Organizations should examine their business requirements before deciding on the right level of availability and redundancy to build into their data centers, as there is a big cost difference between datacenter tiers, said industry players.
In an e-mail interview with ZDNet Asia, Naresh Singh, principal analyst of datacenter convergence at Gartner, said organizations should choose datacenter tiers based on a detailed assessment of the availability requirement of their applications and the optimal infrastructure needed to meet such requirements so that the data center will not be over- or under-resourced.
Noting that a Tier IV data center can cost nearly double that of a Tier III facility, Singh said it is critical to determine accurately the availability requirement based on business metrics. For example, organizations may consider the financial impact from various periods of unavailability and what amount of downtime is considered acceptable by them.
David Blumanis, vice president for regional data center solutions in Asia-Pacific and Japan at APC, which is owned by Schneider Electric, added that a data center's "criticality" is a major decision in the planning process. Businesses need to accurately assess their business objectives and align these with their IT needs as the approach to determining datacenter tiering can be vastly different, he said in an e-mail interview.
He noted that the decision generally comes down to the type of business that the data center is supporting. Compared to small and midsize businesses (SMBs), the uptime requirements in verticals such as telecommunications, finance and health are extremely high due to government regulations and high potential downtime costs and so on, he said, adding that this would affect the level of criticality required for the respective industries.
Moving up tiers a challenge
Datacenter tiers are based on a classification system by The Uptime Institute. The most basic is Tier I which has a single path for power and cooling distribution without redundant components and provides 99.671 percent availability. Tier II refers to a site with a single path for power and cooling distribution and redundancy, and has 99.741 percent availability.
Tier III is composed of multiple power and cooling distribution paths, of which only one is "active". It also has redundant components and is concurrently maintainable, providing 99.982 percent availability.
The highest level, Tier IV, is composed of multiple active power and cooling distribution paths, and is equipped with redundant components. Besides fault tolerance, it also provides 99.995 percent availability.
In an e-mail interview, Karthik Ramarao, head of integrated datacenter solutions at Dimension Data Asia-Pacific, noted that as the infrastructural cost of each tier varies greatly, it is prudent to plan long-term for maximum return on investment.
"Shifting from one tier to another is a process that requires time, effort and money," he said. "While it is technically possible to do so, it involves a lot of 'rip-and-replace', especially when looking to move up the tiers."
Moving from a lower tier to a higher one is especially challenging for an older facility, in terms of obtaining the appropriate components.
Gartner's Singh also pointed out that upgrading a data center from a lower tier to a higher tier is "extremely difficult" as it requires big renovations on the site right from the scratch. "This is because support space required for the raised floor goes up in multiples with every jump in tier," he explained.
In addition, "running multiple delivery paths of electric cables or networks and having multiple sources of power and cooling can be difficult to achieve in the current site building", and the data center owner may have to look at significant re-engineering, he said.
Schneider Electric's Blumanis added that in the Asia-Pacific region where modularor standalone data centers are not the norm, upgrading from one tier to another poses a significant challenge because organizations will need to take into consideration many external factors such as space and power constraints.
"Many data centers in this region are also customized to the individual company's needs. As such, each respective datacenter upgrade will require more time and effort as there isn't a 'one-size-fits-all' approach," he noted.
Due to upgrading challenges, Singapore-based e-government service provider CrimsonLogic is exploring a partnership with a Tier IV datacenter operator.
In an e-mail interview with ZDNet Asia, Quah Chin Yong, the company's vice president of continuity and operations, said the company's current Tier III datacenter locations have some limitations in upgrading to Tier IV. "We are open to partnering with other Tier IV purpose-built data centers or building our own data center that is [Tier IV-capable] at a suitable location should the need arise as we continue to seek out new business opportunities," he said.
'Five nines' beyond four tiers
While the highest datacenter tier is tied to 99.995 percent availability, industry players said it is possible to build a facility with 99.999 percent availability.
Gartner's Singh, however, pointed out that there is only "a short list" of Tier IV data centers certified by the Uptime Institute. The majority of users, he said, opt for Tier III or lower designs.
According to him, a data center with 99.999 percent availability is achievable but it requires stringent and consistent adherence to the high availability principles and requirements outlined by Uptime Institute. In addition, there is a need to minimize human errors by greater automation and standardization of the various operations practices involved in running the data center, he added.
Organizations that want and target "five nines" availability, typically will "lose huge money and brand equity if a transaction fails to go through due to unavailability", he observed.
Dimension Data's Ramarao noted that there is a distinction between "five nines" at a business application or a facility level. "It is possible to have a Tier III data centre with 'five nines' business applications, by having the right IT infrastructure in place," he said.
Blumanis believes that instead of focusing on achieving 99.999 percent uptime, companies should put in place disaster recovery plans for business continuity. "A minute of outage on the servers may translate to an entire day of recovery for the company," he noted, adding that business continuity can allow organizations to minimize downtime and losses.
Organizations, he said, can take a lead from banks and telcos which have a production data center and a disaster recovery data center. This arrangement enables a seamless failover transition, should there be a failure in their data center.