In the past two years, I have had the pleasure of scoping and executing more than 130 data center relocations. During this time, I have participated in every facet of the relocation life cycle, including application and business process discovery, application bundling, scheduling, logistical planning, physical move execution, and vendor recertification. Regardless of where one is within this life cycle, hidden and often unforeseen gremlins lay waiting for the opportunity to sabotage your move.
OK, well, maybe it’s not that dramatic, but the result of these gremlins may be equally as fatal to the success of your data center move. Understanding the risks of any data center relocation event, anticipating them, and ultimately planning for contingencies are the best complements to even the most well-planned relocation event.
Planning relocation: everyone is a stakeholder
More often than not, I have observed reluctance on the part of organizations when it comes to investing in a proper planning regimen before physically executing a move. Across the last 100+ relocations I have managed, no less than 60 to 70 percent of the total relocation cost was attributed to planning.
Frankly, physical logistics involves both art and science; planning is all science. You must know the implication of removing any cable, network device, server, storage volume, database, application, service or person within your core business during your move.
Planning a data center move often begins within the IT or facilities department. In recent years, IT growth and the significant increase in power and heat compaction have stressed data centers to the point of exhaustion when it comes to power and cooling. Many data centers have yet to reach physical capacity before the supporting environmental systems reach maximum capacity.
This reality is usually the basis for a facilities-driven need to relocate to an area that can better accommodate the power and cooling demands-and often comes to the surprise of the IT department. Because of the delicate nature of the service provider relationship that IT has with both the facilities department and the business units, a successful relocation project must have stakeholder representation from each of these three areas of the organization.
In even small and midsized data centers such as those with fewer than 100 racks, IT assets are often placed in a seemingly random manner. At one point there may have been an attempt to group assets logically (by application, department or function, for example) but, over time, gating factors such as power density, form factor and other considerations resulted in assets being placed wherever they would fit.
Consequently, the physical plants resemble an interwoven fabric of delicate communication paths, risking service disruption if any one of these paths is disturbed. This is metaphorically consistent with the integration of workflows, services and applications supported by the physical infrastructure. Because of this, data center migration planners must rely heavily on the quality of the documentation about the application environment, as well as the tribal (and often undocumented) knowledge within the minds of the system designers, administrators and owners.
Logic, then Physics
Logic, then physics
Relocations can be achieved in a number of ways and often employ a combination of the following three methods:
Method No. 1: The lift and shift
The simplest, the lift and shift, involves taking a verified, successful backup of a system, powering it down, moving it and powering it back up.
Method No. 2: The swing move
Another, more complex method is the swing move. This method entails setting up temporary systems at the target site and replicating data to those systems in order to shift an application or service to the target site quickly-and then powering down and relocating the equipment from the source site.
The temporary equipment is retired once the service or application is again running on its original equipment. This method is commonly used when the time it takes to physically relocate a system exceeds the organization’s tolerance for downtime of the application or service.
Method No. 3: The logical move
Another method that is gaining widespread popularity is the logical move, which does not involve physically relocating any assets. Logical moves are used for existing virtual machines or as an opportunity to migrate physical systems to virtual platforms. Many organizations find that data center relocation creates opportunities to gain increased efficiencies such as those that come from consolidating physical systems.
Moves of this type involve setting up platforms to host VMs at the target data center, performing Physical-to-Virtual (P2V) migrations at the source site, and transferring those virtual instances to the target site over high-speed links. The VMs are then started at the target site and users are pointed to the applications and services running at the target site.
Regardless of the method employed, an organization must always go through the exercise of defining the logical components of each service and application. It must map those components back to physical devices. This process, commonly referred to as application bundling, allows the relocation planners to develop a picture and logical sequence of events that must take place in order to move an application or service. It also helps to flesh out the logical and physical dependencies that applications and services have on one another.
Preparing the Target Site
Preparing the target site
Many organizations will find that some portion of existing assets is not worth moving. This may be because the assets have reached end-of-life (EOL), are being virtualized or are otherwise decommissioned. Regardless of the reason, any change in the number or type of assets that will reside in the target site will have an effect on the physical and environmental requirements of the target site.
The bundling exercise will indicate what needs to move, what doesn’t, and what accommodations may need to be made for swing equipment. From this, a physical space plan can be derived and the necessary calculations for space, power and cooling can be performed.
Since most equipment racks are designed to be stationary, the systems must be removed from the racks prior to physically relocating them. As previously mentioned, a bundle may be composed of systems from multiple racks at the source data center. This means that there may not be a free rack available to move with the systems of a given bundle. For this and other reasons, most relocations will require some new racks to be predeployed at the target site in order to receive incoming systems from the source.
Some organizations may opt to leave all racks behind at the source and deploy a uniform system of new racks at the target. This offers many advantages that may outweigh the costs such as standardization, the ability to easily integrate environmental sensors or rack-level security, and a uniform aesthetic in the target site.
When specifying the requirements for the target site, it is important to keep in mind the opportunity for improvements and the ability to correct shortcomings that developed over time and as a necessity to sustain growth in the source site.
One area in particular is the cable plant (OSI Layer 1); data center relocation gives an organization the option to hit the reset button on how they deploy and manage Layer 1. Companies must pay special attention to prepatching each rack with the proper color coding and labeling scheme prior to relocating systems to the target. Having each rack prepatched will save a considerable amount of time and minimize troubleshooting headaches on move days when time is precious.
Execution: Scheduling and Logistics
Execution: scheduling and logistics
Compared to planning, move execution can be surprisingly less painful, especially with the right logistics partner. While it is possible for an organization to perform the necessary discovery, arrive at a solid set of move bundles and develop a timeline, few organizations specialize in the complexities of scheduling, managing and coordinating the logistics of data center relocation.
Scheduling requires tight coordination with business units and application owners. The schedule must be governed by the organization’s tolerance for downtime of the application or service that is moving.
Downtime begins when the application or database is taken offline-and includes the time necessary to relocate assets, perform and verify a backup, replicate data (when required), power down the system, pack and transport, rerack and reinitialize. Within this timeline, a rollback plan must also be factored.
Some platform vendors may require you to have vendor assistance with powering down and reinitializing a system. This fee-for-service arrangement-often referred to as recertification-is required so that there is no lapse in warranty and maintenance. Many equipment vendors will offer relocation services, which include the recertification service.
However, it behooves an organization to compare the costs of an OEM move versus a non-OEM move with recertification services taking place after the fact. When using a proven relocation partner in conjunction with an OEM for recertification, risk is appropriately mitigated and the costs are commonly reduced.
Insured and secure transportation of assets is also critical to risk mitigation. Data center assets differ from furniture in the way they must be handled, packed, secured and shielded from electrostatic and electromagnetic damage. As such, only qualified personnel and the appropriate packing materials should be used when transporting data center equipment.
Any party transporting data center assets must provide full replacement value insurance for theft, damage, or loss. Plus, this insurance must be applied to each conveyance (that is, each vehicle or vessel transporting equipment) as opposed to tying the insurance to the event itself, which may be comprised of multiple conveyances.
The industry average default insurance is approximately 60 cents per pound of cargo, which won’t cover the loss of any data center asset. Ensure that your carrier or relocation partner can provide conveyance-based insurance equal to or greater than the value of the conveyed assets.
Data Center Relocations Today
Data center relocations today
Over the past two years, I have seen a dramatic spike in the frequency of data center relocation requirements. This is expected to continue for at least the next few years because the convergence of increased physical compaction of IT systems and the mean age of a typical data center will continue to force many data centers into early obsolescence.
More than half of the data centers I have seen relocated in the past two years are facilities that are seven to 10 years old. The typical planning horizon for a commercial building is 20 years. Today’s equipment power densities were not considered seven to 10 years ago. These power densities continue to be on the rise, with some analysts predicting greater than 40kW per rack and beyond. Unless your data center could easily handle that density today, chances are you, too, will be moving sometime in the next few years.
Data center relocations are an exceptionally high-risk concept. The level of effort put into the planning phase by those who specialize in orchestrating these migrations is directly proportionate to the amount of risk mitigated. It is an exercise that requires collaboration with areas of the business with which one might not normally interface.
Keep in mind that everyone is a stakeholder and approach the discovery exercise with patience and an open mind. These two guiding principles are critical to uncovering all of the information needed to accurately identify dependencies and understand the sequence of events required to move your assets. Chances are you’ll probably need to assemble a team comprised of more than just two guys and a truck.
Kris Domich is the Principal Consultant of Data Center Solutions for Dimension Data North Americas. Kris has over 16 years of technical and consulting experience in data center migration, design, management and operations. His data center migration clients include many Fortune 100 companies as well as federal, state and local agencies. In addition to the United States, his repertoire spans continents, with significant experience in Europe, Middle East, Asia, Australia and Latin America.
Recognized globally for a strong business acumen and profound technical knowledge, Kris has served as a direct advisor to the executive leadership of numerous Global 1000 companies. Kris is a regular speaker at international trade shows and events, and often publishes articles on data center design, electro/thermal considerations, virtualization and other data center trends. He can be reached at firstname.lastname@example.org.