Choose by separating who owns and manages the hardware, who runs the facility, and how people need to access resources.
A hypothetical requirement for an existing server shows why colocation may be worth checking first—and which questions remain unanswered until the service terms are known.
Data center fundamentals series
What should you decide before comparing the three?
Start with the requirements, not the model names. An external facility for your own server and an online service running on provider-owned hardware solve different problems. Write down four answers before comparing them:
- Must the organization keep using hardware it already owns?
- Who can maintain and monitor the servers, including when something fails?
- Who must arrange and operate the space, power, cooling and physical access?
- Is physical access to equipment and a carrier connection required, or is access to on-demand computing resources enough?
The Fiber Optic Association — The FOA Reference For Fiber Optics - Data Centers -
IBM — What is Cloud Infrastructure? | IBM
IBM — What Is a Data Center? | IBM
Who owns the equipment, and who operates it?
In an own-facility, or on-premises, model, the organization keeps its IT infrastructure and data on site and is responsible for deployment, monitoring and management. It must also assign responsibility for facility needs such as power, cooling, physical security and communications connections. The label does not, by itself, say whether employees or contractors perform every task.
In traditional colocation, the customer owns and manages the equipment but leases space in another operator’s facility. The Fiber Optic Association describes colocation facilities with power, air conditioning, security and carrier access; that description does not establish what any particular site will provide.
Some colocation providers also offer optional management and monitoring. Do not assume that leasing space transfers server operations to the facility operator.
A managed data center is a different arrangement in IBM’s description: the customer leases dedicated server, storage and network hardware from the provider, which handles management and monitoring. That changes both the equipment and management boundary compared with traditional colocation.
In public cloud, the provider owns and manages the underlying hardware and infrastructure, while users access computing resources on demand over the internet. That is not a way to bring an existing customer-owned server into the provider’s facility.
Nor does cloud necessarily mean a shared virtual server: IBM describes cloud servers as physical, virtual or a combination, and offers a dedicated bare-metal service as one example. Whether a server is dedicated and whether the customer owns it are separate questions.
Specify what kind of cloud access is needed, too. IaaS provides resources such as servers, storage and networking; PaaS provides a platform for developing and running applications; SaaS provides ready-to-use applications. These categories help identify what is being requested, but do not settle every management duty in a particular service agreement.
IBM — IBM Cloud Bare Metal Servers
How does an existing-server requirement change the shortlist?
Consider a hypothetical organization that must keep using and directly managing servers it already owns. Assume it cannot readily provide on-site space and cooling, and it needs a carrier connection. This is a screening example, not an account of a real site or a measured capacity requirement.
Own facility
- What the requirement suggests
- Keeps the existing servers, but the assumed space and cooling problem still needs a solution.
- What remains to verify
- Who will provide and operate the required facility functions.
Colocation
- What the requirement suggests
- Allows the organization to investigate placing its own servers in leased space while retaining direct management.
- What remains to verify
- Whether the site can accept the equipment and provide the needed power, cooling, carrier connection, physical access and any optional management service.
Public cloud
- What the requirement suggests
- Offers resources on provider-owned hardware, not placement of those same customer-owned servers.
- What remains to verify
- Whether keeping the existing hardware is genuinely required, or the workload could instead use different resources.
Colocation is the first arrangement to investigate under these assumptions, not a confirmed fit. In particular, neither a general description of colocation nor the name of a facility establishes its available capacity, access rights or service commitments.
What if keeping the servers is optional?
Change the hypothetical requirement: suppose the organization does not need to own or bring in those servers, and on-demand access to computing resources matters more. Cloud then belongs on the shortlist.
The next questions are whether the workload needs virtual or dedicated physical servers, and whether it needs infrastructure resources, an application platform or a ready-to-use application. Those choices do not, on their own, establish price or suitability for the workload.
What must be confirmed before making the choice?
Once ownership and access requirements have narrowed the options, check each candidate’s written terms: facility services, carrier connections, who handles equipment during a failure, any optional management, and charges. Model definitions tell you which questions to ask; they are not promises made by an individual facility or cloud service.
Resolving whether the existing hardware must stay in use prevents a comparison between options that do not meet the same requirement.
Sources
- IBM — What Is a Data Center? | IBM
- IBM — What is Cloud Infrastructure? | IBM
- IBM — IBM Cloud Bare Metal Servers
- The Fiber Optic Association — The FOA Reference For Fiber Optics - Data Centers -
No comments:
Post a Comment