Can Private Cloud Storage Accelerate Enterprise Cloud Adoption?
The barrier to mass cloud adoption by enterprises is realistically two-fold. First, the amount of data that is stored on premise would take an exorbitant amount of time to move to the cloud and would in many cases not be financially beneficial. (we're not even mentioning security and privacy concerns about data and a whole slew of other things). The second thing is, even if we move all the data into the cloud, the communications link would dictate that the compute (VMs) be as close as possible to this data, because most enterprise applications and databases are latency-sensitive. This would suggest that you now need to move your compute and your storage to the same provider, and that leads to vendor lock-in.
There are many options that can address some of these caveats and overcome some of these situations, but there really isn't an elegant one yet for this problem. Well, what if certain data centers could establish direct, high-speed, low-latency connections to the major cloud service providers like Amazon, Azure, vCHS, IBM, and others? It would plug one of the two challenges I mentioned earlier, but then we still have the issue of storage.
Let's stretch the "what if" a bit more: What if these data centers also offered customers the ability to host their storage arrays in a fully managed offering? The customer still buys the storage arrays, but they never see them. Instead, the storage is managed on their behalf and they can pay as they grow. So, instead of buying all the storage upfront, they buy what you need now and then as they need more, it is made available to them. And the customer dictates SLAs and the provider bills accordingly. This would now centralize the storage in a location that can be accessed by several cloud service providers over high-speed, low latency links, thereby doing several things: avoiding lock-in, addressing the security and privacy concerns, and enabling enterprises to move applications and databases into the cloud without worrying about latency or a degraded user experience.
Now, before every vendor under the sun jumps on my comments and says we are already doing this, read carefully what I am suggesting. I understand that today you will place an array on premises or at a customer co-located space. I understand that you are willing to manage it on their behalf, that is great too. What I am suggesting is, take the customer out of the data center business and out of the co-location business altogether. The customer buys the array and where it is stored is not their concern as long as it meets certain criteria. What but the customer needs is that the storage array need not be in a customer cage. Instead, it could be wherever as long as the storage is theirs and managed on their behalf.
Back to the scenario I described earlier: We now have centralized storage in a location that is accessible by cloud providers and they are offering it at a high speed with low latency, which means the compute, your VMs, are now free to live on any cloud. So, they could be on Azure today, and tomorrow VMware introduces a limited-time offer where the VMs could be cheaper to host. Guess what? Migrating those VMs is now very easy. Maybe, the day after tomorrow Amazon offers cheaper prices, so you move back to that provider. Your data is still in the same place, but you are not locked in with your cloud provider anymore.
In this model, we are blending the traditional data center model with the cloud for a best of breed solution. In such a scenario, you can enable services like DaaS and not worry about user experience or performance. You can move your tier-1 apps to the cloud and not worry about performance.
What do you think? Is this the way we are going to migrate to the cloud and get out of the data center and infrastructure-owning business altogether? Where do you see challenges? What am I missing? Please share in the comments here.
Posted by Elias Khnaser on 03/10/2014 at 3:11 PM