Every data center project begins with the business plan and business requirements in mind. This is fundamentally the starting point and the first building blocks of a data center project. An enterprise with intention to build a data center facility for their own internal use may have an easier time to establish and understand this starting point. But for a Data Center service provider offering co-location and hosting services, then it gets a little bit complicated.
In reality, regardless of how closely the initial design matches with the business requirements, in the long run, data center facilities rarely achieve the efficiency, capacity and operational targets defined in their initial designs. This is because a data center environment is never static and like a living organism, it is always evolving. In the day to day operations, there will be people / human traffic, periodic changes, expansions, upgrades, introduction of new equipment, de-commissioning of aged equipment, and other external influences. As time goes by, newer technologies, with possibly higher density equipment, requiring substantial incremental power and cooling capacity, will replace aged equipment. Pressures to maximize utilization and reduce costs, will lead towards consolidation of data centers into few locations, adoption of virtualization technologies, causing changes in operational procedures, etc. All these are part and parcel of the evolution of a data center environment, and no matter how you strive to achieve an equilibrium, as soon as you think you have reached an equilibrium, the operating conditions will likely change again.
For Data Center service providers, who are like hotels, they will rarely know when their next customers will be coming through their doors, or what kind of equipment these customers will be hosting at their facility. How do you plan the design of a facility for such requirement when these requirements are not known at present day?
The rule, therefore, is to design for flexibility, modularity and scalability. There are a number of design criteria to consider and if we could apply this rule for each of the design criteria, then, it is possible to design a facility that will be adaptable and flexible enough to accommodate current and future demands of the facility.
Design criteria to be considered:
- Fault tolerance requirements: Uptime institute’s Tier classification specifies the level of fault tolerance for data centers and the tier level requirement will drive the design specifications. For a colocation service provider, the tier specification must correspond closely with the service offerings’ SLA. It is also possible for a service provider to design a multi-zone facility with varying tier levels.
- Visualize the layout of the facility by considering the structural room layout, i.e. location of walls, service corridors, access points / doors, support columns, viewing gallery (if necessary) and key utility connections, and mapping that structural room layout to a design racks layout which depicts the location of IT equipment racks in rows. IT equipment racks in the layout should be take into account the structural room layout constraints, airflow path, containment strategy and access ways.
- Power source and distribution: Review your utility feed and capacity, understand how best to distribute this capacity for your data center and building use. The design criteria for power subsystems should include modularity of UPS, backup diesel generator, PDUs and flexible and scalable distribution of power to the IT racks.
- HVAC systems: Defining appropriate strategies that are flexible and modular will help address future deviations and changes in loads of the data center. For instance, using an underfloor distribution of air may be effective means of cooling a heat load of up to 4.5kW/rack given a maximum of 700 CFM, but will be challenged if the heat load is more than that. Strategies to adopt closely coupled cooling with redundancy through multiple units of cooling systems and interlaced configuration of CW loop or DX loop, will help.
- Do not neglect fire protection systems (for detection and suppression) because a data center has a significant risk of electrical fires. For every segregated area / zone, there should be a string of detection devices, e.g. if you have raised floor, there should be detection devices beneath the raised floor. Although pre-action dry sprinklers may be mandated by local regulatory bodies, be mindful that valves do misbehave over time.
- As the cost of constructing a data center vary significantly, depending on whether it is a greenfield site, leased building, retrofitting an old data center facility, etc. and what tier of fault tolerance is required, expected density and payloads, etc. Sometimes, it may make sense to conduct a value engineering study of the construction costs and associated life cycle costs to determine what are the cost elements that are closely associated with the different levels of data center reliability, fault tolerance, and upsizing the capacity for future needs.
- Think about the operational model: How would the data center facility be managed in the future? Will there be a single organization / department who will manage all of IT services, Network Infrastructure and Facilities? Or will it be separate organizations? How do you formulate the strategy for an effective Building Management System (BMS) and would you integrate that with your Operational Support System (OSS)?
Obviously, there are many more criteria and factors to consider in planning a new data center. The above are just examples for thought. There is no short-cut to this planning exercise other than going through the this thought process.
Tags: Business, Cost, Data Center, Density, Design, Plan, PUE