This the third article in series on DCK Executive Guide to Custom Data Centers.
Like any large scale project, when commissioning a data center design, whether standard or custom, a clear understanding of the responsibilities and points of contact (POC) and/or project managers (PM) need to be carefully selected and agreed to by all involved parties. It is highly recommended that the POC or PM for the organization that is purchasing or leasing the data center be generally familiar and have some experience with the operation and basic technologies of a data center. This is especially important for a custom design, and simply appointing an “all purpose” internal POC or PM without any specific data center experience should be avoided if at all possible. If such a qualified person is not available internally, consider utilizing a qualified independent consultant to act as the POC or PM or at the very least a trusted advisor. While they do not have to be an engineer, they do need to be able to fully understand what is being asked of the bidding data center design and build firms and the implications of their responses, questions or change requests as the designs are developed.
Before delving into the details, let’s first clarify the general data center categories and terms; standard, build to suit and of course a custom design.
Standard Data Center
While there really is no such thing as a generic “standard” data center, it generally involves a design that follows common industry standards and best practices. This usually covers the layout of the rows of cabinets, typically capable of supporting a moderate power density, then selecting the tier level of infrastructure redundancy and the total facility size commensurate to your organization’s immediate and future growth expectations. This type of data center is readily available for lease or purchase (please see part 1 of this series “Build vs Buy”) and is built using standard equipment and straightforward designs.
Build-to-Suit Data Center
The “Build to Suit” term and other similar marketing names such as “Turn Key” and “Move-In Ready” are used by some data center builders and providers in the industry. While the name sounds like, and would seem to imply a completely custom design, it generally offers a somewhat lower level of customization within certain limits of a basic standard design. This should be given serious consideration, since in many cases it may meet some or most, if not all of your specialized requirements, with a minimal cost impact. Also by keeping within the basic framework of a standard design, it would be less likely to face early obsolesce should a normal traditional technology refresh occur.
Custom Design Data Center
Like a custom built race car, designed and built for performance, a custom data center should represent a technically leading edge, tour de force design. In the case of a data center, the extreme performance is typically manifested in the form of higher flexibility, reliability, energy efficiency and power density, or some combination thereof.
Hardly a week goes by without some headlines in the data center publications announcing a new custom built data center based on a radical new design, most commonly by a high profile firm in the Internet search, social media and cloud services arena, such as Google, Facebook, or Microsoft. It is important to understand that these are typically based on very large scale dedicated applications and may involve specialized custom built hardware for use in so called hyper-scale computing. As an example, Facebook and Google utilize unique custom built servers (each has their own different server design), which do not have standard enclosures and require special matching cabinets, as well as specialized power and cooling systems.
This results in some technical and financial advantages, primarily related to lower cost per server and better overall data center energy efficiency. However, before embarking down the path of a highly customized data center design, it is important to understand that it requires a sufficiently large scale and IT architecture. It also may limit the general ability to support standardized racks and IT equipment. Let’s look at some emerging trends in custom data center designs.
Hybrid and Multiple Tier Levels
Tier levels generally refer to the level of redundancy and fault-tolerance resulting in a projected level of availability rating for a data center (1 lowest, 4 the highest).
One area of customization that is becoming more popular is the incorporation of multiple tier levels of infrastructure redundancy within the data center. This can lower costs and may increase energy efficiency by creating a lower tier level (i.e. Tier 2) zone for less critical applications, while still providing a high level of redundancy (Tier 3-4) area for the most critical systems and applications.
There are also those data center operators and owners that do not feel that they have to exactly follow all the requirements of the tier level system, but may prefer to use selected concepts and have a hybrid design. This allows them the flexibility to allow for greater level of redundancy of the electrical systems (i.e. 2x[N+1] dual path system — comparable to a tier 4 design), while using a less complex and lower cost cooling system, with only N+1 cooling components (for more details on tier levels please refer the “Uptime” section in part 1 “Build vs Buy”).
Of course, once you have begun to explore a custom design you may choose to mix the multiple and hybrid design schemes to match you organizations various applications and systems requirements and may also lower your CapEx and OpEx costs.
There is also a growing trend to try to segregate hardware by environmental requirements. Systems such as tape backup equipment in particular requires tight environmental control, yet does not require much actual cooling or power density. By isolating them from other hardware such as servers, you are able to properly support and maintain the reliability of more sensitive disk based storage and tape library equipment, by tightly controlling the temperature and humidity. This also improves the energy efficiency of the cooling system for other more robust hardware such as servers, or the new solid state storage systems, by allowing for raised temperatures and expanded humidity ranges (for more on this please refer to part 3 “Energy Efficiency”).
Containerized Data Center
The data center in a container is an alternative that is beginning to find some traction in the data center industry. These can be either an add-on to a traditional facility or the basis for an entire “data center”, based primarily on containerized or modular prefabricated units. Some designs are based on a core power and cooling infrastructure system meant to support these systems that are weather-proof units which can be placed on a prepared slab and then connected to the core power and cooling systems. While some other containers may require a warehouse type building to shelter them and again need to be to be connected to the core support systems.
Although similar in concept, it is important to distinguish the difference between actual container units and some modular data center systems. It is important to note that containerized solutions or modular systems are not necessarily an inexpensive alternative to a traditional brick and mortar data center facility. They are typically best suited to very high density applications of tightly packed mostly identical hardware, typically thousands of small servers or several hundred blade servers, configured to deliver hyper-scale computing. Their main attraction is for those large organizations that required the ability to respond quickly to rapid growth in computing power and also to a certain degree to minimize initial capital expense, by just being able to add containers or modules, on an as needed basis.
Regardless, it is important to note that whether you consider a container or a modular system, they still have to be installed at a data center facility that will support and secure them and that the overall facility infrastructure must be pre-designed and pre-built for the total amount of utility power, generator back-up capacity, as well as power conditioning (i.e. UPS typically required for most containers), and in some, but not all cases, a centralized cooling plant.
Containers can be part of a hybrid custom design, based on a more traditional data center building as a core primary data center building which is relatively standard. However, the overall facility infrastructure has pre-allocated space, as well as power and cooling infrastructure for containerized systems which can then be easily added as needed for rapid expansion.
Open Compute Project
There are also some resources for “non-standard” or leading edge “outside of the box” designs. One in particular is the Open Compute Project (OCP) which has published its highly energy efficient basic designs and specialized IT equipment specifications. While not every organization is an ideal candidate for all the elements disclosed in the OCP designs, some aspects of the designs can be chosen selectively and incorporated into a custom data center. Some data center providers offer to build a data center based on the OCP designs.
You can download a complete PDF of this article series on DCK Executive Guide to Custom Data Centers courtesy of Digital Realty.