Data Center Compromises

Featured Image

A Symphony of equilibriums…

Life is impossible without compromises. Data centers, like any other walk of life, are resolutions of a chain of compromises. Usually people believe that the main compromise in data centers is between cost and availability. It’s not that simple, unfortunately. To compromise, organizations typically undermine important measures in order to render maximized outputs for a given KPI. So what are some important data center compromises? What are these compromises costing us? What are the opportunity costs for each or in simpler terms, what are we giving up to achieve a certain other? These are all questions that need to be answered for a well-balanced and equilibrium data center.

Availability vs. Efficiency

These days it has become a ritual for data centers to compromise high-efficiency for high-availability and vice versa. Usually, if you found an “availability-centric” data center, you would find it hard to simultaneously sense tangible efficiency measures implemented. With the same token, when you visit “efficiency-centric” data centers, availability can be at grave risks. We’ll take free cooling as an example. Free cooling is a viable solution if it’s engineered right and your data center node is located in the right environment. Depending on free cooling alone to facilitate a mission-critical data center node can be catastrophic to your availability requirements. This is due to the fact that you can’t simply open the windows and enjoy the fresh air. By opening the windows, you are allowing unwanted expenses and threats such as dust, moisture, EMI, heat, cold, etc. to creep into your data center. So what you ought to do is to bring the outside air into specialized chambers, then, sample and measure the air against your acceptance thresholds. Accordingly, you must subsequently treat the air to your ideals. This means that if the air is too hot, you need to pay to cool it; if it’s too cold, you need to pay to heat it (this is why you can’t just build your data center in the north pole). If it’s too dry, you must humidify it, and if it’s too moist, you must dehumidify. If you are sitting on volcanic belts or in dust prone environments, the cost of filtering and shielding yourself against all the potential threats can be greater than what you originally planned to save via this approach. Ignoring these important and necessary facts and simply trying to be “efficient” will hamper your availability and even prove that your efficient cooling methodology is not so efficient after all. On the contrary, installing and operating excessively redundant units of power, cooling, and other equipment in a single location whereby your application is supposed to be supported by a series of locations may not be the most efficient choice. This is another critical reason that your data center must be application-centric and you must always see yourself and every choice you make as an integral part of the application ecosystem.

When application delivery is your true aim, then exerting too much in too little and vice versa doesn’t make sense anymore.

Safety vs. Security

Sometimes over emphasis on the security aspect of your data center can adversely affect site and personnel safety. Likewise, certain safety emphasis can put security at risk. For example, there are technologies that can provide further security to a data center node by emitting radiations. But these can be harmful to your people and your equipment. So utilizing them is out of the question. Likewise, implementing some safety measures i.e. uncoordinated and unjustified safety drills without regard to security protocols can put the data center at serious security risks. Other examples could be when your people need to make an emergency call for an ambulance from inside a faraday cage that is aimed to prevent EMI, EPO button that may disable the elevator half-way when people are still inside, authentication system that fails from the inside when people are trying to exit during emergencies, or electrified fencing to secure the data center perimeter, which can put people’s life in danger. Thus, it is crucial that your safety and security teams work hand-in-hand to provide a well-balanced environment that is simultaneously safe and secure for your data center and its resources.

Resilience vs. Capacity

Assume that you have a facility that runs on a pair of 1MW generators, designed for a 2N parallel configuration. As you add IT load your eventual actual live load ends up to be ~ 1.5MW. In this case, do you still have the 2N resilience that you had planned for? Surely not! You simply compromised your resilience to satisfy your capacity needs. This is why a data center’s infrastructure cannot be measured in isolation from its operational parameters. Understanding this fine element, underlines the importance of capacity as an integral part to your ecosystem. Having redundancy without taking the actual capacity needs into consideration is meaningless.

Having redundancy without taking the actual capacity needs into consideration is meaningless.

Therefore, a data center is required to balance its resilience requirements with its capacity needs. This is analogous to the story of roads and transport in our mega cities. Having roads is one thing, but being able to use them when you need them the most, without sitting in traffic for two hours, is something else. Many metropolitan residents experience this all the time. The road is there, wide-open when you don’t need to use it situated at work or sleeping at home, but as soon as you want to get to work or drive home, traffic simply wont move. So it is not enough to just have roads, what’s needed is a road system with ample capacity that can handle the traffic/load during peak. The story of our obsolete roads and transport system can be the subject of another discussion. What’s important here is that data centers cannot be measured by the number of UPSs, chillers, CRAHs or generators they posses. Rather, how these overall systems service the required load and still remain resilient is key.

Physical Security vs. IT Security

Sometimes you wonder what goes into the minds of policy-makers when they make the data center entrances and visits so difficult and cumbersome at some facilities, while many of the same data centers severely lack proper IT security for their hosted applications. If we spend as much money, time and effort that we spend in fences, cages, barriers, cameras, mantraps, security counters and security personnel into IT security of our data center, then, and only then, our security investment will result in true outputs, not show business! At the same time, all the IT security can prove useless if physical security is compromised. Again, the problem is the lack of application ecosystem-mindset and improper understanding of what a data center truly is and ought to deliver. If the goal of security is to secure the application ecosystem, then all the attention will not be vested over fencing just the generators or just in firewalls. This calls for a balanced approach towards security that will have the security teams and policy-makers understand the consequences and to work simultaneously on both the physical as well as IT aspects of the application ecosystem security.

Infrastructure vs. Operations

The best infrastructure will not save you from data center shortfalls when you have not invested as much in your people. Likewise, the best people will not protect you when you have not vested as much attention into your infrastructure. You cannot give a non-redundant, non-efficient, unsafe, unstable and insecure infrastructure gear to your people and expect them to deliver high marks of service, including 99.999% availability, no compromise on security, absolute safety, OpEx reductions, etc. simply because you hired the best and the brightest. In such cases, even Superman can’t help you. At the same time, you cannot hire unqualified personnel and expect them to run your state-of-the-art data center flawlessly. Both notions are disaster scenarios. Your people’s knowhow, expertise quality and quantity combined with your predefined effective operational policies and procedures, must be aligned with the quantity and quality of your infrastructure for your organization to maximize its potentials and optimize its use of its vested resources.

Your people’s knowhow, expertise quality and quantity combined with your predefined effective operational policies and procedures, must be aligned with the quantity and quality of your infrastructure for your organization to maximize its potentials and optimize its use of its vested resources.

IT vs. Facility

Usually organizations are either IT bias or facility bias, whereas balancing both is the recipe for success. You can pretty much get a sense of the data center architect’s background as soon as you walk into a site, by witnessing how focused or distracted the data center is on certain aspects. If you just spent $50million on your new IT assets, have you made sure that your site and facility are geared to render you the level of service your new IT gear deserves? If not, you just wasted your money. Similarly, if you just built a brand new data center facility for an IT gear that you know is a far cry from the facility that will be hosting it, or for an application that doesn’t even need such level of availability and resilience that you built into your new facility, again you have wasted your company’s resources. Always getting the facility teams to align their developments with the ‘true’ needs of the IT and ensuring that IT comprehends the realities of the facilities is a challenge. But it is crucial that both ends of the compromise are fully aware of the consequence of the choices made, as they relate to each other.

Time vs. Money

So often, projects are delayed due to budgets not getting released in time, contractors and vendor payments being delayed, or IT managers end-up settling for inferior designs and technologies due to lack of funds. All of this costs data center experts time, and time is money. Also, often enough, to save money, we spend less time in properly designing or planning our data center. That lack of ample preparation gets us wrong assessments of the realities we will ultimately face and thus everything in the lifecycle of the data center becomes an unanticipated challenge rather than a routine and foreseen task. Therefore, we end up spending both further time and money into the same subject and in damage control instead of a progressive and efficient operations regime that becomes our operational culture. The more time we spend in advance preparations, the less money we will end up spending during the life of a data center; and the more funds we provide to our necessary needs, the less time we need to invest in alterations and getting around problems.

CapEx vs. OpEx

It is a fact that lower initial cost (CapEx) doesn’t necessarily warrant lower total cost of ownership. Data center is not something you buy once and shelf it. It is a live being that will continuously generate cost/benefit results by the measure of its live operation. Data center is one of those things that cost of running it can end-up being more than building it. It is not enough to look at the price tag of your systems to decide if such system is financially feasible for your organization. What drives cost of ownership is the upfront costs as well as operating costs over the life expectancy of the data center [TCO = CapEx + (OpEx * DC Useful Life)]. Therefore for every system or component of your data center, you must find the right balance between both capital as well as operating expenditures to run an equilibrium data center.

Data center is a business of equilibriums!

All of the above mentioned parameters could also be interlinked and served against one another. For example, running low on capacity can directly influence your efficiency; implementing higher security can compromise your efficiency; higher resilience could hamper safety, and so on. Furthermore, the above and certainly more factors are KPIs that you must balance to achieve your equilibrium data center. Hence, there are tons of controversies in each state of data center design, build and operation. From your site selection, to cooling and power design, to IT parameters and operational disciplines, all the way to consolidation and green initiatives, they all require a series of equilibrium decisions that results in a balance of the above.

If you have been involved in any one of the mentioned scenarios and failed to achieve a balance, don’t worry you are not alone. Many fall short in aligning or making their data center compromises feasible. What’s important for you is to obtain a clear grasp for the opportunity-cost of your every choice and the consequences you would expect thereof. Remember, that data center is a business of equilibriums! This means that you must achieve a state of assurance by convening perfect balance between diverging forces. Your application ecosystem is a symphony of equilibriums, a balanced harmony of your optimized choices…

Your application ecosystem is a symphony of equilibriums, a balanced harmony of your optimized choices…

List of Comments

No comments yet.

Leave A Comment