The Public Cloud is widely popular but can be very expensive. Rising costs, complex pricing models, and the inability to predict monthly bills are forcing many organizations to look for ways to save money. A recent article in The Register told the story of a SaaS provider that saved $1.2 million per year by repatriating some of its workloads from AWS to servers hosted at a colocation facility.
Companies may just be realizing that cost savings expected from moving to the Public Cloud are not happening. In fact, costs may be out of control. The key to saving money is to identify workloads that are less costly when placed in a colocation facility or at an on-premises data center versus those that are cost-effective and/or require the special properties of the Public Cloud. It sounds simple but it does take some analysis.
Where Public Cloud Is The Best Choice
Public Cloud is great for some workloads. Websites that experience rapidly changing user demand can take advantage of the elasticity and self-service provisioning capabilities of the Cloud to scale up or down as needed. Though most websites are good candidates for the Public Cloud, there are some exceptions. If your website primarily serves a local geographic area and if latency is a concern, then the Public Cloud may not be the best choice. Placing web servers in a local colocation data center will reduce latency and provide better response times for website visitors who are located nearby.
Software development teams frequently benefit from the flexibility of the Public Cloud. Compute, memory, and storage resources can be added as needed to help the DevOps team keep on schedule. QA can spin up test environments to perform continuous testing. Tests can be run against different IT configurations to predict performance. Once software has been completed, there may no longer be a need for the cloud resources used during development and testing. Savings can be achieved by periodically checking to see which Public Cloud resources are still needed after certain phases of development are completed.
Email servers are also a good choice for the Public Cloud. O365 and Google Workspace leverage the Public Cloud for their email and office application services. Clients like the universal accessibility and convenience of these SaaS offerings. Software updates and patches are automatically provisioned. Data backup is also included, at least for a modest period of time. However, clients requiring email backups beyond 90 days may need to set up an offsite data backup server at a colocation or on-premises facility.
In general, the Public Cloud is the best choice when workloads vary in terms of the compute and storage resources they require. Conversely, colocation can generate cost savings when the IT requirements of a particular workload are more predictable and constant.
Where Colocation Can Save Money
Database servers are usually the best place to look for savings from the Public Cloud. Not only are many databases stable in size, but Public Cloud companies usually charge data transfer (egress) fees every time data is exported from their storage systems to a device that is not in their cloud environment. CAPS has clients that have repatriated databases from the Public Cloud to achieve more than $100,000 a year in savings by hosting their database application at our colocation facility.
Other server types that can be placed cost-effectively at a colocation data center include Domain Name System (DNS) servers and Dynamic Host Configuration Protocol (DHCP) servers. These servers have modest and stable processing and storage requirements. It may be possible to host these types of workloads on older, lower performance servers to minimize the investment required.
It is possible to save money on spiraling Public Cloud expenses. Evaluate the types of workloads you are hosting in the Public Cloud to determine if you still need some of these Public Cloud resources and if repatriating these workloads can provide meaningful savings. Of course, the cost of hosting new workloads in both the Public Cloud and at a colocation data center or on-premises should be considered before deciding where they should reside.