Notes from the Consultant's Jungle

The Tier Model for Mission Critical Facilities, created and governed by The Uptime Institute, is the most pervasively referenced data center tier model.  It is, however, not the only tier model for data center facilities, though it enjoys the majority of mind-share in this regard.  We have written about The Uptime Institute’s data center tier model rather extensively in this forum.

The four-tier model from The Uptime Institute was developed through thoughtful analysis and extensive empirical data from facilities of member organizations.   One of the reasons that enterprises have gravitated to this model is that it gives guidance as to what level of availability is necessary for certain business models and business characteristics.  While this is helpful, the level of abstraction of this framework of guidance is high.   A coarse application of guidance can lead to inaccuracy of planning.  In the case of data center projects, this exposes the possibility of over building or at least spending that is not accurately targeted.  For example, a coarse application of these guidelines could cause planning for a facility designed for one segment of the business’ applications at the expense of all the other enterprise applications.

Interesting also is the fact that while a facility may be uncertified or even non-certifiable to a particular tier level, its true availability performance often outperforms even higher tier ratings (especially in the case of quality providers).  The Uptime Institute model has sometimes drawn criticism because it excludes factors contributing to operational excellence, as well as risk management factors that vary significantly based on geographic location alone (as well as other things). The Uptime Institute is working on modifications to its guidance for precisely these reasons, and we look forward to those developments.

In the mean time,we now have the new BICSI-002 standard released (finally) in June 2010.  The newly released standard is “BICSI-002-2010, Data Center Design and Implementation Best Practices.”  This document is long awaited by the industry, and will likely be adopted as an ANSI standard as well.  The document contains advice relevant to IT telecommunications management, security management, operations management, facilities, A&E, and Construction.

Our focus today is with what BICSI has advised regarding data center availability models, and we’ll leave the rest for later discussions.  There is a lot to like in the new BICSI work, and we’ll share it at a very high level for you here.  Along the way we’ll be drawing comparisons to The Uptime Institute’s data center tier model.  This is because our readers are known to be familiar with that model and it’s therefore a useful barometer.  We in no way mean to slight either organization in doing so.

The BICSI-002 standard includes three Annex sections that are subtitled, “informative,” and as such is technically separate from the standard itself.  However, this is where BICSI delivers its guidance on this topic.  For our discussions here, we’ll acknowledge that disclaimer but still refer to this tier model as the “BICSI model,” for the sake of convenience.

The BICSI model establishes five facility availability classes- F0 through F4.  These somewhat align, by the way, to The Uptime Institute’s tier classification model if you agree that the lowest level (only basic environmental and energy requirements) is implied.  While the specification of each tier level is articulated differently than that of The Uptime Institute, what we think is valuable to the enterprise is the process steps involved in reaching a determination of which tier level is appropriate for the business.  This is in contrast to identifying a business model from a list of characteristics that most closely match the enterprise.

Rather than four broad definitions of business characteristics, the BICSI model leads one through an analysis of the mission critical risk level for the business applications, an estimate of corresponding downtime (planned and unplanned), impact analysis of downtime, to arrive at the necessary facility availability class.  The standard document even includes a worksheet reminiscent of a simple tax form that one can use for each mission critical application in the enterprise to map against a corresponding facility class (availability tier).

Each of the BICSI model’s availability classes are defined in terms of four areas:

1. Component redundancy increases reliability by providing redundancy for critical, low-reliability components within the systems.
2. System redundancy increases reliability even further by increasing redundancy at the system level.
3. Quality control ensures that high quality is designed into the facility and thus reduces risk of downtime due to failure during initial installation or premature wear.
4. Survivability refers to reduction of downtime risk by protecting against external events such as physical forces, security breaches, and natural disasters.

Given that, the five BICSI availability classes are defined as follows (and we borrow from the BICSI-002 specification).

Availability Class F0

The objective of Class F0 is to support the basic environmental and energy requirements of the IT functions without supplementary equipment, with avoidance of capital cost as the dominant driver.  In this case, there is a high risk of downtime due to planned and unplanned events.  However, in F0 facilities, it is assumed that maintenance can be performed during non-scheduled hours and long downtime has very little impact on operations.

Tactics for Class F0

Component redundancy:     none

System redundancy:              none

Quality control:                      standard commercial quality

Survivability:                           none

Application:   A critical power distribution system separate from the general use power does not exist in this model.  There is no backup generator and no redundancy of any kind for power or air conditioning exists.

Typical facility performance characteristics for Class F0

Annual maintenance windows:   >400 hours

Target availability:                           <99%

Scope of impact:                                Local data center

Availability Class F1

A Class F1 facility supports the basic environmental and energy requirements of the IT functions, and there is a high risk of downtime due to planned and unplanned events.  however, in Class F1 facilities, maintenance can be performed during nonscheduled hours and the impact of downtime is relatively low.

Tactics for Class F1

Component redundancy:     none

System redundancy:              none

Quality control:                      standard commercial quality

Survivability:                           none

Application:  The critical power distribution system would include power conditioning, but no redundancy of any kind is used for power or air conditioning.

Typical facility performance characteristics for Class F1

Annual maintenance windows:    100-400 hours

Target availability:                            <99%

Scope of impact:                                 Local or regional data centers

Availability Class F2

The objective of Class F2 is to provide a level of reliability higher than that of a Class F1 in order to reduce the risk of downtime due to component failure.  In a Class F2 facility, there is a moderate risk of downtime due to planned and unplanned events and maintenance activities can be performed during unscheduled hours.

Tactics for Class F2

Component redundancy:   provided for critical components

System redundancy:            none

Quality control:                    premium quality for critical components

Survivability:                         moderate hardening for physical security and structural integrity

Application:  In this Class, the critical power system includes redundancy in those parts of the electrical distribution system that are expected most likely to fail.  These would include any products that have a high parts count or moving parts, such as UPS, controls, air conditioning, and generators.  Additionally, it may be necessary to specify premium quality devices that provide longer life or better reliability.

Typical facility performance characteristics for Class F2

Annual maintenance windows:    50-99 hours

Target availability:                            99.9%

Scope of impact:                                 Local, regional, or multiregional data centers

Availability Class F3

The objective of Class F3 is to provide additional reliability and maintainability to reduce the risk of downtime due to natural disasters, human-driven disasters, planned maintenance, and repair activities.  Maintenance and repair activities will typically need to be performed during full production time with no opportunity for curtailed operations.

Tactics for Class F3

Component redundancy:   provided for critical and non critical components, as well as to increase maintainability. Redundancy is not provided where the compone is part of a redundant system already.

System redundancy:           may be provided without component redundancy

Quality control:                   premium quality for all components

Survivability:                        significant hardening for physical security and structural integrity

Application:   The critical power system must provide for reliable, continuous power even when major components or subsystems are out of service for repair or maintenance.  To protect against unplanned downtime, the power system must be ale to sustain operations while a dependent component or subsystem is out of service.

Typical facility performance characteristics for Class F3

Annual maintenance windows:      0=49 hours

Target availability:                              99.99%

Scope of impact:                                   all data centers, even some local data centers with high availability

requirements and low maintenance windows, may need to be Class F3.

Availability Class F4

The objective of Class F4 is to eliminate downtime through the application of all tactics to provide continuous operation regardless of planned or unplanned activities.  All recognizable single points of failure from the points of connection at the utility to the points of connection at the critical loads are eliminated.  Systems are typically automated to reduce the chances for human error and are staffed 24×7.  Rigorous training is provided for the staff to handle any contingency. Compartmentalization and fault tolerance are prime requirements for a Class F4 facility

Tactics for Class F4

Component redundancy:   provided for all critical components and to increase maintainability.¬† Redundancy is also provided for noncritical components.

System redundancy:           provided with component redundancy so tha overall reliability is maintained even during maintenance activities.

Quality control:                   premium quality for all components

Survivability:                        all building systems are self-supporting in any event and are protected against the highest levels of natural forces.

Application:  The critical power system in a Class F4 facility must provide for reliable, continuous power even when major components or subsystems are out of service for repair or maintenance.  to protect against unplanned downtime, the power system must be able to sustain operations while a dependent component or subsystem is out of service.

Typical facility performance characteristics for Class F4

Annual maintenance windows:     0 hours

Target availability:                             99.999%

Scope of impact:                                  multiregional or enterprise-wide data centers

One may notice that the availability class definitions touch upon the characteristics of the MEP infrastructure as well as the operational scenario applied to the facility.

This BICSI model does not directly map one-to-one with the tiers of The Uptime Institute model, but many comparisons are immediately noticeable.  In future articles, we will be drilling down on these comparisons in more detail.

We’d appreciate your comments and thoughts about the BICSI model, especially as it relates to your interest in the The Uptime Institute model.