I have written a couple of posts (here and here) about Datrium around Tech Field Day 14 back in May; at that time I was intrigued by their fresh and unusual approach to resolve the challenges associated with both the traditional “non-converged” and the “hyper-converged” infrastructure philosophies, but at the very same time I expressed my concerns about the maturity of their solution. I was eager to see their promising technology mature and today I am very pleased to acknowledge the efforts that Datrium have been making since that day, away from the glamour of the spotlight. Datrium just made a big push and only three months after their TFD showcase they introduced not one but two major technology updates.
The first big announcement is from one month ago – so please forgive me for the delay – and it is about extending Datrium’s support beyond VMware. Datrium started as a VMware-only platform: the level of integration with vSphere I witnessed at TFD14 was so clean and polished that DVX almost looked like a native vSphere feature – I am not talking about the UX experience only but about the overall experience: provisioning VMs, checking an managing performance, working with snapshots etc. was a breeze. But limiting themselves to be VMware-centric would have been too much of a risk for Datrium, so let’s welcome the support of new environments, specifically Red Hat RVH (with KVM on RHEL 7.3), CentOS oVirt (with KVM on CentOS 7.3) and Docker Containers (on bare metal CentOS and Red Hat hosts and with the Datrium Docker Volume plugin). By doing so Datrium positions themselves as – probably – the one single converged platform that can serve at the same time vSphere, RHV/oVirt and Docker workloads, and all of this with the specific benefits of the DVX unique architecture.
The second announcement is from yesterday (15 Aug 2017) and it is probably even bigger. Until yesterday it was only possible to scale DVX Compute Nodes (up to a maximum of 32) without any possibility to scale out Data Nodes. Enter Split Provisioning: it is finally possible to scale independently Compute Nodes (up to 128) and Data Nodes (up to 10).
This means that not only storage capacity will potentially scale up to a capacity of 1 PB but also IOPS (up to 18 M 4K Random Reads) and throughput (up to a whopping 200 GB/s for 32K Random Reads), all of this linearly because of Datrium’s architecture that moves the storage controller (and therefore the IO capacity) from the array to the compute nodes.
These two milestones represent a huge advancement allowing customers to deploy multi-hypervisor infrastructures with different size options, from a small “one-rack” deployment up to a large one for Service Providers use cases.
If I had concerns in the past about the maturity of Datrium’s offering, they are dissipating fast: well done Datrium and let’s see what else you are planning for the future!
PS: Datrium will be at VMworld US, so be sure to visit their booth for a demo or a deep dive chat… even more if you are a vExpert, as you will find a nice surprise waiting for you!