With the increasing focus on controlling capital and operating expenses (capex and opex) while enabling data center growth that fits the immediate business needs (“right-sizing”), modularity has become a hot design trend. But modularity can, and in many cases should, extend beyond simply replicating only the same elements (power-distribution units, racks, cooling units and so on): a heterogeneous approach can garner tremendous cost savings over a homogeneous one. In particular, multiple availability tiers brings right-sizing to the various data center tasks rather than treating the facility as a monolithic structure designed to accomplish what amounts to a single task.
Traditionally, a data center might be ranked as Tier I, Tier II, Tier III or Tier IV, meaning that the entire facility offers a certain level of redundancy to support a certain availability (say, 99.99% in a mission-critical case). Chances are, however, that some tasks in a given data center require less availability than others; in these cases, jobs for which 99% availability is sufficient constitute a tremendous waste of resources when they operate on 99.99%-availability infrastructure. Thus, subdivision of the design into availability tiers can greatly reduce capex, since less infrastructure is dedicated to tasks that don’t need it, and opex, thanks to declines in a range of ongoing costs such as maintenance, financing, energy and so on.
This design approach is effectively an extension of the multi-zone data center, which offers “a method of reducing capital expense dramatically since the electrical/mechanical equipment required was often reduced, while still maintaining scalability in the data center,” according to Gartner Research VP David J. Cappuccio. Multi-zone data centers support differing power-density levels in different areas, enabling reduction of infrastructure like power distribution and cooling for certain zones while providing more for the zones that truly need it. In addressing multiple availability zones, Cappuccio said, “Rather than build a tier 4 fully redundant data center that supports all mission critical systems, and everything else, why not build a tier 4 zone that supports mission critical (which may only be 15% of my overall workload), and assign tier 1, 2, or 3 status to other areas in the same building?” He suggests that a design combination of power zones and multiple tiers can cut capex by some 40%.
Why Use a Multi-Tiered Approach?
The official requirements of a Tier IV data center versus a Tier I data center are huge—to say nothing about the difference relative to a data center that avoids the requirements of the Uptime Institute’s tier classification system altogether. For instance, according to a summary in an HP white paper (“The Multi-tiered Hybrid Data Center”), a Tier IV facility targeting 99.995% availability (less than 30 minutes of downtime per year) must incorporate 2N generator and UPS systems, dual-active utility feeders, a 2N mechanical system and so on. For Tier I, which targets 99.671% availability (about 30 hours of downtime per year), the requirements are much more lax: the generator is optional, only a single UPS is called for and so on.
The difference in capital costs alone between the highest and lowest tiers is staggering. If a data center is running, say, just 10% mission-critical tasks, 90% of the services would be overprovisioned with fault-tolerant infrastructure if the entire facility is designed exclusively for Tier IV operation. By creating a multi-tiered modular design, that 10% could be served with the required infrastructure while providing only what is needed (maybe a single backup generator instead of two—or, more broadly, N instead of 2N) for less critical functions, all while following a strategy of adding infrastructure only when it is needed.