At Nasuni we treat cloud storage as a component for our enterprise storage solution rather than infrastructure. This begs the question of what makes something a component versus infrastructure. It helps to look back at infrastructure that has been around for so long, it has become almost invisible.
Think of electricity. In Tom McNichol’s compelling AC/DC The Savage Tail of the First Standards War he describes the showdown between Edison and Westinghouse over the standard for power in the US. A good portion of that battle raged around the power generators, the equivalent of the huge cloud storage data centers of today, but a critical component of the war was fought around distribution. Edison realized that without the entire eco-system to support the use of power, it would be impossible to win. So, Edison set out to create not only the power generators, but also the transformers, transmission lines, power meters, connectors and plugs, etc. In today’s world we see a plug in the wall and when we want power we just plug in.
Now, let’s look at cloud storage providers. They’re the generators with a huge amount of power (capacity) available on demand, but using this capacity is not as simple as plugging an appliance into the wall. Today’s operating systems do not know how to talk to these systems natively. Clouds also have odd behaviors like eventual consistency models that break the ways our computers have used storage for the last several decades. While the Internet is certainly part of the distribution solution, cloud storage needs to be a standard that is available to a multitude of devices on demand to be considered infrastructure. While there has been some consolidation in the cloud storage providers around their APIs, but they continue to rely on concepts like eventual consistency to reach their massive capacity and distribution goals. Even if all the providers could agree on a common API, it would take a very long time for new storage interfaces to be adopted – every client, every operating system, every server must be updated by all the different vendors to support the new interface. That can take decades, and that is with most of the interfaces being derivatives of the other. I worked on SCSI interfaces in the 80’s. Today in the enterprise we’re using Fibre Channel and iSCSI for block storage, both of which use the SCSI protocol. For unstructured file storage, CIFS and NFS have reigned supreme for decades with no significant challenges.
Today’s challenge differs from the one Edison faced in that we already have the devices to “plug in” and the plugs are defined (iSCSI, Fibre Channel, CIFS, NFS) yet we cannot connect them. This means that cloud storage is not an infrastructure that we can simply plug into. It must be used as a component to build a solution that can become infrastructure for your business. There are implications to this conclusion. If cloud storage is not infrastructure but a component, then it must be resold. While, like electricity, there may be a generation charge and a delivery charge, you wouldn’t want to have to deal with both units independently. If your electricity goes out, you don’t want to have to try to figure out if it’s a generation problem or a distribution problem and then call different companies based on what you think the problem may be. Plus the two parties may not agree.
Cloud storage must be resold and it must be a component, and as a component it should not be something the users need to deal with directly for billing or support or functionality.
The cloud is an extraordinary new component from which to build storage infrastructure. The cloud allows you to achieve global data distribution, have access to unlimited capacity on demand, achieve unbelievable levels of protection, and preserve as many versions of your data as you wish. The challenge is in the packaging and delivery. Check out how Nasuni uses the cloud as a component to deliver enterprise storage for the new age.