Marine mammals like dolphins breathe just like us. But while humans can breathe even when our conscious mind is asleep, dolphins are equipped with a voluntary respiratory system and therefore must keep part of their brain alert to trigger each breath. To avoid drowning, it is crucial that marine mammals retain control of their blowhole, which is a flap of skin that opens and closes under the dolphin’s voluntary control. Although still a matter of discussion, most researchers feel that in order to breathe, a dolphin must be conscious and alert to recognize that its blowhole is at the surface. So, how does a dolphin sleep?
The bottlenose dolphin sleeps by shutting down only half of its brain, along with the opposite eye. The other half of the brain stays awake at a low level of alertness. This attentive side is used to watch for predators, obstacles and other animals. It also signals when to rise to the surface for a fresh breath of air. After approximately two hours, the animal will reverse this process, resting the active side of the brain and awaking the rested half. This pattern is often called cat-napping.
So, applying the same concept on data center facilities, how do we maintain and service the critical infrastructure components in our data centers without reducing the number of capacity units required to support the data center?
As most of us are aware, Uptime Institute’s Tier Performance Standards are a widely accepted definition of a four tier classification approach of data center design topologies. One of the terms used in the Tier Performance Standard is Redundancy. This term refers to the components beyond the number of capacity units required to support the computer equipment. For example, if the number of units of capacity required to support the computer equipment is N, then having redundancy means having more than N units in place.
The beauty about redundancy in the design of a site infrastructure is that besides providing higher availability to mitigate against risk of component or distribution path failure, it can be configured to provide an avenue in planning for a safe way to maintain and service the critical infrastructure components.
The Tier Performance Standards defines Tier-I and Tier-II as having single distribution path which does not support concurrent maintenance. For Tier-III and Tier-IV classification, however, provides for multiple distribution path and at least N+1 component redundancy level, hence the ability and support for concurrent maintainability.
Lets take an example – if you have 20 racks of IT equipment in the data center and each rack contains 4.5kW of payload. That will be a total of 90kW of payload and equivalent heat. Imagine you adopt a cooling architecture (regardless whether it is row-based or rack-based or room-based), and you group the racks together into two rows and align them properly with a common hot-aisle (contained). Now, imagine each of the cooling units you adopt has a capacity of 15kW, which means you’ll need 6 units of these for the 90kW payload. Hence, N=6. An N+1 redundancy means you’ll need 7 units installed. With 7 units, you’ll be able to rotate the operation of any six units, and at any point in time, able to take one of them out for servicing or maintenance. Doing this will not affect the cooling capacity required for your payload. Hence, it’ll keep your operations alive and you’ll still be “conscious” (just like the dolphins).
Tags: Data Center, Operations, protection, PUE, strategy