Make no mistake: Cloud is here to stay. Most companies now have at least some workloads delivered via a mixture of IaaS, PaaS and SaaS. Even when enterprises can exit from the cloud completely, it's a long and expensive exercise with dubious benefits.
However, organizations across all verticals and of all sizes are adjusting their cloud strategies and roadmaps. During the COVID-19 pandemic, organizations felt pressured and rushed to push workloads into clouds, especially networking functions. That pressure has lessened since the pandemic waned, and the rush has subsided. Now, some organizations have started to migrate workloads from cloud environments back to on-premises data centers.
Hefty cloud costs challenge organizations
Companies that push into cloud too quickly discover that moving networking workloads into IaaS and PaaS environments likely costs more than operating them in-house. On average, the cloud premium for running applications in cloud over on premises is around 24%, according to Nemertes Research.
Although it requires some effort, it is possible for enterprises to save on cloud costs. Nemertes found that the most successful organizations saved an average of 45% on costs. To spend less, organizations can take the following steps:
- Put only cloud-friendly networking workloads into clouds.
- Make workloads they want to migrate more cloud-friendly via refactoring, rearchitecting or replacing with new software.
Cloud repatriation can improve network performance
Companies also pull networking workloads out of cloud for performance reasons, such as to get the network out of a loop or to reduce latency due to distance. Enterprises might want to migrate out of cloud to reduce risk and bolster security by controlling access to a firewall or data set physically, for example.
If a cloud network has a different design than the data center network at the logical level, engineers need to review the effect the differences might have and decide how to mitigate them. When the workload in question is specific to networking, organizations need to make sure the workload is ready and able to participate in the on-premises architecture. Problems with a piece of network infrastructure could affect not just that system, but also everything else in the network.
Network cloud migration needs consideration
Before an organization decides to pull a workload out of a cloud environment and into a data center, they should consider several networking factors, including the following:
- IP address management. Ensure assigned IP addresses that work in the data center don't affect anything within the application moving on premises.
- Network access policies. Ensure the workload can still reach and communicate with the other systems it normally communicates with in its new location.
- User access policies. Ensure that all users who need access can reach the workload in its new location.
Data architecture is another important networking consideration for cloud migration. Network teams should understand from where an application receives and sends data. For example, if other systems that remain in the cloud send a great deal of data to migrated workloads, it could result in excessive data egress costs. If data sovereignty concerns triggered the move out of cloud, the same concerns could also apply to data that flows out of it to another system.
Performance considerations are other factors to evaluate. For example, if the goal of the move is to remove performance variability by getting off the internet and back onto a company WAN with minimal loss, latency and jitter, network teams should ensure the affected user and system populations are on the WAN and aren't mainly in internet-only branches or remote work environments.
Plan a cloud exit strategy with networking in mind
Network teams should have the tools to address most of these questions, especially those related to the volume and direction of data flows, as well as network performance. One area in which many organizations need to add some tools is dependency mapping. It can be a challenge to identify all the other systems with which any given system communicates. But, with careful planning and consideration, network teams should be able to move from cloud back to on premises without negative effects on cost, performance or security.