ASHRAE Journal - July 2013 - 123

components requires data center operators that are better trained and smarter
(i.e., “A” players). Anything less than
highly trained “A” players could lead
to a higher probability of human error
(which would defeat the original purpose
of creating more redundancy in the data
center).

Resultant:
76.5% Reliable
Component 1
(85% Reliable)

Component 2
(90% Reliable)
Series System
Configuration

Redundancy Tiers
The industry frequently uses “tiers”
Component 1
(85% Reliable)
to identify the level of redundancy and
Resultant:
98.5% Reliable
to allow for a comparison of the resiliency of one data center to another. One
common topology is a four tier system
published by the Uptime Institute.
Component 2
(90% Reliable)
The four tiers deﬁned by the Uptime
Institute are:
Parallel System
Configuration
• Tier I – Basic Site Infrastructure;
• Tier II – Redundant Site
Infrastructure;
Figure 1: Two component system configuration in series and parallel.
• Tier III – Concurrently Maintainable
Site Infrastructure; and
Topic
Tier I Tier II
Tier III
Tier IV
• Tier IV – Fault Tolerant Site
Infrastructure.
Active Capacity Components to
N
N+1
N+1
N After Any Failure
A summary of the attributes that make
Support the IT Load
up the four tiers is shown in Table 1.
In practice, it was found that there
1 Active & 1
2 Simultaneously
Distribution Paths
1
2
Alternate
Active
was a signiﬁcant difference between the
original deﬁnitions of Tier III and Tier IV
in terms of cost and return on investment
Concurrently Maintainable
No
No
Yes
Yes
(ROI). Therefore, the market for designs
that fell between these two Tier levels
Fault Tolerance
No
No
No
Yes
grew. The informal category of Tier III+
was coined by data center designers and
Compartmentalization
No
No
No
Yes
operators who had, in response to their
client’s needs, facilities that exceeded
Continuous Cooling
No
No
No
Yes
Tier III but not to the full blown doubling
of equipment and infrastructure man- Table 1: Uptime Institute tier requirements.
dated by Tier IV.
In addition, some equipment vendors introduced an informal
Therefore, TIA-942 essentially allows for a data center
category of Tier 0 to describe data center topology that did not to be deﬁned by a variable tier per subsystem (e.g., a data
include an uninterruptible power component (Tiers I to IV all center could be classiﬁed as T2 E3 A1 M2 where the subscript
assume that there is a backup power source).
next to each letter refers to the Uptime Institute tier for that
The Uptime Institute’s redundancy topology is also referenced particular subsystem.
in the Telecommunications Industry Association (TIA) in their
In practice, it is common for end-users and designers to
data center standard ANSI/TIA-942. Whereas the Uptime In- use a mix and match approach similar to TIA-942. Most data
stitute considers their topology to be at the overall data center centers are some combination of the tier systems and the
level, TIA-942 expands the four tier concept by applying it end-users personal preferences based on years of operational
at the individual subsystem level with the four subsystems experiences.
identiﬁed being:
Finally, it’s important to remember that redundancy is only
• Telecommunications (T);
one component of achieving high availability. Additional
• Electrical (E);
considerations include the aforementioned “human elements”
• Architectural Infrastructure (A); and
(training, operations practices, qualiﬁcations of operations,
• Mechanical Infrastructure (M).
etc.), redundancy of the data center and software services,
July 2013

ASHRAE Journal

123
ASHRAE Journal - July 2013

Table of Contents for the Digital Edition of ASHRAE Journal - July 2013
