When data center hardware becomes outdated, there are a few options. Here are ways that admins can ensure that a migration is efficient and downtime is minimized.
A data center upgrade is like a home remodeling project; it requires planning, a budget and a certain finesse to ensure that existing infrastructure isn't damaged in the process. There's also the option to skip the remodel altogether and head to a condo with homeowners' association fees.
Admins must make a few important decisions before they decide to tackle a data center upgrade. It's an expensive, time-consuming and risky project -- and in the age of the cloud, it's sometimes unnecessary. If admins decide that upgrading the data center is the best route, they should know how to make the process a smooth one.
When to move to the cloud or colocation
The public cloud versus on-premises debate is never-ending, but the pro-cloud argument is growing stronger. A few years ago, on-premises data centers were necessary to maintain a level of security that cloud couldn't provide; but now, public cloud providers are offering higher security protocols to comply with stringent data protection laws.
Admins should also consider the price of upgrading. Expensive hardware might not be worth it if the organization is willing to sacrifice a level of control to a cloud provider. Plus, a cloud provider can offer more IT support and maintenance than a small to midsize organization can afford to keep in-house.
Admins shouldn't simply migrate to the cloud, however, without determining the necessary network bandwidth. Cloud providers need to share internet traffic with other tenants, which can put a strain on individual network resources.
Organizations should also consider colocation facilities to cut the costs of cooling equipment, staffing, hardware and maintenance. The issue of control, however, once again comes into play. IT has a support staff to call when downtime occurs, but they don't have control over how long it takes to remediate the issue. And if an organization didn't properly evaluate the service-level agreement with a colocation provider, then IT can stumble upon hidden fees or caveats when a problem occurs.
To avoid these problems, admins should determine which workloads -- if any -- belong in a colocation facility. They should also make a disaster recovery plan in case of an outage.
How to prep for a data center upgrade
If admins decide to keep workloads on premises and upgrade the hardware, they must ensure that downtime doesn't occur. Or, if it does, they must ensure that it doesn't affect mission-critical workloads.
Admins should also establish a workflow for the server upgrade process.
Traditionally, admins could let end users know in advance about an upgrade and take workloads offline during non-work hours. But this method isn't possible for organizations with remote employees or global organizations that span various time zones. In these cases, admins should move the affected workloads to a public cloud and reroute the traffic before starting the upgrade.
Admins should also perform pre-upgrade testing on new versions of software that they plan to use. Admins can perform the data center upgrade process in a lab environment to vet out bugs and compatibility issues. Service providers can help with this process for organizations that lack the necessary in-house resources.
Make upgrades carefully
Admins should replace data center hardware with redundancy to prevent data loss. For example, admins that need to replace a network switch should have a secondary switch in place before the replacement occurs to prevent connectivity loss.
Microsoft designed Windows servers as failover clusters that support rolling upgrades to address this situation. Admins can perform the upgrade process so that only one server node at a time is offline. However, admins should first ensure that the cluster they are trying to upgrade can run its workloads without a cluster node.
Admins should also establish a workflow for the server upgrade process. The most mission-critical workloads aren't necessarily the first ones they should migrate. It often makes more sense to migrate the least important workloads first to ensure that the migration techniques work properly.
