Reliability Requirements

Definition

A reliability requirement is a dependability requirement that specifies a required amount of reliability, which is a quality factor that is defined as follows:

Reliability

the degree to which something operates without failure under given conditions during a given time period.

Objectives

The typical objectives of a reliability requirement are to:

• Ensure that something will function properly for long periods without failure.
• Thereby minimize any unintentional disruptions in operation (i.e., unscheduled downtime).

Measurements

Reliability requirements are typically specified in terms of the following measurements:

• The mean time between failures (MTBF), whereby MTBF is defined as the average period of time that the application shall continue to function correctly without failure under stated conditions.
• The maximum acceptable probability of the failure during a given time period.
• The maximum permitted number of failures per unit time.

Examples

The following are typical examples of reliability requirements:

• “The application’s mean time between failures shall be at least 1 month.”
• “The mean time between failures for the normal paths through the purchase item use case shall be at least 4 months.”
• “The component’s mean time between failures shall be at least 1 year.”
• “The probability that the component fails shall not exceed .001% per year.”
• “The component shall not fail more than an average of 3 times per year.”

Guidelines

The following guidelines have been found to be useful when producing reliability requirements:

• The scope of a reliability requirement can be:
  • A business enterprise
  • An application
  • A component
  • A center
• Reliability requirements can be identified and specified in term of the following:
  
  

  Component of Requirement	Possibile Values
  Type of Failure	Failure to handle an input Failure to produce an output Failure to produce a correct output Total failure
  Scope of Failure	Total Localized to a functional area Localized to an external actor Localized to a user interface Localized to an API
  State when Failure Occurs	Normal operations Degraded mode
  Response Types	Notify selected users and other systems Log unscheduled downtime Shut down external server systems Switch to degraded mode Shutdown affected capabilities Totally shut down
  Response Measurements	Mean time between failures (MTBF) Mean time to fix (MTTF) Maximum acceptable probability of the failure during a given time period. Maximum acceptable number of failures per unit time.

  
  
• Reliablity requirements should be specified quantitatively.
• Because some capabilities of an application are more critical than others, reliability requirements should usually be restricted to its most important capabilities (e.g., use cases or use case paths). For example, the reliability of customer purchasing and credit card approvals should typically be higher than that of allowing a customer to change their password. Unfortunately in practice, the most time-critical capabilities are often those that are most complex and thus the most likely to contain defects that can lower their reliability.
• Reliability requirements should specify the conditions under which they apply:
  • The application or component capabilities, the reliability of which is being specified.
  • The application or component load (because relability may decrease as the number of simultaneous transactions or the size of the databases increase).
• Reliability requirements are difficult to validate, especially if they require high reliability.
  • One can use statistics to estimate the reliability of an application based on the reliability of its components (as well as certain assumptions about the independence of the failures).
  • One can estimate the reliability of an application or component based on ‘long term’ reliability testing prior to delivery, but the duration of the testing is often so limited by delivery schedules to allow very accurate estimation.
• Reliability is related to availability as follows:
  • High availability implies either high reliability or very low mean time to fix (MTTF).
  • Low availability may mean low reliability, high MTTF, or large amounts of scheduled downtime (e.g., for hardware or software upgrades).