By Gary Chen, Research Manager, Cloud and Virtualization System Software, IDC

Virtual machines (VMs) are the most common way of consuming Infrastructure-as-a-Service. Some VMs are born on the cloud, as in they are designed from day one to be operated in a cloud environment. But there is a huge base of VMs running in enterprise data centers that, over time, customers may want to move to the public cloud. While virtualization is one step to making that easier in terms of abstracting away software from the hardware, in reality it can be deceptively difficult. IDC survey data in the figure below shows the top challenges with moving VMs to the public cloud. In this blog we’re going to focus on the initial challenges in getting the VM to the cloud.


First, a VM may have to be converted into the right format if the cloud runs a hypervisor that is different than the one the VM is currently running on While there is an industry standard format – the Open Virtualization Format (OVF) which defines an immediate, portable VM format – it does not address actual runtime differences among various hypervisors. So the need for conversion remains, and it is more complicated than just converting from one VM disk format to another. Each hypervisor presents different virtual hardware, so drivers may need to be injected for optimal performance. Also VMs may contain environment-specific agents or software such as special guest tools, management/security agents, etc. The old ones will need to be uninstalled and the relevant replacements installed during conversion.

Converter software has varying capabilities in  terms of how it handles these issues. In some cases, customers will need to rely on additional configuration management software or custom scripting to complete the conversion. So, it is no surprise then that VM conversion is a top concern for organizations looking to migrate VMs to the public cloud. Many customers also need the ability to move VMs back and forth between their on-premises and public cloud environments. Unfortunately, when the on-premises and cloud hypervisors are different, conversions may be needed each time that border is crossed. Ideally, the process of moving and integrating VMs across both on-premises and cloud environments needs to be seamless so that workflows can be completely automated. While this is a challenge, it  is necessary in order to prevent conversion issues from slowing down IT processes and hindering cloud usage.

Some of the other top challenges with VM migration revolve around bandwidth, the scarcest and most expensive resource in the cloud today. IDC study respondents indicated that transferring VMs along with their related datasets was another major issue during their migrations. VMs can be quite large and data can be even in the terabytes. Transferring data across the public internet can be somewhat unpredictable and will also be dependent on the WAN connection and the cloud provider’s capabilities. Some ways to improve performance are (note not all may be available, practical, or cost-effective, depending on the situation):

  • Buy a faster WAN connection or possibly buy a faster networking service from the public cloud provider if available.
  • Slim down the size of the VM or dataset. This might not be applicable in many cases or only provide marginal gains, depending on the application or data.
  • Move the VMs or data into an intermediate location that is physically close to and directly connected to the public cloud, a setup that some public clouds now offer.
  • Implement a WAN optimization solution, which can speed up the transfer.
  • Send the data physically via mail! For extremely large datasets, this can actually be faster and cheaper than trying to use the network.
  • Use a storage replication setup, where the transfer can happen in the background or while changes are happening over time. The VM can remain running until the replication finishes, with little downtime during migration.

Stay tuned for the next article, where we’ll delve deeper into issues that can occur once VMs are converted and transferred into the cloud.

View the full list of cloud services at

For future updates, follow us on Twitter and Facebook at @vCloud and