Fundamental Design Criteria for Data Centers
Data Centers are not generic in design; they must be designed to meet the specific needs of the Owner. The requirements vary drastically as the complexity of the data transacted and stored increases and as the data becomes more critical to the Owner’s operation. Even the loss of data in some non-critical systems can be costly but, as the importance of the data increases so should the effectiveness of the Data Center to protect against data loss.
The fundamental criteria for most Data Centers should be: (1) Reliability, (2) Capacity, (3) Functionality, (4) Safety, and (5) Continuance/Recovery. If all of these five items are thoroughly analyzed and the findings incorporated into the design, the Data Center should function well for you over its designed life.
Reliability:
This is where the criticality of the data transacted and fiscal responsibility meet to set the guidelines of the design. To provide the design needed, you must first decide what level of Data Center is required to accomplish your goals. This should be based on the sensitivity of the data and the value of this data. This is the basis for setting all of the design concepts. The higher risk factor from either or both sensitivity or monetary value of the data the more redundancy should be included in the design.
Today, most hardware is equipped to accept redundant services for both power and communications. Many times systems fail because one or the other of these redundant paths is not complete enough to keep all critical system hardware operational (think Single Points Of Failure analysis).
Redundancy should not stop at just power and communications but should be taken into consideration in the design of all infrastructure systems critical to the operation of the Data Center. A failure in an improperly designed HVAC system can bring your facility down just as fast as a power or communications failure.
Many times, in the case of natural disasters, the roof and exterior walls are the first to cause problems. The telecommunications industry has set the standards for building systems reliability with their Exchange Facilities requirements. In many cases these are buildings within a building to where, if the outer shell is damaged, the interior roof and walls provide a second line of protection. This “hardening” of a structure has become easier with the advent of new materials and methods that do not require an interior concrete structure.
Another item that directly affects the reliability of a Data Center is proper grounding and lightning protection systems. The proper dispersal of static electricity can facilitate tremendous strides toward accurate recording of transmitted and stored data. Achieving a balanced ground resistance throughout the Data Center is essential plus achieving as minimum of ohms resistance-to-ground as possible. Achieving five ohms or less to ground is difficult and it must be checked seasonably for compliance to design.
We are also dealing with design concepts that are carryovers from the main frame age and they are not totally applicable when deploying modern technology. One concept that creates many problems is the use of raised floor space for cabling. The high density hardware that is presently being used creates massive amounts of data cabling to any one cabinet. This, plus power cabling in the same space, uses up most of the space under the floor that was design as an airflow plenum space and very quickly becomes unmanageable. Also, in the fluid environment that we have created, systems get moved and enlarged to where cabling has to be rerouted and this makes an even bigger mess. Quire often, when it’s not possible to move a single cable, it is abandoned in place and a new cable drop is run. This same under-floor space, because cabinets are design for airflow coming up and around the hardware mounted within, changes and/or blocks the under-floor air distribution efficiency. This translates into air distribution problems; hot spots, imbalance, etc. Keeping an under floor air plenum clear and balanced with correct air flow up through each cabinet requires a well designed and maintained system. The answer? Overhead cable (power and data) management.
Capacity:
As difficult as it may seem, capacity should be forecasted on a 1 year, 3 year, and 5 year basis. The forecast should be segregated into components as applicable:
The separation of components is necessary to address the different asset growth rates. As form factors reduce in size and the aggregation of computing asset into single chassis units increase, the physical size of the new facility may not have to be that much larger than your present needs except for the physical plant systems. The physical plant assets are guaranteed to increase and, therefore, increased space utilization for these assets must be provided for.
In many cases the design can be formulated so that the construct can be staged over time resulting in a pay-as-you-go budget. Modular design of both physical space and building systems can save you money now and make expansion easy with minimum disruption to operations. Careful planning means a facility that can grow cost effectively.
Functionality:
All of your present systems, hardware and software, must be examined for expectancy, efficiency, and how it will interface with your future expansion plans. With the advent of new technologies, such as blade technology, design should follow a system approach decision. Other decisions to consider such as what applications are to be kept separate, if any, and how the equipment is to be mounted, will constitute the drill-down process that must be followed. Once again, whether or not you might want to locate your communications equipment and data storage hardware in separate areas will have to be addressed.
The security of various functions and personnel access to these functions has to be decided. The higher the value of data transacted and stored the more attention must be paid to its physical environment.
Your Operations Center (OC) should be the nerve center of your entire operation since it monitors and directs all operations. This center, again depending on your operating requirements, can be manned just during normal working hours with on-call personnel for after hours support, or a full fledged 7-24-365 OC.
A functional Data Center must have a building management system sophisticated enough to allow it to monitor all critical physical plant functions to a point where it can prevent many total system failures. In addition, the system chosen should be capable of tracking and scheduling most building preventive maintenance requirements. (An ounce of prevention......)
Safety:
Personnel security is directly related to the value level of data you are transacting. Physical isolation of all critical area should be designed so that only authorized personnel have access and that the areas are not subject to random outside vandalism. The use of personnel traps for access to secure and ultra secure areas must be considered.
Also, give serious thought to a secure entrance that protects your security force from random violence. Various personnel identification systems are available along with CCTV systems that can be deployed down to the individual cabinet level. Give consideration to a comprehensive asset management system whereby all hardware is coded so it cannot be removed without prior authorization.
Fire protection is another area where the design should be looked at differently in a 7-24-365 facility than in others. The methods contemplated must be thoroughly discussed with the Fire Marshal that has control over the facility. Tremendous care must be exercised in designing any of the modern chemical system and especially with those that are under high pressure. There have been cases where a discharge of one of these high pressure systems has caused more damage to the facility than the fire it was intended to extinguish.
Protection of all exterior equipment from vandalism must be considered as well as the placement of all fresh air intakes into the facility.
Considerations should be given for protection of the building from vehicles where drivers have lost control or acts of direct violence using vehicles. Personnel safety should always be considered in areas outside of the building.
Continuance/Recovery:
No matter how well you plan, events can happen to interrupt the operation of your Data Center. With this in mind, disaster planning, preparedness, continuance, and recovery should be a foremost design criteria throughout the design process.
Disaster planning means looking at your facility and all of the hazards that could impact the operation and rank them for their impact on operations. Working by a hazard ranking affords you the opportunity to analyze the damage they could create and what can to be done to minimize operational effect. Part of this process is to evaluate your personnel and establish the groups that can remain (if possible) at the facility throughout the emergency. You list all of the materials and spare parts needed for minor repairs and what is required for personnel comfort. Establishing with vendors and contractors the availability of manpower and critical spare parts for rapid repair after the emergency is critical.
The preparedness phase is making sure you have all of the items on hand that was listed in the planning stage and personnel trained as to where these materials are stored and how to use them in case of a problem.
Summary:
Because there are so many different levels of Data Centers in both size and complexity we are not able to cover all of the issues to consider in the planning of a new Data Center. This is a highly specialized endeavor so an effort should be made to have an unbiased review of your data transacting operation. In most cases your in-house personnel do not have the time to devote to developing this information and your hardware/software vendors that offer their design services will be ever cognizant of the design on their products and, therefore, rarely able to address “foreign” equipment. It’s your data center that you have to operate when the construction is complete and a significant amount of capital dollars has been expended.