I recently read with interest an article on the future of hybrid architectures by Mike Uzan. To put it bluntly, Mike states that hybrid architectures are always a bad choice. He suggests that hybrid is a compromise between a legacy concept and a new disruptive one, and that in the long run, the newer concept wins 100% of the market, leaving any hybrid solution at a disadvantage.
I couldn’t agree more about the inadequacy of hybrid architectures and would add hybrid cloud to his list, as I recently stated in a previous post.
Where my opinion and that of Scality differ from Mike’s is on the extent to which the world will become entirely all-flash anytime soon.
We are moving to a digital world where the most business value will be created in a digital form. Therefore, businesses need to store more and more data, in the form of IoT, user-generated data, videos, documents, backups, archives of everything, etc. Storing all this in all-flash, even with dedupe – and much of the unstructured data cannot be compressed further or deduped – is prohibitively expensive.
At Scality, we see a world with two types of storage leveraging a multi-cloud (not hybrid cloud) approach.
The first type is one with super low latency (sub-millisecond) for relational databases, virtual desktops, home directories, virtual machine farms, and big data analytics. With such low latency, systems do not have time to go from machine to machine on an Ethernet network. Data needs to be available then and there. All-flash solutions, like those from Kaminario, 3Par or Pure Storage, and hyperconverged systems, like those from Simplivity, are the right architectures for this.
The second type of storage meets the need for capacity storage, for storing the massive volumes of content generated by the digital era, starting with plain old backups, videos, and data lakes. For Global 2000 enterprises, media & entertainment companies, the healthcare industry, and service providers, storing this mass of data on all-flash or hyperconverged systems is not economically viable. The right architecture for this is a distributed system, where the data is stored on cheap industry-standard servers using very large hard drives, and where self-healing software guarantees that the data is safe and the load is balanced automatically. This is what each of the cloud giants is doing, and this is what Scality delivers with Scality RING for enterprise, with all the necessary enterprise bells and whistles, starting with authentication and security.
We don’t believe in hybrid cloud. We do believe in multi-cloud. We don’t think that Global 2000 enterprise will move 100% of their IT to the public cloud. Again, this is just not economically viable. When data is active, when a compute instance is used 100% of the time, public cloud is much more expensive than a well-designed private cloud infrastructure (and you need the skillset in-house).
The Future is Multi-cloud
What we see for the future of Global 2000 IT is an infrastructure composed of multiple clouds. Large enterprises won’t be able to sustain the digital transformation with their traditional IT infrastructure, even if it is virtualized. Simply stated: it is not going to be competitive. In an era where most of our global economic value is being created digitally, if your costs for digital infrastructure are higher than your competitors’ costs, you cannot win the race. Global 2000 organizations will need to transform their IT into “Cloud IT,” with a mix of private clouds and public clouds where it is cost-efficient. They will need to move data from one cloud to another programmatically based on workloads. For example, if you produce videos, you probably want your main repository to be in-house.
When you want to distribute a video through the Internet around the world, you would copy this video to a public cloud, use the burst compute to transcode the video in multiple formats, and use the cloud CDN to distribute the video efficiently. If your distribution campaign is over, you can simply delete the video from the cloud thus halting additional storage costs. And when your video becomes obsolete and you want to archive it, you would move it to an archival cloud like Amazon Glacier. It is indeed data movement, but it is not tiering the way your grandpa used to do it. The newly-launched Zenko (www.zenko.io) product provides exactly this, managing unstructured data in a multi-cloud world.