Many enterprises are migrating workloads away from on-premises data center occupancy to a combined cloud and colocation deployment often termed “hybrid cloud.” This requires careful planning to insure efficient, reliable, and flexible platform adoptions supporting future IT goals.
Many enterprises have aging on-premises data centers with major critical systems nearing end of life, and some owners won’t approve the sizable capital investments required to change that. Users have also realized that small and mid-sized self-operated data centers (less than 2 MW) aren’t as economical or environmentally friendly as the 20- to 50-MW data centers being built at scale by large colocation and cloud providers.
For these reasons, and to benefit from modern cloud and colocation benefits, many enterprises are migrating workloads to a combination of public cloud and colo, often with a “cloud first” directive from corporate IT leadership. However, deciding which applications should go to the cloud and which to colocation is complex. AFCOM’s 2022 State of the Data Center Report, written by Bill Kleyman, executive vice president of digital solutions at Switch, revealed the industry is going through a cloud and data center rebalancing, indicating that 87% of respondents either have a hybrid model today or will have one within the next 36 months.
![cloud and colocation data centers](/ext/resources/Issues/2023/02-February/Killian-hero-slide1.jpg)
Image courtesy of Kirk Killian
What goes where?
Many mature enterprises are evaluating the applications and data sets that could be easily placed in a public cloud — or easily updated for cloud suitability — and those that work better in colocation, essentially asking the question, “What goes where?”
Cloud-native applications written specifically for cloud deployment migrate well into a public cloud, taking advantage of cloud scalability and multifunctional cloud features, like AI and big data mining, which efficiently use data in different ways once it is captured within a public cloud. However, “cloud native” is a nebulous term — ask 10 industry experts for the definition of cloud native, and they’ll give 10 different responses.
Existing applications with usage that varies significantly over time (think retailer websites during the holidays or concert tour ticket sales) are also good candidates for public cloud deployment, where scalability — both up and down — are cloud hallmarks. Containerized workloads are excellent candidates for migration into the cloud, between public clouds, and from public cloud back to colocation or on-premises. The cloud also works well for software development and testing, where projects can spin up and down as needed, and disaster recovery functions that aren’t location-specific.
Unfortunately, many custom-coded and legacy IT applications that existed well before the modern public-cloud era, including COBOL and mainframe programs, do not migrate easily or efficiently into public cloud platforms. Some legacy programs can be rewritten for the cloud with significant investment and effort, and enterprises are therefore choosing to move them to colocation instead. Other existing applications face performance degradation or skyrocketing, unbudgeted, variable costs when placed in a public cloud. High-performance computing with user-customized hardware clusters is also typically a poor match for public cloud environments. Further complicating this analysis, many IT leaders anticipate moving applications and data back and forth between cloud and colocation in coming years.
![Application development and cloud migration](/ext/resources/Issues/2023/02-February/Killian-hero-slide2.jpg)
Photo by Jason Goodman on Unsplash
Start with the end in mind
For many users, the first step in developing a hybrid cloud and colo deployment plan is forecasting application needs over both short- and intermediate-term horizons. This scoping analysis includes quantification of current and planned future applications and data sets along with identifying the required hardware and related critical power requirements. While each enterprise’s applications are unique to its specific IT architecture and business, many applications fall into two broad categories: those that tend to operate better in cloud and those that tend to operate better in colocation.
Other issues shape the decision process between cloud and colocation, including deployment timing; costs; audit/compliance; staff skill sets; environmental, social, and governance (ESG) goals; and geographic factors. However, each enterprise prioritizes these qualitative issues differently.
Deployment timing
The cloud can offer a significant deployment timing advantage for many applications. Cloud provisioning is typically rapid (for “cloud ready” applications), but colocation deployment can take longer, especially given supply chain shortages affecting hardware availability and colocation suite customizations.
Unbudgeted costs
While many enterprises have enjoyed cost savings from cloud adoption for some applications, other applications face wild and unbudgeted cost variances. Some enterprises prefer predictable platform costs in a stable colocation environment compared with fluctuating costs that attract negative management scrutiny.
Repatriation
An underpublicized downside to the public cloud is repatriation fees, a surcharge imposed when users move applications and data sets from the public cloud back to on-premises facilities or colos, even for justifiable reasons, including inferior performance, excessive cost, or audit and compliance issues. The 2022 Cloud Protection Trends Report by VEEAM indicates that 89% of users intend to move some cloud workloads back into controlled environments.
![Colocation data centers](/ext/resources/Issues/2023/02-February/Killian-hero-slide3.jpg)
Image by Creative Art
Security and business continuity
Some enterprises incorrectly assume that placing workloads in the public cloud inherently solves security and business continuity concerns, but proper security and continuity planning are still required for both cloud and colocation platforms. IBM’s 2022 Cost of a Data Breach Report indicates that 43% of organizations were only in early stages or had not yet started applying security practices to safeguard their cloud environments. That same report revealed that 45% of organizational data breaches occur in the cloud. In The Uptime Institute’s 2022 Global Data Center Survey, 35% of respondents indicated significant performance issues or downtime for applications in a single public cloud availability zone. Cloud providers offer multiple availability zone solutions but at additional cost.
Latency and edge
Applications required to run in specific locations for regulatory, telecom, or business continuity reasons find colocation very attractive. Colocation providers typically offer effective solutions addressing telecom latency concerns, such as a paired “active-active” site topology in almost all major locations. In other cases, applications may require deployment in edge markets to be near customers or other business partners. AFCOM’s 2022 State of the Data Center Report revealed that 83% of respondents have implemented or plan to implement edge placement in the next three years, and the number of locations with colocation options dwarfs the number of locations with public cloud availability zones or local zones.
![Existing applications](/ext/resources/Issues/2023/02-February/Killian-hero-slide4.jpg)
Photo by Radowan Nakif Rehan on Unsplash
Regulators playing catch up
Heavily regulated industries may find that some of their applications don’t yet migrate well to public cloud because regulators are still adjusting to approving applications hosted outside a physically controlled environment. Enterprises requiring physical control of their hardware often prefer a colocation environment where security, audit, and compliance procedures are well-documented, especially where a user can lease an entire colocation suite mirroring the physical control of a traditional on-premises facility.
Staffing needs
Enterprise IT staff skill sets that have worked well for on-premises and colocation may not be ideal for cloud adoption, requiring retraining or staff replacement to meet evolving cloud deployment initiatives. IT staff replacements are often risky to overall effectiveness and application uptime and expensive in an inflation-intensive labor marketplace. Unfortunately, multi-decade IT employees take valuable experience and knowledge with them if terminated.
Successful cloud adoption can require comprehensive IT organizational changes.
"We see many companies that adopt the public cloud with the hope of increasing the pace of innovation, releasing more products to market, and scaling effortlessly” said Peter Roosakos, CTO of Foghorn Consulting. “But if IT governance, tooling, and processes are not transformed along with infrastructure, companies find that their investment does not yield the expected benefits. Foghorn recommends that companies plan appropriately to ensure modern management techniques accompany modern infrastructure. Key focus areas include infrastructure as code, automated provisioning and deployment, FinOps [cost reporting, allocation, and control], and shifting security left in the development process."
Make it greener
Many enterprises have ambitious ESG goals. Because data centers consume a huge quantity of electricity compared with other operational segments, they also attract disproportionate scrutiny. Public cloud providers have been leaders in promoting energy and resource efficiency, but their general opacity on specific facility placement has made this benefit less traceable than colocation options that provide detailed and auditable statistics at the facility level for meeting corporate ESG goals. Most cloud and colocation facilities are environmentally friendlier than legacy on-premises data centers.
![cloud or colocation](/ext/resources/Issues/2023/02-February/Killian-hero-slide5.jpg)
Image courtesy of Kirk Killian
Colocation flexibility
Colocation providers have tailored their services to users executing a hybrid cloud strategy, seeking a shared slice of the application placement pie.
“Enterprise requirements for colocation and cloud are rapidly increasing as digital transformation reshapes the global economy,” said David Stinson, executive vice president of enterprise sales for QTS Data Centers. “Traditional 500-kW enterprise deployments have expanded into multi-megawatt environments in just the last few years. As data center demand increases, enterprises must seek a data center partner that can offer high-availability, low-latency network ecosystems managed via advanced service delivery platforms.”
Private cloud
Many colocation providers now offer a private cloud as an attractive bridge to enterprises moving toward the public cloud, which are also facing regulatory hurdles or technology, like telecom latency. In this option, users migrate a portion of their applications to the colo provider’s private cloud within the same building as their colo suite. Later, after the operational kinks have been worked out for each application within the private cloud deployment, the user can evaluate moving that portion of the workload into a public cloud or full colo environment. Colo providers have generally been flexible with spend-shift contract policies.
Hybrid cloud and colo mega-campuses
In prominent locations, like Northern Virginia, Chicago, and San Jose, public clouds operate on the same huge campuses as colocation providers. Increasingly, large enterprises are seeking to deploy a mix of colo and cloud in close proximity using low-latency circuits to efficiently and cost-effectively move data back and forth.
Navigating the hybrid cloud and colocation path can be challenging, so many enterprises engage an advisor to assist in scoping, planning, and executing a comprehensive deployment plan. Experienced advisors assist in selecting solutions and negotiating procurement contracts, usually delivering net cost savings far exceeding any professional services fees.
Some IT architecture experts feel the industry is still in the early innings of public cloud adoption, but it's clearly a great platform solution for many applications. The challenge for most enterprises lies in evaluating and testing applications for cloud suitability while flexibly supporting placement among both the cloud and colocation. The key strategy question for the next few years appears not to be “cloud or colo?” but, instead, “what goes where?” while optimally using both.