I tend not to write much about VMware these days, mainly because the company has teams of professional bloggers to do that on their behalf. However, I intended to write about VMware and disaster recovery after attending Cloud Field Day 7 in April 2020. VMware’s DR strategy reminded me very much of an “old school” approach that emulates array-based replication. With the acquisition of Datrium, it looks like that gap has now been plugged.
Datrium is a startup that developed a data mobility solution for virtual workloads. DVX, the initial product from the company, implemented a form of disaggregated HCI, with shared storage and local caching in each server of a virtualisation cluster. The concept seemed to be a struggle for some people to understand. However, it became clear that the underlying technology of de-duplication could be used to implement data mobility and specifically efficient disaster recovery. You can listen to an introduction to DVX on this podcast.
- #48 – Introduction to Datrium DVX (Sponsored)
- Modern Storage Architectures: Datrium
- Datrium delivers data mobility with Automatrix
Datrium has progressively focused on DR as a service (DRaaS), using VMC (VMware on AWS) as a secondary target for data recovery. Towards the end of 2019, Datrium introduced DRaaS Connect, the ability to replicate and protect data into a VMC environment without the need to have DVX in the primary data centre. This solution could have been both a blessing and a curse for the company (more on this in a moment).
You can listen to our Storage Unpacked recording on DRaaS Connect here.
VMware SRM is designed very much like old-school array-based replication. One of the biggest challenges in implementing the technology is the need to have an entire VMware environment on the receiving end of the deployment. This scenario wasn’t a problem for large enterprises, where either each cluster runs active/active or one environment houses test/development systems.
The public cloud has gradually changed that architectural model. This transition started with the ability to backup virtual machines into the public cloud, quickly followed by the ability to re-instantiate workloads directly into cloud virtual instances. In both cases, no virtual instances are required until the workload needs to be activated in the cloud. This design is a model many backup and DR companies offer today.
VMC for DR
VMC and Datrium changed the model again. For IT organisations that wanted to retain a look and feel of their on-premises infrastructure, VMC offers this experience. The same VMware tools your teams already know. The crucial difference is that Cloud DVX and Automatrix implements a highly automated failover process that manages all of the challenges of networking and data consistency when failing back from a DR position.
- #63 – Datrium CloudShift
- #101 – Datrium Automatrix with Brian Biles and Tim Page
- S01E04 – Datrium and Cloud Mobility
The other significant benefit in the Datrium solution is the elimination of pre-configured DR equipment. Customers can choose to either spin up VMC when they need it or pay for a “pilot light” minimum viable DR infrastructure and expand quickly in a DR situation. This design vastly reduces the cost of maintaining a recovery option while offering more flexibility for DR testing.
I mention DRaaS Connect towards the top of this post. This feature enables customers to get the benefit of DVX DR without installing DVX as a hypervisor storage platform. From my conversations with the company, my understanding is that DRaaS Connect was seen as a beachhead into customers that might not initially use DVX.
We could reflect on the decision to develop DRaaS Connect as both good and bad. It’s good for customers who get a great DR solution with no commitment to DVX. It’s bad for Datrium as they potentially sell fewer DVX systems. It’s really bad when VMware comes calling because the key asset that is likely to be retained in the acquisition is DRaaS Connect. VMware offers local storage, SAN, NAS and vSAN storage for vSphere, so DVX is almost certainly a dead product and that’s a disappointment all round.
One of the more interesting aspects of the announcement is what this acquisition says about the positioning of VMware with respect to storage and their parent company. VMware has pushed vSAN heavily. This solution is also the primary storage component of VMC. Now VMware can provide super-efficient DR without any shared storage involvement. Is this strategy an attempt to provide some protection if VMware is spun out as a separate company? I’m not entirely sure.
The Architect’s View
Neither VMware nor Datrium published terms of the acquisition. This is generally a bad sign, indicating a distressed sale. I do hope Tim, Brian, Sazzala and the remainder of the team managed to come out ahead. I’ve enjoyed writing about the Datrium story since inception and hope that some of the DVX technology gets retained in some form. Those who remember Virsto will have been saddened by the disappearance of the technology. However, it still lives on somewhere within ESXi. Perhaps DVX will have the same fate.
For VMware, this is a win-win. The company has a more elegant DR solution than SRM that competes with the likes of Zerto – without having to acquire a Zerto to achieve it. VMware now also has a data mobility solution that extends out of the private data centre and allows data migrations between public clouds. This capability alone could be very interesting.
I can’t help wondering if this acquisition will be the first of many, for whom the coronavirus dealt a fatal blow to their expansion plans. In times like these, the strong generally get stronger, and that isn’t always a good thing.
Copyright (c) 2007-2020 Brookend Limited. No reproduction without permission in part or whole. Post #237e.