backup appliance Rubrik doesn’t have to worry about burning through CPU and it’s efficiency, instead it has the freedom to use all of the CPU to squish down copies of data so RF3 doesn’t really become an issue.

Being scale-out allows Rubrik to use multiple streams to backup data going to different CPU’s. This is very important when stunning VM’s for something like VSS snapshots and reducing backup windows.

I look forward to seeing Rubrik support other hypervisors and hopefully the Acropolis hypervisor sometime in the future ( I don’t have a road map, I am just wishful thinking).

Speak Your Mind

*