p. 143

3 - 2 Evaluation of Reliability

3-2-1 Concept of Reliability

Various information processing systems are operated in a modern society. Conversely, it can be said that a modern society cannot be set up without an information processing system. It is but obvious that an information processing system playing such an important role is required to have high reliability. The concepts for constructing a highly reliable system are classified as shown below.

(1) Concept of a fault

(a) Fault tolerant (fault tolerant technique) This is a concept of creating a mechanism for minimizing the effect of a fault on the basis of the assumption that a fault will definitely occur in a device or system created by a human. A system that is constructed on the basis of this concept is called a fault tolerant system.

• • Fail-soft
  This is a type of fault-tolerance and is a concept based on which the operation is continued by partially restricting the effect of the fault. An example of this concept is degraded operation where the main important functions are not stopped even if several functions have been done away with. In the case of a power failure in a hospital, the lighting is kept to a minimum and the power of the private power generator is concentrated on artificial life support. · 1 question

(b) Fault avoidance (fault avoidance technique) This is a concept that a fault is not allowed to occur in a system at any cost. Since in reality it is not possible to prevent a fault from occurring, it can be referred to as a concept (or technique) that the possibility of the occurrence of a fault is brought as close to zero as possible.

(2) Concept of safety

(a) Fail-safe This is a concept of stopping the system in order to avoid the impact of the occurrence of a failure on other parts by giving priority to safety. An example of this · 1 question

p. 144

concept is the control system of a traffic signal. When a failure occurs, all signals are turned to red to prevent accidents.

(b) Foolproof This is a concept of preventing an operator from making a mistake so that the mistake of the person operating the system does not affect the overall system. Examples of this concept include washing machines that do not operate unless the lid is closed and the checking and correction of input data for any mistakes. · 1 question

3-2-2 Index of Reliability

There are indexes and indicators that evaluate such as the reliability of an information processing system. This set of indexes and indicators is called RASIS which is obtained by arranging together the first letter of the following five items. It is also called RAS which is obtained by arranging together the first letter of the top three items.

Among these five indexes and indicators, R (Reliability), A (Availability), and S (Serviceability) are evaluated by using the scale shown below.

(1) MTBF (Mean Time Between Failures) This is the mean operation time from the recovery of the system (or device) from a failure until the occurrence of the next failure. It is used to represent R (i.e., Reliability). (2) MTTR (Mean Time To Repair) This is the mean time from the occurrence of a failure in the system (or device) until its recovery. It is used to represent S (i.e., Serviceability). (3) Availability This is the probability of the normal operation of the system (or device) and is used to represent A (i.e., Availability). Availability indicates the ratio of normal operation of the system per unit time and can be calculated using MTBF and MTTR as shown · 2 questions

p. 145

below.

The probability that the system (or device) is not operating with respect to the probability that the system (or device) is operating (availability) is called non-availability. Non-availability indicates the ratio of the system failure per unit time and can be calculated as shown below.

p. 146

3-2-3 Availability and Failure Rate of the System

Generally, a system is configured by several devices. At this time, the availability of the overall system can be calculated from the availability of each device.

(1) Availability of a series system

In the case of a series system, the system operates normally as a whole only when all devices are operating. In the case of a series system shown in Figure 2-10, the operating status of each device is summarized as shown in the table below.

According to this table, the system operates as a whole only in the case of 1) when both devices are in the operating status. Therefore, the availability of a series system is calculated by the expression below.

Similarly, even if the number of devices is increased to three or four, the series system operates normally only when all devices are operating. Therefore, the availability of the series system is calculated as the result obtained by multiplying the availability of all devices that make up the

p. 147

system.

Since the availability is less than 1, the value keeps getting smaller with each repeated multiplication. That is, the higher the number of devices connected in a series system, the lower the availability becomes.

(2) Availability of a parallel system

In the case of a parallel system, the system may operate as a whole even when not all devices are operating. In the case of a parallel system shown in Figure 2-11, the operating status of each device is summarized as shown in the table below.

At this time, the functionality of the system is conserved (or the system operates as a whole) even if only either of the two devices is operating, and thus, the system operates as a whole in the case of 1) through 3). Thus, the availability of the overall system can be calculated by adding up the probability of 1) through 3). However, since in this way the calculation becomes tedious, it is assumed that among all operating conditions, the system is in a non-operating status only in the case of 4), and the total probability of 1) through 3) is calculated by subtracting the probability of

• 4) from the total probability “1” of all operating conditions. Therefore, the availability of a parallel
  

p. 148

system is calculated by the expression shown below.

Similarly, as shown in Figure 2-12, consider a case in which three or more devices are connected in parallel.

In this case, the following combinations can be assumed in consideration of the minimum number of devices with which the system operates normally as a whole.

In a parallel system, the more devices you connect, the higher the availability becomes.

p. 149

(3) Availability of a series/parallel mixed system

The availability of a system in which both a series system and a parallel system are mixed is calculated on a part-by-part basis, and then the availability of the overall system is calculated.

(4) Failure rate of the system

Failure rate is the probability of the occurrence of a failure (or a fault) in a device or system. Since a failure occurs once after the device has been operating for a fixed period of time, the failure rate of the individual device becomes the reciprocal of MTBF.

p. 150

Moreover, when several devices (or components) have been combined together, the overall failure rate becomes the total sum of the failure rate of each device (or component).

If the change in the failure rate of the hardware is represented by a graph with the elapsed time plotted on the horizontal axis and the failure rate (or occurrence frequency of failure) plotted on the vertical axis, this results in a bathtub curve in most cases.

p. 151