top of page

Avoid 8 Data Center Migration Mistakes

Data center migration mistakes are common. The reason for this is that data center migrations don’t happen every day. Data center staff might participate in one or two data center moves in their entire career, and some staff never experience a data center relocation.

Data center migration mistakes are easy to avoid if they are on the radar before the data center move occurs. Here are nine to avoid, with steps to take to ensure they don’t happen.

1. Incomplete Infrastructure Assessment

Data center staff must know all the details about everything running in their current data center. This includes knowing how many workloads are physical, how many workloads are virtual, how many workloads use local storage, and how many workloads use shared storage.

The data center migration project manager should know how many workloads will work the same way in the new data center and how many workloads will work differently (from physical to virtual, moved to shared storage, archived, etc.). This level of detail helps ensure that new infrastructure purchases and costs for the new data center are calculated accurately.

Infrastructure assessments should include rack-by-rack, unit-by-unit documentation for each device and its applications. To make things even easier for the data center move and for setup at the new data center facility, all operational and technical interdependencies should be added to the assessment. An example would be that application, web, and database servers for a related application would be moved as a bundle.

2. No Dedicated Data Center Migration Project Manager

A dedicated data center migration project manager is a necessity. All data center staff should know who that person is. The manager will have overall responsibility for the move, will authorize the next steps to take place, and communicate with data center staff during each phase of the relocation.

Not having a dedicated data center migration project manager could lead to such things as disagreements about whether the relocation can continue forward successfully. The net result would be unnecessary and costly downtime, which can be avoided if one person has the authority to make the call.

3. Unrecognized Dependencies

Recognizing dependencies is critical to a successful data center migration and can result in failures in the process. Dependencies can include something as simple as multiple redundant power supplies connected to the same uninterruptible power supply (UPS) or several applications using the same database. Knowing dependencies at all levels is key to planning a successful data center relocation.

For example, if the UPS is turned off before one of the power supplies is moved to a different power source, the server will fail. If data center staff believe the database is being used with just one application and they shut it down along with that one application, the rest of the applications using it will fail. This mistake can be avoided by generating a complete and up-to-date application dependency map.

4. No Step-by-Step Data Center Move Procedure

The data center migration project manager must create a detailed procedure for every step of the data center move. Nothing should be assumed or taken for granted. Data center relocations are often extended processes, and with detailed procedures to follow, and critical steps can be missed.

A step-by-step procedure should spell out exactly what to do from the first step in the data center move to the last step. It's critical to include definitions of what success is for each step. As data center staff become tired, especially during long moves, it can be tempting to take shortcuts just to finish faster.

A well-defined procedure that requires both data center staff and the data center migration project manager signoff for completion ensures that the data center relocation will be done correctly and successfully.

5. Inaccurate Time Estimate to Complete the Migration

A common problem among IT professionals is underestimating the time it will take to complete projects. This is because they’re assuming perfect-world scenarios instead of real-world scenarios. Data center staff and the data center migration project manager can adopt this kind of thinking as well.

In a perfect world, nothing goes wrong. No unforeseen problems occur. In the real world, though, things can and do go wrong. If adequate time for addressing these isn’t built into the time estimate for the data center move, then the amount of time estimated to complete it will be too low.

To avoid this, plan for variations that will require extra time to address during each step of the data center relocation. It’s better to assume too much time than to assume too little.

6. No Test Plan

Data center migrations can fail entirely if a test plan is not in place. The test plan should define in detail what success is for each step, how each step is validated, and who signs off on each test’s success. Having a test plan ensures accountability. When done right, when the new data center goes live everything will work as expected.

7. No Back Out Procedure

Data center migrations don't always happen. There are various reasons for this, but this possibility must be considered and included in the same procedure that governs the data center move. A back-out plan may seem like a lot of upfront work, but it will be invaluable if the data center relocation must be postponed or abandoned.

The data center migration project manager should detail the kinds of failures that will trigger the back-out plan and who has the authority to activate the back-out plan.

8. Wrong-Sized Server Racks

Server densities have increased dramatically. This will affect data center migrations. Before new high-density server racks are purchased for the new data center, verification must be done that the data center has the ceiling height to accommodate them, the floor strength to support them, the power density to sufficiently power them, and the cooling capacity to adequately cool them. Overlooking any of these factors could result in underutilized larger racks, which translates to wasted capital outlay or, worse, having to return them.

bottom of page