peshkova - Fotolia

Evaluate Weigh the pros and cons of technologies, products and projects you are considering.

Uptime Institute tier classification system faces new rivals

Because the Uptime Institute focuses primarily on power and cooling, alternative data center certifications have emerged. Their chances of replacing Uptime, however, seem slim.

The Uptime Institute tier classification system has been the subject of debate for nearly as long as it has existed. For decades, it was the primary standard by which data center managers measured levels of efficiency and system availability in their facilities. Now, new alternatives to the Uptime Institute tiers have emerged as potential competitors.

Common criticisms of Uptime

Some argue that Uptime's tier classification system covers only the power and cooling infrastructure. But elements such as the network and security are just as critical to the availability of a data center. Critics also question whether the tier definitions have kept up with technology -- particularly alternate power designs and multisite rather than internal site redundancy.

Two tenets that are sacred to Uptime generate the most criticism. First, there are no partial tiers. If facility designers overlook one element of the electrical or mechanical infrastructure that prevents a system from being concurrently maintainable, they fall into a lower tier. Second, only the Uptime Institute can provide a legitimate tier certification for a facility -- which is accompanied by a significant charge. The Uptime certification is a performance-based approach -- not a prescriptive one -- so there's no checklist to guide designs.

Alternative data center tier classification systems

Many organizations and standards have attempted to replace the Uptime Institute's certification system, particularly the Telecommunications Industry Association (TIA) ANSI/TIA 942-A standard. TIA-942 began as a cabling standard and covers a wider spectrum than the Uptime documents. It's also an official standard recognized by the American National Standards Institute (ANSI), whereas the Uptime Institute's tier standard topology is not. Although the TIA-942 standard has had some success -- more in Europe than in the United States -- design specifications do not cite it nearly as often as the Uptime Institute tier levels.

In addition, two other alternative tier systems recently emerged.

Some argue that Uptime's tier classification system covers only the power and cooling infrastructure. But elements such as the network and security are just as critical to the availability of a data center.

In June 2017, Switch, a Las Vegas-based data center provider, unveiled an extended standard called Tier 5. Switch Founder and CEO Rob Roy does not consider the Uptime Institute's highest rating -- Tier IV -- sufficient enough for a best-in-class hosting, cloud or colocation site. Roy hired two of the original authors of the Uptime tier classifications to help create and run his new Data Center Standards Foundation and develop Switch's Tier 5 criteria. The Switch rating adds components such as carrier redundancy, locations of power and cooling equipment and security. Switch intends to form a nonprofit body that will provide this proprietary rating system and its certifications.

Switch is a commercial enterprise, and one would assume it will market its new classification system to competitors. Whether those competitors would want Switch to evaluate their data centers remains to be seen.

The Green Grid's new Open Standard for Data Center Availability (OSDA) could be Uptime's fiercest competitor. The Green Grid developed widely accepted metrics, such as Power Usage Effectiveness (PUE), for tracking power efficiency in data centers, and its members are among the highest-profile companies in the industry. Therefore, anything The Green Grid promotes could find some interest.

OSDA is intended to be nonproprietary and will build on the same fundamental classifications as the Uptime Institute Tiers, TIA-942 levels and others. Its hierarchy is built upon four core levels: basic nonredundant, basic redundant, concurrently maintainable and fault-tolerant.

OSDA emphasizes three considerations that The Green Grid believes other classifications overlook. First, like Switch's new standard, it incorporates the use of alternative energy sources, such as solar, wind, fuel cell and tidal, as main energy sources, as well as innovative approaches to substation design that might provide capabilities similar to more conventional designs. Second, the standard acknowledges architectures that are based on multiple networked data centers for redundancy -- a basic tenet of mega-scale operations, such as Google -- instead of only on individual power and cooling redundancies. Lastly, it provides a way to classify data centers that implement availability features beyond their present classifications, but not enough to meet requirements at the next level. This addresses one of the most well-known complaints of the Uptime's rigid approach.

Test your data center certification knowledge

What's the highest certification level offered by Switch? Do you know the generator capacity requirements for critical load, cooling and life safety to achieve OIX-2 data center certification?  Take this quiz to certify your knowledge.

OSDA is only in preliminary form, but The Green Grid intends to release it in stages throughout 2017, with ongoing collaboration and improvement in 2018. However, since The Green Grid is a voluntary organization, accomplishing this might take longer than expected, and it could be problematic to provide actual certification services.

Will Uptime tier classification be replaced?

Uptime's had too much traction for too long. The four-tier concept's simplicity is a major appeal, and anything more comprehensive will also be more complex -- which has been the downfall of other attempts. That simplicity is a double-edged sword, however, since many data center designers that overlook the simplicity of the Uptime tier classification think they've designed a Tier IV facility, when it may not even be Tier II.

Still, it's doubtful that any alternatives to the Uptime tiers will act as a replacement anytime soon. They might influence some revisions, but even that is questionable.

Next Steps

Stay up to date on data center metrics, standards and certifications

Know these data center standards and metrics

International Data Center Authority program also looks to challenge Uptime

Dig Deeper on Data center design and facilities

Join the conversation


Send me notifications when other members comment.

Please create a username to comment.

Do you think emerging standards, such as Switch and The Green Grid, could eventually replace the Uptime Institute's certification?
For all the praise being heaped on ODSA in this article it is worth nothing that ODSA is not a Standard in any formal sense and is not recognised by ANSI or any other formal national standards body. This is also the case with the new suggestions from Switch and the more established Uptime Tier Topology. Merely putting the term "standard" in the title does not make it one!
It is a pity and a significant oversight that the article did not mention CEN/CENELEC EN 50600 as this is a genuine international Standard in the true sense, which is specific to data centres and endorsed by the national standards bodies of all EU member countries. It is published, available in multiple languages, does allow for partial levels for the primary elements and is currently undergoing revision to produce the second release. While this is currently very much European based there is a strong possibility of wider international adoption by formal standards bodies. I would suggest this is one to watch very closely. 
The issue of "Certification" is also somewhat contentious. In reality there is no genuine certification programme specific to the data centre sector which has either 3rd party governance or endorsement from a formal national standards body. The terms "Certification" in the world of genuine standards has a very specific meaning and can only be applied to standards that have been specifially designed to be certified such as ISO 9001 or ISO 27001 etc. When people refer to certification in a data centre specific context what they generally mean is a "Letter of Compliance". On this basis there remains room for a genuine certification programme within the sector.

Here's a response from the author: First, this article was specifically about proposed expansions of or challenges to the Uptime Institute Tier Classifications, which remain the most often quoted method of stipulating expected Uptime in the world. (We just got a request from China that specified by Uptime Tier Level!) I have personally seen nothing about EN 5600 presenting any kind of a serious challenge to Uptime Tier Levels, and have certainly seen nothing in which EN 5600 was listed as a Reference Standard for a project. It may be more widely referenced in Europe, but in the US, and even in the UK, we haven't seen it even mentioned.

The "emphasis" placed on OSDA was due simply to the fact that it comes from The Green Grid, and nowhere in the article was the word "standard" capitalized. TGG calls it a "standard". That's their label, not mine, and it would be improper to identify it differently than the originators. Yes, there was a fair amount of verbiage devoted to their proposed approach. That is because TGG has some very bright and powerful members, and their development of the PUE Metric indisputably put them on the global map. Like the Uptime Tier Levels, PUE is the most widely referenced metric for its purpose in the world. Although other metrics they have developed have not gained that kind of traction, anything TGG proposes has to be taken somewhat seriously.

If EN 5600 is truly gaining traction in Europe, and if it is actually becoming more widely referenced than the Uptime Tiers, then perhaps an article should be written comparing it with Uptime, TIA-942, and other classification approaches. But that was not the purpose of this article, and I will be first to admit that I do not have the first-hand range of knowledge as to what is being most widely used in Europe to draft it.

Collocations would say that Uptime certification is an expensive piece of marketing paper. There are well known operators with excellent data centers and they don’t see the need for this. Having multiple standards / certifications does not mean that the DC is well operated. It just means that the team is very well prepared for the evaluation. What the market really need is something that enables end users to easily evaluate data centers.

Bob's article certainly is a required discussion for both operators and those that underwrite risks for IT-dependent businesses.  However, the article's focus misses the central concept of "Standard" as defined by the various accredited Standards organizations (ANSI, IEEE, IEC, ISO, CENELEC, etc.).  As well stated by ANSI, a true Standard document vs a quasi-marketing tool (either a focused product such as UTI or focused cause celebre such as TGG and Switch) must be complete and have requirements of due process: "Openness; Lack of dominance; Balance; Coordination and harmonization; Notifications of standards development; Consideration of views and objections; Consensus vote; Appeals; Written procedures and Compliance with normative standards organizations' policies; and Administrative Procedures."

While Bob's inclusion of ANSI/TIA-942-B (consists of 134 pages) as the Standard is good, -942 provides primarily the requirements for the design of data center telecommunications infrastructure.  The most useful Standard that we and our Clients have been using is another ANSI Standard: ANSI/BICSI-002-2014 (consists of 500+ pages).  

BICSI-002 provides a wide range of information, recommendations, and requirements regarding all aspects of designing a data center.  BICSI-002 provides best practices that exceed the minimum requirements of TIA-942 and a significantly wider range of data center subjects not covered in -942.

BICSI-002-2014 is the second cycle (ANSI requires a 3 year renewal or expire cycle) and was written, edited and voted on by WW SMEs (large and critical inputs from APJ and EU entities).

ANSI/BICSI-009 Data Center Operations Standard is being developed and may be published in 2018.  Again the WW sourcing of writing/editing/voting SME is large.  It will establish a public, written, voted reference for operation and maintenance of the data center after it is built.

Thanks, Bob! Great coverage of one of my favorite subjects.
One of my pet peeves relates to the design of Tier IV systems that DO meet the intent of that class (as defined by UI), but are in fact less reliable than Tier III, or in some cases, even Tier II systems merely due to the overly complicated design. One can't forget that Tier IV systems are supposed to be "self-healing" upon detection of a fault. As well intentioned as that may be, that approach adds complexity, which may add to the probability of failure! The fault itself can be due to human error, or a large list of other possible factors - these include the increased complexity of the design, construction, commissioning, controls, operations, and maintenance. Oh, yeah - you can't forget about those pesky humans that need to be paid to support this extra effort associated with the increased complexity! There's a good reason why Tier IV facilities can cost upwards of two times the cost of the lesser tiers. Perhaps twice the cost is not enough... 
Let's talk!
FULLY agree. We often do bring up that issue of over-complexity driving increased risk. So many T3 and T4 sites have had the automated solutions ending up being the root cause of outages (e.g.: 451 Main/SFO, etc.). Not that I don't like complexity, but this can be better done with two T3 or even T2 sites with modern IT infrastructure and SW. Plus a complicated SW solution can be easier modified/corrected when an issue is identified vs a major MEP correction to an embedded issue that maybe wasn't even identified -- DURING COMMISSIONING. Commissioning is probably the key element of risk elimination that is very often NOT done right or extensive enough. It would seem that if one spends the money for a T4 DC, one would spend an even higher percentage of funds to execute an even higher level of intensity and completeness of commissioning (NOT just a "lights out" test).