We have discussed many of the benefits of cloud computing. However, the balance between benefits and disadvantages might at some time shift to a situation where an organization wants to stop using cloud or switch to another cloud service provider.

There may be many reasons to not continue the cloud service. The reasons can be categorized into voluntary leave or involuntary leave.

Reasons for a voluntary leave can be:

Discontinuing the services of a cloud provider can be involuntarily as well. The main reason for that to happen is when the cloud provider goes broke or decides to discontinue their service.

This happened, for example, in 2013 when Nirvanix ran out of money. Nirvanix had a limited portfolio. It offered only storage as a service. Customers initially got two weeks' notice to transfer their data from the Nirvanix data centers to another location, but the time was later extended to 4 weeks. Some customers had 10 to 20 petabytes of data in use, which was be difficult to migrate within two weeks—especially if there is no plan and when data needs to be moved over limited bandwidth connections.

In April 2011, Iron Mountain, another cloud storage provider announced it would discontinue its cloud-based services. However, customers were given over 12 months to move data to another location.

One of the steps any organization that intends to use cloud-based services should take is to prepare their cloud exit strategy. However, less than half of the customers of cloud services actually have an exit strategy. Why is this? Probably it is not cool to have such a plan. The same applies for disaster recovery, backup verification, and so on. Those are all on a top priority to-do list but are the first items to be removed or demoted when time and or budget become issues. A cloud exit strategy should contain information that makes it clear when to exit and how to exit.

Organizations should think about when enough is enough and it is better to exit—think about maximum loss of data, maximum loss of availability, or lost revenue. These need to be monitored to be able to judge whether the service level is breached or not. Do not test only on outages but also on performance. However, in most SLAs of cloud service providers, there is no mention of a guarantee of performance.

Also, try to determine the future of the provider. What is the roadmap? Are new features being added at the same pace as other providers? How frequently does the provider publish press releases announcing these new features? If there is silence for many months, something could be wrong.

So if you decided to move out, there are two options: either back source (bring the workloads back to your on premise infrastructure, the opposite of outsourcing) or find another provider.

One of the advantages of using IaaS over PaaS and SaaS is that you probably do not have to perform any conversion of data. Your data is included in virtual disk files that also contain the application and the operating system. As long as your new cloud provider is able to host that type of virtual disk, file migration is not that difficult. The only challenge is moving the data out and into another location and changing network configuration.

Organizations considering cloud should make sure they are able to import and export data to cloud providers' data centers using external media such as a NAS or a bunch of USB drives. This is the only way to import of evacuate large amounts of data in an efficient and timely matter. Make sure the data on that external media is encrypted when unauthorized staff needs to have access to that media for import or export. Check for RAID compliancy when all the data is on USB drives and one fails.

Export and import operations of data would mean that virtual machines most likely need to be shut down during the export and import operation. It will be very difficult to perform some sort of virtual to virtual conversion between different cloud providers while workloads remain active.