Do we really need TIER IV data centers?

The DatacenterDynamics Global LinkedIn group had a heated discussion recently over the relevance of data center Tier certification

24 January 2013 by Penny Jones - DatacenterDynamics

Dr Carlos Garcia de La Noceda, Operations Executive: Is investment in a TIER IV 99.9999999 ad infinitum facility really necessary? We sent men to the moon and back with tin foil, chewing gum, paper and pencil calculations, walkie-talkies (practically) and less computing power than my Casio watch! Is there really a need – for technical reasons, not commercial or investor pipe dream – for such a facility, especially when the organization has mirror/back-up systems elsewhere?

Amod J, director, critical environments, Hong Kong: For most enterprises, a Tier III data center with a firm commitment from management for continued support for maintenance, spares for critical components and training of staff to increase retention should be sufficient. Beyond that, it is “bragging rights”. 

Rahul Shewale, consultant, Capgemini, India: It all depends on the kind of reliability and service level agreements with the end client under which the servers in the data center are hosted. If we are looking at robust and reliable data centers, Tier IV is a must, but if we go in terms of efficiency and infrastructure required, Tier III is more efficient and cost-effective.

Dr Carlos Garcia de la Noceda: Tier IV – or what is defined by the Uptime Institute in its expanded interpretation of ANSI/TIA-942 – is a panacea. The expense, in all aspects required for a so-called Tier IV installation, is a waste in my opinion. You can build a Tier III(+) data center with a more reliable infrastructure than a “box standard” Tier IV facility. You need to consider the reliability and quality of the equipment, maintenance regimes, personnel training, staffing, procedures, safety, etc. Each of those can increase risk, reduce reliability, or increase it if done properly.

Rhys Jones, senior electrical engineer, JCA Engineering: I wouldn’t say that Uptime has an expanded interpretation of TIA, given that the scope of its Tier assessment is actually more narrow than the national standard. I think you also need to be a tad careful about the use of terminology... In terms of reliability, it is demonstrable that ‘N’ generators are more reliable than ‘N+1’ generators; however, the resilience of the latter negates the catastrophic impact of a failure to one set.

I think the bigger question should be whether there is a need to have a single redundancy topology in your data center, and also in the age of distributed computing whether the Tier of a single facility is actually relevant. I will expand on that comment in the context of the kind of client that is likely to build a Tier IV data center (as opposed to a critical facility based on Tier IV). Generally, the client that can justify the cost of a Tier IV facility is going to be a large multinational enterprise operator that, in this day and age, is likely to be able to run concurrent IT functions distributed across sites utilizing a highly virtualized infrastructure.

Given what has been publicized regarding Deerns’ Guiding the Cloud I would have thought that a ‘follow-the-moon’ approach could be utilized such that a globally distributed cloud platform could direct IT process to facilities that are benefiting from reduced-cost energy tariffs during reduced demand, and also seeing night-time setbacks in terms of ambient temp. In this sense, if the platform is distributed over enough sites, why even bother with Tier III? I can see a benefit to having adequate amounts of redundancy in terms of fans, when utilized in an indirect air-cooling solution such that fan set-back to minimum on the curve can be achieved to maximize affinity law savings without detriment in performance.

Henk Waardenburg, consultant, physical IT infrastructures, Netherlands: I think an organization needing a Tier IV data center is either doing something very secret or does not have its Distributed Processing Policy in good working order. Furthermore, Tier IV is impossible in some parts of the world (like mine) due to the simple fact that no location with two separate grids owned and run by at least two different providers can be found.

When I reflect on all the problems I’ve seen in the past caused by equipment that was supposed to protect the IT gear, I come to the conclusion that multiple Tier I data centers in a Distributed Processing environment are far more beneficial.

Dr Carlos Garcia de la Noceda: The answer, in my opinion, is that each data center solution must be carefully studied and considered on a case-by-case basis, taking into account as many variables as possible, but always using the sustainability of the customer’s business as the ultimate litmus test. “Tier” is trying to “box” a complex set of decisions and solutions into neatly marketable packages.

William Angle, principal, CS Technology, Baltimore: The difference between a Tier III+ and a Tier IV data center is the containment of the mechanical and electrical gear already bought to supply Tier III+ reliability. Therefore you are arguing over a small increase in cost for the fault-tolerance built into a Tier IV facility. I have some depth of experience in the field, as I was part of the design team that built the first Tier IV facility in the world, and we knew at that time the reasons why we had decided to achieve a fault-tolerant, concurrently maintainable facility. The further development of the operational side of this field is as important as the design, given almost 60% of all infrastructure failures occur because of people, not because of infrastructure equipment failure.

It is a true statement that there are Tier III facilities that are more reliable than some Tier IV facilities, because of the high-reliability engineering assembled to train, document, provide procedures and properly identify the operational elements of the data center. It is also true that the Tier levels do not address environmental and ballistic hardening, security, structure, electrical density, efficiency and many other aspects of data centers that contribute to cost, and subsequent reliability and efficiency of the asset. The requirements of the business case will determine where the capital is spent, and to what level the infrastructure should be designed.

This quickly became one of the most popular debates on the DatacenterDynamics Global LinkedIn Page. We had far too many responses to include in this edition. Join the debate. Resgister at LinkedIn.com/DatacenterDynamics Global.

CONNECT WITH US

Sign in


Forgotten Password?

Create MyDCD account

Regions

region LATAM y España North America Europe Em Português Middle East Africa Asia Pacific