Alan Zeichick | Developer Content Strategist | March 9, 2023
On-premises data centers create a burden on today’s companies. Whether a global business or a midsize entity with a single data processing center, the racks of spinning disks, humming fans, and blinking lights provide a drag on finances, a staffing headache in a tight labor market, a potential choke point for business agility, and a worrisome security liability. As a contrast, cloud computing has proven itself as offering all the benefits of a business’s on-premises data center, while reducing those drawbacks. The big question is: How to manage the transition, and in particular, close down that legacy data center with minimal business disruption?
A data center is a building that contains an organization’s computing resources, including hardware, such as servers, storage, and networking equipment, plus infrastructure for power delivery, cooling and heating equipment, and systems for data backups and disaster recovery. The networking includes high-capacity connections between devices in the data center, between the data center and the internet, and, if an organization has multiple data centers, high-capacity connections between the buildings. The data center’s servers drive public-facing operations, such as ecommerce and websites, and run internal applications, such as enterprise resource planning (ERP), customer relationship management (CRM), human resources, email, and production workflow.
Data centers can also be designed to run highly specialized, industry-focused applications, such as high-speed financial trading systems; engineering tools, such as computational fluid dynamics used in automotive or aerospace; or sophisticated modeling and simulation frameworks used in pharmaceutical drug discovery. A data center is built for physical security to prevent intruders from entering to tamper with the hardware and sophisticated cybersecurity to protect against hacking.
A large organization might have several data centers designed to balance workloads, improve performance, minimize outages in the event of disasters, and comply with data residency and other compliance requirements. Smaller organizations might have a single data center in their headquarters building or located offsite in a colocation or “colo” facility. A data center may have thousands of servers, plus the requisite racks, wiring, networking gear, hard drives for storage, backup storage resources, power generators, cooling towers, and administrative workspaces.
An organization can’t expect to set up a data center once and enjoy its benefits for years to come without further investments. Like any physical structure, a data center requires constant attention, maintenance, and improvements, all of which can be expensive and require expert staffing. Data center maintenance can become more difficult over time, as the facility ages and its infrastructure and equipment cease being state of the art. The organization either needs to handle these challenges internally or hire service providers to manage them. By contrast, in a cloud computer scenario, many of these challenges fall onto the cloud provider; the customer can simply use the service, which they get in a pay-for-use model.
Inside a data center are hundreds to tens of thousands of computing devices, each with hard drives, power supplies, fans, and other parts that inevitably break. They also have the near certainty of being obsolete within a few years. While there is always some excess capacity and “hot spares” that can be automatically used in the event of failure, staff are kept busy replacing disks and power supplies, diagnosing intermittent failures, making upgrades, applying security and performance patches, swapping components, reconfiguring gear, and otherwise performing maintenance. In addition to the computing hardware, a data center owner needs to maintain the physical building, which needs sophisticated physical security controls, reliable electricity distribution, heating and cooling, cabling, fire suppression, internet and internal network connectivity, and more.
Patching servers and routers against new security vulnerabilities is an essential cybersecurity requirement, but many organizations’ data center staff struggle to keep up with those patches, leaving systems vulnerable to attack. Several widely publicized customer data breaches have been blamed on data center systems that were left vulnerable despite patches being available.
Over time, the computing requirements of a healthy organization will almost certainly increase, and this requires the acquisition and installation of entirely new racks of hardware, replacing existing hardware with more powerful equipment. The process of specifying new equipment, ordering, building, shipping, installing provisioning, and integrating that equipment, can take weeks or months—perhaps even longer. This precludes rapid action, unless the data center already has sufficient excess capacity to handle anticipated future needs. If the computing needs grow beyond the physical size of a data center, organizations face a major real estate and capital investment decision.
Once hardware and other infrastructure have been installed in a data center, it is difficult—and likely expensive—to get rid of it, especially as computing needs change over time. Scaling down can also be an issue in the short term; for example, if server capacity is required to handle the loads of seasonal transactions, what do you do with that equipment the rest of the year? (Fun fact: Some of the earliest online services began as far back as the 1980s as a way for companies to monetize their unused excess computing capacity.)
The good news is that a data center’s physical security needs normally change slowly: Perhaps keyed locks were replaced by keypads, and keypads by smart badges, and smart badges by retinal scans or fingerprints. The fast-moving challenges are in digital security, whether the cybercriminals who infiltrated an employee’s laptop via phishing are executing a brute-force attack against a website or leveraging a newly found exploit against server operating systems or device firmware. Data centers require constant security monitoring, as well as new tools to protect against fast-evolving threats.
Despite the latest in tech automation, data centers still need people. From round-the-clock staffing for a security operations center (SOC) to specialists who design upgrades to the data center’s systems, a data center requires skilled technologists. In some cases, those are employees; in others, they may be contractors or work for a managed service provider. Hiring, retaining, and training such talent is expensive, whether for the organization or the service provider. In some markets, organizations have trouble finding sufficient qualified candidates.
Data centers have served enterprises well for decades, providing the computing infrastructure to enable the tremendous growth of online services, drive automation and worker productivity, and spur digital transformation. Compared to cloud computing, however, data centers are legacy systems that increasingly look like a liability. There are many reasons to plan data center exit strategies. Those reasons include:
Technological advances and business opportunities may convince you it’s time to migrate from a physical data center into the cloud. But the actual timeline may hinge on non-tech issues—such as the lease on a data center facility. Such leases are long term, often five to twenty years. The end of that lease provides an opportunity for cost savings. The process of exiting a leased data center may take a year or more, but the lease date can provide an important incentive to revisit hardware purchase plans, maintenance contracts, service provider agreements, depreciation schedules, and staffing plans, and map a plan to the cloud.
Organizations grow and evolve over time, organically and through acquisitions. It’s not uncommon for there to be multiple data centers, each serving a different business function, sometimes spread across the country or the world. Acquisition legacies can lead to different facilities hosting separate enterprise resource planning (ERP) applications or database servers, for example, or with one facility housing legacy ecommerce transaction systems while another is dedicated to a new system for inventory management. Such sprawl is a missed opportunity to benefit from economies of scale. But it’s often technologically or operationally untenable to consolidate those multiple data centers into a small number of owned facilities. However, moving one data center into the cloud, and then moving another, may accomplish the same goal with great savings and without the upheaval of a physical move or the need to procure larger facilities for consolidation.
Hardware has a limited life in the data center world, and that’s often due to obsolescence rather than outright failure. The anticipated or planned end of life for a data center’s equipment can provide a timeline or opportunity for a migration. What causes obsolescence? Sometimes the systems can’t handle new usage paradigms, such as routers that aren’t equipped to deliver edge computing services. Sometimes the equipment is inefficient compared to new models, such as database servers with lower-capacity storage or which use slower rotating hard drives instead of faster solid-state drives (SSDs). Perhaps the computing architecture has changed: An organization might be moving away from x86-based servers to ARM-based devices. There may also be longevity concerns around security, energy efficiency, compliance, disaster preparedness, or even the condition of the building itself.
What is the lifespan of the equipment in a data center? For the major facility infrastructure, such as well-maintained heating, cooling, power, and security systems, it might be a decade or longer between significant upgrades. For individual equipment in racks, a good rule of thumb is three to five years before replacement with new, improved gear.
When gear’s time is up, replacing racks of old hardware can be expensive, labor intensive, and disruptive to business operations. Organizations facing a major hardware refresh due to end-of-life equipment may find the cloud migration more appealing and cost effective. This cutoff point also provides a natural schedule to aid with planning.
The easiest and best way to overcome the challenges of operating a data center environment is to exit that data center entirely—and migrate its applications, data, and services to the cloud. With a cloud migration, capital expenses (CapEx) become manageable operating expenses (OpEx). Your architects, engineers, developers, and administrators can focus on the parts that matter: delivering services, solving business problems, and reacting to business opportunities.
The parts of data center operations that don’t provide a competitive advantage for you—providing electricity, providing cooling, providing bandwidth, stringing cables, replacing burned-out parts, diagnosing intermittent networking issues, coping with denial-of-service attacks against the internet router, planning for flood or fires, hiring hardware technicians and security guards—aren’t your problem. Those responsibilities are borne by the cloud provider, which has economies of scale beyond all but the very largest enterprises.
What’s more, because cloud providers are inherently flexible, your systems administrators can spin up new servers or other capabilities in minutes instead of months. When peak-capacity services aren’t needed, they can be released (along with their costs) until they are needed again. When services are required for a limited time, such as during end-of-year financial closes or seasonal events, they can be added for a few days, weeks, or months without requiring extra hardware purchases.
Exiting a data can be more complex than, say, moving offices. It’s vital to avoid any disruption to services provided to customers, partners, and employees. Doing so requires knowing exactly what services your data center currently provides and how it provides them.
For each of those services, your organization will have to make key decisions: Move them as they are into the cloud (sometimes called “lift and shift”); move them into cloud native infrastructure, such as from an on-premises database system to a cloud database; or redesign those applications entirely to use new service-oriented paradigms that are more scalable, but which may take longer to create. With applications, such as ERP and human capital management (HCM), you can also replace conventional on-premises apps with cloud-based software as a service. With those software-as-a-service (SaaS) apps, you don’t need a data center because the entire application is in the cloud and can be accessed by customers or employees directly via a browser.
To make a retreat from the data center easier, here are eight key steps to take during the process.
You can’t move what you don’t know about. This means cataloging the hardware, software, and services, including external service providers. Know the contracts. Know where data is stored; what format it’s in; where the backups are; what dependencies systems have on each other; and if there are specific concerns around areas, such as compliance or data residency. Automated asset discovery tools will help with this process. There will be surprises.
Now is the time to scrutinize leases, depreciation schedules, service provider contracts, and licenses. Some of the existing providers may move with you into the cloud; others may not be able to, or you may not want to bring them with you.
If this hasn’t been done previously, make sure everyone in your organization is onboard and aligned with the new strategy. This is likely a board-level decision.
Look at each of the applications and determine what the cloud architecture will look like. What about data: Will the format or structure of the data need to be modified to fit new applications or storage paradigms, and how will that get done? How will data be moved? Which systems need access to which other systems, and how will that be handled in the cloud? How will end users access applications, data, and services? What will the security architecture look like?
Not all your staff’s skills will be applicable to the new paradigm, and not all employees will be interested in a new model. However, many of your employees will be excited about the potential for working with cloud services and could be retrained to gain the required expertise and certification.
Cloud providers and third-party vendors offer a wide range of tools, from automated software systems to sophisticated consulting, to assist with the migration and data center exit. While your organization knows its assets and needs, cloud providers and specialists have experience with migrations. Be sure to leverage that expertise.
You can’t move everything at once, or at least you shouldn't. Start with the easiest, most-siloed applications, data, and services to gain confidence, experience, and quick wins for your team and employees. Phased migration may be tricky in that some systems will be in the cloud, and others will still be in the data center for weeks or months—and yet they must interoperate seamlessly even while different parts are migrated. The extra work of making sure nothing breaks during these transitional phases will provide resilience now and in the future.
Once the full migration is complete, shut down the old data center, dispose of any remaining assets, and look into the bright new future.
On-premises or colocated data centers are legacies of an older form of computing, and it’s time to move beyond them to the cloud. While company-owned data centers helped organizations enter the initial era of computer-fueled automation and data-driven decision-making, data centers now are expensive to operate, require significant ongoing maintenance, need constant upgrades, are inflexible when it comes to meeting challenges, and can hold organizations back from seizing opportunities. The process of exiting a data center will take months, and the time to begin planning is now—before too much more money is spent on maintaining and upgrading those existing facilities.
There are many reasons to exit your data center, including cost savings, agility, scalability, reliability, and security. On all these fronts, Oracle Cloud Infrastructure (OCI) has delivered results for companies, letting them completely exit their data centers or significantly scale down their data center footprint. OCI’s second-generation cloud is designed for modern applications and offers deployment models for public and private clouds, including cloud services that reside inside a company’s own facilities. OCI also provides a wide range of capabilities, including multicloud integrations and high performance computing for high-volume AI workloads.
What is the first step in exiting a data center?
When you are planning to exit a data center, conduct a thorough survey of applications, data, services, users, and security requirements. Everything on that survey will require a migration plan, whether it’s to “lift and shift” the existing applications and data into the cloud, choose new applications, or build new applications from scratch.
What is the lifespan of equipment in a data center?
Major parts of data center infrastructure, such as HVAC (heating, ventilation, and air conditioning) systems, power distribution, and physical security systems, could last a decade or longer with regular maintenance. The computation equipment, such as servers, routers, switches, and storage, are good for three to five years, as a rule of thumb, before becoming obsolete.
Who is responsible for security in the cloud?
Physical security of the cloud infrastructure—the servers, network infrastructure, and so on—is managed by the cloud providers. Responsibility for securing the software and services is shared between the cloud provider and the enterprise.
How long does it take to exit a data center?
Plan on a full data center exit to take months. A larger IT infrastructure could take years. It all depends on the size of the data center, its complexity, and the amount of data. Much of that time will be consumed by taking a thorough inventory, developing plans, creating and testing new software (if required), and training. Like with moving offices, the actual migration and exit itself is a relatively short phase, once all the planning is complete.
Learn more about how OCI's approach creates a differentiated cloud experience and delivers on the promise of modern cloud economics.