Network Evolution

Building the infrastructure for the changing face of IT


Manage Learn to apply best practices and optimize your operations.

Portable, cloud-based virtual data centers? Sure, why not?

It was once unthinkable, but virtual data centers that can be migrated between cloud providers may soon be real, thanks to SDN and other advances.

I'm wondering if the battle between Infrastructure as a Service (IaaS) cloud platforms is even worthy of debate. That's because it seems that the ultimate goal of any IaaS platform is to give organizations the ability to abstract the data center and run its resources from anywhere, on any service. I'm asking a larger question: What happens when you are able perform a physical-to-virtual (P2V) conversion of the entire data center?

Keith TownsendKeith Townsend

Let's fast forward the conversation around cloud computing by a few years. What happens when Moore's Law, which dictates that computing capabilities double every 18 months, catches up with cloud? There will be an unavoidable effect on cloud computing. If we draw an intersection between the virtual data center and cloud computing, we come up with a model of computing that's much more abstract than it is today.

Today, we have to consider compute, network and storage compatibility as part of an organization's IaaS strategy. When we have complete abstraction, however, compatibility becomes less of an issue. Just as server virtualization lessens the reliance on the underlying hardware, abstraction will lessen the reliance on the individual IaaS cloud platform for virtual data centers.

I don't believe the portability of the virtual data center from provider to provider with a click of a button is too far-fetched of an idea. General x86 compute is already being used in many areas of technology that are not related to traditional use cases. Just a few years ago, it would have been hard to envision being able to run dozens of instances of Windows on a single physical server. X86 virtualization has become so powerful that I was able to install an entire virtual desktop infrastructure (VDI) session on my laptop. It would have also been difficult to believe that x86-based systems would someday threaten to commoditize the networking industry via software-defined networking (SDN).

The virtual data center and the future vision of cloud computing are both deeply dependent on the abstraction of the network, better known as SDN. SDN is enabled by the ability to virtualize the control and data planes at acceptable speeds for production networks. This separation is, of course, enabled by the progression of the software and hardware layer of the x86 platform.

The foundation of this vision for cloud-hosted virtual data centers is that the x86 virtualization used in IaaS will become so powerful that you will one day be able to run your entire data center in the cloud without making major changes to the underlying logical infrastructure. Imagine being able to perform a P2V conversion of your data center to Amazon Web Services' (AWS) IaaS environment or to VMware's vCloud Hybrid Cloud Service.

The ability to migrate your data center wholesale will come as the underlying network virtualization and compute capabilities continue to grow. However, cloud platforms and delivery are about more than just virtualizing the underlying subsystems. They're also about providing the ability to deliver services via a self-service interface and to provide metered-usage capabilities. Virtualization alone doesn't provide these capabilities.

The cloud platform, be it open source (CloudStack, OpenStack or Eucalyptus) or proprietary (VMware, BMC or AWS), provides these additional services. Future IaaS capabilities will allow for more portable infrastructure resources. There still is a need to understand which cloud platform is best for your organization, both in the long term and short term.

Article 6 of 6
This was last published in March 2014

Dig Deeper on Data Center Networking

Join the conversation

1 comment

Send me notifications when other members comment.

Please create a username to comment.

I agree that we will be going into an era where with a click of a button you can decide where to run your workloads. And probably soon you can decide where to run your (virtual) datacenter. Having said that, there are a few things that will need some attention.
1. To become truly hybrid, you need to be able to seamlessly run and move your workloads from one location / provider to the other. This is not yet the case as different providers run different IAAS platforms that are not compatible. E.g. you cannot seamlessly move a v-Sphere workload in your private environment to AWS, you need some form of migration or reinstallation to make that work. So either this gets solved in the coming years or you need to pick your favourite platform and stick to this platform in your private / public environments.
2. Moving workloads to a different location / provider is not going to be the issue. It is the data that is connected to that workload. Traditional applications need their data in close proximity to be able to perform. So when you move your workload, you also have to move your date. And in large quantities that will offer some challenges to keep the workloads online while you move them. You may start the move with the click of the button, but the actual move itself may take hours or even days, just to get the data across. Modern applications are much better equipped to handle distributed data so will be much easier to move.
3. Last, but not least, I do not believe in a P2V move of a datacenter. It is going to be V2V or C2C (cloud to cloud). I think that if you want to move a datacenter this is going to be limited to workloads that are already virtualised. So if you want to move physical workloads, you first need to virtualise them and then they are ready to be moved around.

Get More Network Evolution

Access to all of our back issues View All