There are two basic models for moving a data center: push/pull (physical move) and virtual/application migration.

Push/Pull (aka Lift and Shift)

Push/pull relocations generally take place with smaller data centers and are the riskier and less expensive of the two options. Less sophisticated than a migration, push/pull at its most basic level simply disconnects the data center components, ships them to the new location and reassembles them. Its most obvious downsides are that the data center isn’t in use during the move and there’s a large degree of unpredictability once the components are disassembled and reassembled, both of which can be major issues for the clients that depend on it.

To ensure a successful transition using the push/pull method, the data center staff needs to create a meticulous inventory of the components in each rack so they can be reinstalled in the same slots at the new location. It’s also critical to set up network connectivity before the move, as well as map the correct power assignments. After the move, it’s important to note power utilization once the new data center is online. Power should be balanced throughout the infrastructure from UPS to the rack level. Also, balancing airflow at the rack level is essential to ensure proper environmental conditions and to maximize efficiency.

During the physical move, several considerations need attention. If populated racks are being moved, it’s important to establish the rack’s static and dynamic load ratings. Dynamic load ratings are almost always lighter, and a rack structure is prone to failure if overloaded (its casters are usually the weakest part). Racks should be packed in shock pallets, then loaded into the moving trucks. Even small bumps and jostles can have a profound impact on equipment that’s never been moved before. Because data center components seldom move when in use, tip guards are often overlooked but should be used as well. Whether rolling around a facility or anchored in the truck, racks are liable to tip when in motion, and these inexpensive pieces of hardware can save thousands of dollars by preventing a rack from falling over.

Migration

A data center migration takes much longer and is much more meticulous than a push/pull move. There are no easy steps or short cuts. Unlike its opposite, a duplicate data center is created to maintain uptime during a migration. The computer environment of the production data center is duplicated and is tested at least three times for network connectivity and applications before the move can take place. Once the test results are satisfactory, the data center environment is frozen (figuratively speaking). The hardware environment, networking environment and computer environment will remain static until the move is complete. Updating the records and applications can occur after the move, and it’s important to avoid updating firmware at this point to prevent conflicts from occurring. Duplicating the data center to ensure continuous uptime and meticulously testing the environments are the major reasons why a migration is more time consuming and expensive than a push/pull move, but they also go far to ensure that the move will be complete with a minimum of problems for the data center tenants.

Once the move is finished, the equipment in the original production data center must be tended to. Disks should be wiped and storage arrays cleared of applications to maintain security. Next, some of the hardware, such as large server racks, can be sold for the value of the metal. More-sophisticated equipment—processors, storage arrays, networking equipment and structured cabling—can be refurbished and resold.

Best Practices

Whether employing a push/pull or migration, a few best practices are applicable to your data center move:

• Plan everything. The first step of any successful relocation is to make sure a delivery path is clear from the beginning of the move until every piece of equipment is plugged in at the destination.
• Employ a clear manager throughout the process. To ensure a smooth relocation, one person needs to ultimately be in charge to make the difficult decisions, answer questions from everyone involved and document the details during the process.
• Don’t make major changes on moving day unless absolutely necessary. The goal is to avoid chaos. It’s more important to take the time to plan all the details ahead of time to avoid making rash decisions with unknown implications once the move is under way.
• Track everything. You never want to be waiting on equipment or wondering where it is. The use of RFID for server racks and GPS on moving trucks will go far to maintain order once the move has started.
• Protect equipment using conditioned trucks. Almost every data center component benefits from being cool. Use air-conditioned trucks, even if moving a short distance, to help minimize equipment malfunction later.
• Make physical security a priority. Be aware of current situations. Conduct background investigations of all people involved and consider an armed escort and safe route mapping from pickup to the destination site.
• Use movers with experience in IT. It may sound obvious, but there’s a major difference between using an experienced firm and using general household or corporate movers. Firms that specialize in moving IT equipment may be a little more expensive, but they’re bargains compared with replacing damaged components. If large and/or heavy equipment is part of the move, you can also hire professional firms with riggers who specialize in moving delicate technology.

No matter which method you use, relocoating a data center is a complex process with many moving parts. Careful planning and preparation are paramount, as is investing in the best practices that will help ensure a successful conclusion. Once the new data center is up and running, you’ve already laid the foundation for the future when your company expands again.

About the Author

Gary Wong is the Director of Applications Engineering for Instor Solutions. He has more than 30 years of data center experience, including financial-sector and multitenant data center operations, and he has held positions in professional services, applications engineering, product management and sales engineering with Eaton, Schneider Electric and others. Gary’s expertise includes data center master planning, fit-out, product implementation, thermal analysis, airflow management and fire protection. Gary holds a BS degree in environmental engineering from Worcester Polytechnic Institute.