For any organization, data center migration can be tricky. But that's especially true for the do-it-yourselfers at small- to medium-sized businesses (SMBs); they often don't have the resources to involve high-priced specialists. Instead they need to do the best with the resources they have.
Prather said the common challenge in all of the moves is the unknown. "That's why it is so important to plan, plan, plan and then plan some more," he said. In one "worst case," Prather said the bankruptcy of a co-location facility forced an unplanned move of one of his data centers within a matter of days. Thanks to good luck in immediately locating an alternate location, the move succeeded. However, he noted, "It took three months before things were really back to normal." More recently, Prather's data center moves have been the result of more leisurely and thorough planning as well as the ability to prepare a whole new facility and then simply move the data, to minimize the impact and reduce the possibility of outages.
Prather said the first step should always be to be clear about your bandwidth needs and then talk to your telecommunication suppliers. "Moving telecom assets might take as little as a few weeks but for higher capacity systems you might need three or four months of lead time -- particularly in rural areas," he said.
Data center migration checklist
Prather's data center migration checklist of preparation steps includes:
- Review support contracts to make sure they will still be valid in the new location.
- Determine who will do the physical moving and reconfiguration: your own staff, a consultant or others.
- Review service-level agreements (SLAs) with outside customers (if any), particularly with regard to liability for any customer supplied or customer owned equipment.
- Document and label everything.
- If possible, consider simplifying your infrastructure well in advance of a move.
- Be aware of system interdependencies and plan the sequence of the move accordingly.
- If possible, set up a duplicate facility in the new location rather than moving and setting up the same equipment again.
- Be aware of and try to eliminate any single points of failure in both old and new locations.
- Develop and use redundancy. For example, a disaster recovery (DR) location where you can keep critical apps and provide failover.
- Consider moving to storage or server virtualization -- or both. "One could set up a virtual platform including storage, virtualize the physical servers and move the virtual platform to the new location while keeping the physical setup at the old location until the virtual platform is up at the new location," said Prather.
- Back up everything in preparation for the move. "Hard drives are notorious for failing when they have been shut down, cooled, moved and restarted," he said.
In the final analysis, said Prather, moving is a lot like preparing a disaster recovery and business continuity (BC) plan -- you need to identify your most critical applications and have contingency plans in place.
While suggesting a broadly similar approach, Paul Taylor, an IT architect at WSP Group Plc in London, recommended that the planning process should be divided into two phases. The first phase is to determine what is wanted from the new data center. For example, this might mean developing a spec sheet for the new location or ensuring that a chosen location can meet requirements (i.e., space, power, security, other facilities, growth capacity and location). The second phase is to execute the move.
"Some years ago I moved our two-racks worth of kit between floors in the same data center ... [and] it was too hard to get all the communications suppliers to turn up on the same Saturday to move everything at once," he said.
In another move, last year, Taylor set up a new data center to supplement the existing London data center. The main criteria turned out to be how much power was available per rack and getting communication circuits installed. "The project was also badly tangled up with our head office relocation and a new WAN, so circuit delivery was a real problem," he added.
Moving toward a greener data center
Finally, there is the experience of Steve Kolbe, President of ANALYSYS, a Baltimore-Washington-based information technology services company. Their most recent data center move, concluded in April, aimed in part to achieve a goal of "carbon neutrality" through a transition to wind power and purchase of Renewable Energy Credits.
Kolbe and his team decided to accomplish the move themselves. They began planning in late December 2008, and completed the implementation in mid-April 2009. The move took about 40 hours of work, about half which was completed outside of business hours. Due to the design of the system, the process was mostly a matter of provisioning the new virtual systems in the data center, transferring data, switching over production and decommissioning the old systems.
Of course, moves aren't always that smooth, which begs the question of whether to do it yourself or hire someone else. Kolbe said it depends on many factors. "If you hire an outside consultant, be sure they have experience in this type of work and be sure you know how equipment will be transported and by whom," he said.
For his part, Prather said hiring outside help may be a must if you have a small staff. However, he said, at a minimum, "You should make sure preparing for the move and coordinating it is the full-time responsibility of at least one person." Otherwise, you are looking at trouble. "Moving is a challenge but if you properly plan and prepare, it's manageable," he added.
For those embarking on a data center move, there are a few tools that may help. For example, data migration can be assisted by product such as Brocade's Data Migration Manager (DMM) EMC Corp.'s Rainfinity, (a NAS virtualization appliance that can be deployed for migrating data), Hitachi Data Systems' (HDS) Universal Replicator and open-source rsync. In addition, Shunra Software offers Shunra VE, a tool that helps predict the impact of moving or reconfiguring a data center on application performance, backups and other aspects of operations.
About this author: Alan Earls is a Boston-area freelance writer focused on business and technology, particularly data storage.