Datrium is a company that I have heard of, but I am not overly familiar with. Even though they were founded in 2012, admittedly the first time I really paid attention to them was at VMWorld 2016. Now in all fairness, I was talking to the engineers there as I was picking up my vExpert Raspberry Pi. Then again recently, at the Toronto VMware UserCon, I struck a conversation with an SE to get a better idea of what Datrium does, but more specifically, what makes them different.
We’ve all heard of HyperConverged, but Datrium bills itself as an Open Converged solution. So what do they mean by that? Their flagship product, DVX, takes the premise of shared storage and turns it on its head. The idea is that you bring in a DVX, which can be expanded up to 32 nodes, and you tie your existing compute into it. Each compute node (that is, your own servers) will require some flash cache it in. From a virtualization perspective, vSphere is supported, and a separate vCenter plugin is required.
This plugin is your gateway to manage the DVX, which given that it is a storage solution for virtual environments is nice when compared to having a separate console. As part of its management tasks, the vCenter plugin performs the orchestration between the DVX and the compute nodes. What I mean by that is it orchestrates the data in such a way that all reads come from the local flash cache, and all writes go back to the centralized DVX box. Because this plugin is managing your data and performance at this point, you can easily bring nodes down for maintenance or replacement without needing to worry about availability. In such a case, the DVX will populate the read cache on the relevant hosts, move the workloads there, and keep on serving up your data.
Ludicrous Insane Speed
One of the unique things about DVX is that the more hosts you have, the better performance you get. With traditional arrays, you typically see a decline in performance as overhead increases due to tasks and orchestration. With the DVX, however, because each host has its own read cache and resource reservation, the trend in performance go upwards.
By default, DVX will reserve up to 20% or the resources on a host in order maintain its level of performance. But what do you do when fast isn’t fast enough? Well, obviously you go to “faster” … but I’m not in marketing, so that’s not a good name. Instead, Datrium implemented a handy “Insane Mode” button which ups the reservations to 40% per system. So what kind of difference will you see with this? Well according to ESG Lab, they saw between a 49% and 200% increase in performance. Not bad for a toggle switch!
I haven’t even touched on Datrium’s “BlanketEncryption” mode. For that, I recommend checking out Jon Hildebrand’s preview post where he does cover off that feature.
Datrium is taking part in their first Tech Field Day event next week in Boston. The event is running May 11th and 12th and will be live streamed on www.techfieldday.com.
Disclaimer: I was invited to participate in Tech Field Day as a delegate. All of my expenses, including food, transportation, and lodging are being covered by Gestalt IT. I did not receive any compensation to write this post, nor was I requested to write this post. Anything written above was on my own accord.