Você está na página 1de 10

Hindawi Publishing Corporation

Journal of Applied Mathematics


Volume 2013, Article ID 482391, 10 pages
http://dx.doi.org/10.1155/2013/482391

Research Article
(𝑚, 𝑛)-Semirings and a Generalized Fault-Tolerance
Algebra of Systems

Syed Eqbal Alam,1 Shrisha Rao,2 and Bijan Davvaz3


1
College of Computers and Information Technology, Taif University, Taif 21974, Saudi Arabia
2
IIIT-Bangalore, Bangalore 560 100, India
3
Department of Mathematics, Yazd University, Yazd, Iran

Correspondence should be addressed to Syed Eqbal Alam; syed.eqbal@iiitb.net

Received 21 July 2012; Revised 31 December 2012; Accepted 31 December 2012

Academic Editor: Ray K. L. Su

Copyright © 2013 Syed Eqbal Alam et al. This is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly
cited.

We propose a new class of mathematical structures called (𝑚, 𝑛)-semirings (which generalize the usual semirings) and describe
their basic properties. We define partial ordering and generalize the concepts of congruence, homomorphism, and so forth, for
(𝑚, 𝑛)-semirings. Following earlier work by Rao (2008), we consider systems made up of several components whose failures may
cause them to fail and represent the set of such systems algebraically as an (𝑚, 𝑛)-semiring. Based on the characteristics of these
components, we present a formalism to compare the fault-tolerance behavior of two systems using our framework of a partially
ordered (𝑚, 𝑛)-semiring.

1. Introduction and by Dudek [21]. Recently the generalization of algebraic


structures is studied Davvaz et al. [22, 23].
Fault tolerance is the property of a system to be functional In this paper, we first define the (𝑚, 𝑛)-semiring (R, 𝑓, 𝑔)
even if some of its components fail. It is a very critical (which is a generalization of the ordinary semiring (R, +, ×),
issue in the design of the systems as in Air Traffic Control where R is a set with binary operations + and ×), using 𝑓 and
Systems [1, 2], real-time embedded systems [3], robotics [4, 𝑔 which are 𝑚-ary and 𝑛-ary operations, respectively. We pro-
5], automation systems [6, 7], medical systems [8], mission pose identity elements, multiplicatively absorbing elements,
critical systems [9], and a lot of others. Description of fault- idempotents, and homomorphisms for (𝑚, 𝑛)-semirings. We
tolerance modeling using algebraic structures is proposed also briefly touch on zero-divisor free, zero-sum free, addi-
by Beckmann [10] for groups and by Hadjicostis [11] for tively cancellative, and multiplicatively cancellative (𝑚, 𝑛)-
semigroups and semirings. Semirings are also used in other semirings and the congruence relation on (𝑚, 𝑛)-semirings.
areas of computer science like cryptography [12], databases In Section 4, we use the facts that each system consists
[13], graph theory, game theory [14], and so forth. Rao [15] of components or subsystems and that the fault-tolerance
uses the formalism of semirings to analyze the fault tolerance behavior of the system depends on each of the components or
of a system as a function of its composition, with a partial subsystems that constitute the system. A system may itself be
ordering relation between systems used to compare their a module or part of a larger system, so that its fault tolerance
fault-tolerance behaviors. affects that of the whole system of which it is a part. We
The generalization of algebraic structures was in active analyze the fault tolerance of a system given its composition,
research for a long time; Timm [16] in 1967 proposed com- extending earlier work of Rao [15]. Section 2 describes the
mutative 𝑛-groups; later Crombez [17] in 1972 generalized notations used and the general conventions followed.
rings and named it as (𝑚, 𝑛)-rings. It was further studied Section 3 deals with the definition and properties of
by Crombez and Timm [18], Leeson and Butson [19, 20], (𝑚, 𝑛)-semirings. In Section 4, we extend the results of
2 Journal of Applied Mathematics

Rao [15] using a partial ordering on the (𝑚, 𝑛)-semiring of (iii) Let 𝑥1 , 𝑥2 , . . . , 𝑥𝑛 , 𝑎1 , 𝑎2 , . . . , 𝑎𝑚 be elements of set R,
systems: the class of systems is algebraically represented by and 1 ≤ 𝑖 ≤ 𝑛. The 𝑛-ary operation 𝑔 is distributive
an (𝑚, 𝑛)-semiring, and the fault-tolerance behavior of two with respect to the 𝑚-ary operation 𝑓 if
systems is compared using partially ordered (𝑚, 𝑛)-semiring.
𝑔 (𝑥1𝑖−1 , 𝑓 (𝑎1𝑚 ) , 𝑥𝑖+1
𝑛
)
2. Preliminaries
= 𝑓 (𝑔 (𝑥1𝑖−1 , 𝑎1 , 𝑥𝑖+1
𝑛
),..., (7)
The set of integers is denoted by Z, with Z+ and Z− denoting
the sets of positive integers and negative integers, respectively, 𝑔 (𝑥1𝑖−1 , 𝑎𝑚 , 𝑥𝑖+1
𝑛
)) .
and 𝑚 and 𝑛 used are positive integers. Let R be a set and 𝑓
a mapping 𝑓 : R𝑚 → R; that is, 𝑓 is an 𝑚-ary operation.
Remark 3. (i) An 𝑚-ary semigroup (R, 𝑓) is called a semia-
Elements of the set R are denoted by 𝑥𝑖 , 𝑦𝑖 where 𝑖 ∈ Z+ .
belian or (1, 𝑚)-commutative if
Definition 1. A nonempty set R with an 𝑚-ary operation 𝑓 is
𝑓(𝑥, ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
𝑎, . . . , 𝑎, 𝑦) = 𝑓(𝑦, ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
𝑎, . . . , 𝑎, 𝑥),
called an 𝑚-ary groupoid and is denoted by (R, 𝑓) (see Dudek 𝑚−2 𝑚−2
(8)
[24]).
for all 𝑥, 𝑦, 𝑎 ∈ R (from Dudek and Mukhin [26]).
We use the following general convention.
The sequence 𝑥𝑖 , 𝑥𝑖+1 , . . . , 𝑥𝑚 is denoted by 𝑥𝑖𝑚 where (ii) Consider a 𝑘-ary group (𝐺, ℎ) in which the 𝑘-ary
1 ≤ 𝑖 ≤ 𝑚. operation ℎ is distributive with respect to itself, that is,
For all 1 ≤ 𝑖 ≤ 𝑗 ≤ 𝑚, the following term
ℎ (𝑥1𝑖−1 , ℎ (𝑎1𝑘 ) , 𝑥𝑖+1
𝑘
)
𝑓 (𝑥1 , . . . , 𝑥𝑖 , 𝑦𝑖+1 , . . . , 𝑦𝑗 , 𝑧𝑗+1 , . . . , 𝑧𝑚 ) (1) (9)
= ℎ (ℎ (𝑥1𝑖−1 , 𝑎1 , 𝑥𝑖+1
𝑘
) , . . . , ℎ (𝑥1𝑖−1 , 𝑎𝑘 , 𝑥𝑖+1
𝑘
)) ,
is represented as
𝑗
for all 1 ≤ 𝑖 ≤ 𝑘. These types of groups are called autodistrib-
𝑓 (𝑥1𝑖 , 𝑦𝑖+1 , 𝑧𝑗+1
𝑚
). (2) utive 𝑘-ary groups (see Dudek [27]).
In the case when 𝑦𝑖+1 = ⋅ ⋅ ⋅ = 𝑦𝑗 = 𝑦, (2) is expressed as
3. (𝑚, 𝑛)-Semirings and Their Properties
(𝑗−𝑖) 𝑚
𝑓(𝑥1𝑖 , 𝑦 , 𝑧𝑗+1 ). (3) Definition 4. An (𝑚, 𝑛)-semiring is an algebraic structure
(R, 𝑓, 𝑔) which satisfies the following axioms:
Definition 2. Let 𝑥1 , 𝑥2 , . . . , 𝑥2𝑚−1 be elements of set R.
(i) (R, 𝑓) is an 𝑚-ary semigroup,
(i) Then, the associativity and distributivity laws for the
𝑚-ary operation 𝑓 are defined as follows. (ii) (R, 𝑔) is an 𝑛-ary semigroup,
(iii) the 𝑛-ary operation 𝑔 is distributive with respect to the
(a) Associativity: 𝑚-ary operation 𝑓.
𝑓 (𝑥1𝑖−1 , 𝑓 (𝑥𝑖𝑚+𝑖−1 ) , 𝑥𝑚+𝑖
2𝑚−1
) Example 5. Let B be any Boolean algebra. Then, (B, 𝑓, 𝑔)
(4) is an (𝑚, 𝑛)-semiring where 𝑓(𝐴𝑚1 ) = 𝐴 1 ∪ 𝐴 2 ∪ ⋅ ⋅ ⋅ ∪ 𝐴 𝑚
𝑗−1 𝑚+𝑗−1 2𝑚−1
= 𝑓 (𝑥1 , 𝑓 (𝑥𝑗 ) , 𝑥𝑚+𝑗 ), and 𝑔(𝐵1𝑛 ) = 𝐵1 ∩ 𝐵2 ∩ ⋅ ⋅ ⋅ ∩ 𝐵𝑛 , for all 𝐴 1 , 𝐴 2 , . . . , 𝐴 𝑚 and
𝐵1 , 𝐵2 , . . . , 𝐵𝑛 ∈ B.
for all 𝑥1 , . . . , 𝑥2𝑚−1 ∈ R, for all 1 ≤ 𝑖 ≤ 𝑗 ≤ 𝑚
(from Gluskin [25]). In general, we have the following.
(b) Commutativity:
Theorem 6. Let (R, +, ×) be an ordinary semiring. Let 𝑓 be an
𝑓 (𝑥1 , 𝑥2 , . . . , 𝑥𝑚 )
𝑚-ary operation and 𝑔 be an 𝑛-ary operation on R as follows:
(5)
= 𝑓 (𝑥𝜂(1) , 𝑥𝜂(2) , . . . , 𝑥𝜂(𝑚) ) , 𝑚
𝑓 (𝑥1𝑚 ) = ∑𝑥𝑖 , ∀𝑥1 , 𝑥2 , . . . , 𝑥𝑚 ∈ R,
for every permutation 𝜂 of {1, 2, . . . , 𝑚} (from 𝑖=1
Timm [16]), ∀𝑥1 , 𝑥2 , . . . , 𝑥𝑚 ∈ R. 𝑛
(10)

(ii) An 𝑚-ary groupoid (R, 𝑓) is called an 𝑚-ary semi- 𝑔 (𝑦1𝑛 ) = ∏𝑦𝑖 , ∀𝑦1 , 𝑦2 , . . . , 𝑦𝑛 ∈ R.
𝑖=1
group if 𝑓 is associative (from Dudek [24]); that is, if
Then, (R, 𝑓, 𝑔) is an (𝑚, 𝑛)-semiring.
𝑓 (𝑥1𝑖−1 , 𝑓 (𝑥𝑖𝑚+𝑖−1 ) , 𝑥𝑚+𝑖
2𝑚−1
)
(6) Proof. Omitted as obvious.
𝑗−1 𝑚+𝑗−1 2𝑚−1
= 𝑓 (𝑥1 , 𝑓 (𝑥𝑗 ) , 𝑥𝑚+𝑗 ),
Example 7. The following give us some (𝑚, 𝑛)-semirings in
for all 𝑥1 , . . . , 𝑥2𝑚−1 ∈ R, where 1 ≤ 𝑖 ≤ 𝑗 ≤ 𝑚. different ways indicated by Theorem 6.
Journal of Applied Mathematics 3

(i) Let (R, +, ×) be an ordinary semiring and 𝑥1 , Elements 𝑥1 , 𝑥2 , . . . , 𝑥𝑛−1 ∈ R are called left zero-
𝑥2 , . . . , 𝑥𝑛 be in R. If we set divisors of (𝑚, 𝑛)-semiring (R, 𝑓, 𝑔) if there exists
𝑎 ≠ 0 and the following holds:
𝑔 (𝑥1𝑛 ) = 𝑥1 × 𝑥2 × ⋅ ⋅ ⋅ × 𝑥𝑛 , (11)
𝑔 (𝑥1𝑛−1 , 𝑎) = 0. (18)
we get a (2, 𝑛)-semiring (R, +, 𝑔).
(iii) (R, 𝑓, 𝑔) is called zero-sum free if
(ii) In an (𝑚, 𝑛)-semiring (R, 𝑓, 𝑔), fixing elements 𝑎2𝑚−1
and 𝑏2𝑛−1 , we obtain two binary operations as follows: 𝑓 (𝑥1 , 𝑥2 , . . . , 𝑥𝑚 ) = 0 (19)

𝑥 ⊕ 𝑦 = 𝑓 (𝑥, 𝑎2𝑚−1 , 𝑦) , always implies 𝑥1 = 𝑥2 = ⋅ ⋅ ⋅ = 𝑥𝑚 = 0.


(12) (iv) (R, 𝑓, 𝑔) is called additively cancellative if the 𝑚-ary
𝑥 ⊗ 𝑦 = 𝑔 (𝑥, 𝑏2𝑛−1 , 𝑦) . semigroup (R, 𝑓) is cancellative, that is,
𝑓 (𝑥1𝑖−1 , 𝑎, 𝑥𝑖+1
𝑚
) = 𝑓 (𝑥1𝑖−1 , 𝑏, 𝑥𝑖+1
𝑚
) 󳨐⇒ 𝑎 = 𝑏, (20)
Obviously, (R, ⊕, ⊗) is a semiring.
(iii) The set Z− of all negative integers is not closed under for all 𝑎, 𝑏, 𝑥1 , 𝑥2 , . . . , 𝑥𝑚 ∈ R and for all 1 ≤ 𝑖 ≤ 𝑚.
the binary products; that is, Z− does not form a (v) (R, 𝑓, 𝑔) is called multiplicatively cancellative if the 𝑛-
semiring, but it is a (2, 3)-semiring. ary semigroup (R, 𝑔) is cancellative, that is,
Definition 8. Let (R, 𝑓, 𝑔) be an (𝑚, 𝑛)-semiring. Then 𝑚-ary 𝑔 (𝑥1𝑖−1 , 𝑎, 𝑥𝑖+1
𝑛
) = 𝑔 (𝑥1𝑖−1 , 𝑏, 𝑥𝑖+1
𝑛
) 󳨐⇒ 𝑎 = 𝑏, (21)
semigroup (R, 𝑓) has an identity element 0 if
for all 𝑎, 𝑏, 𝑥1 , 𝑥2 , . . . , 𝑥𝑛 ∈ R and for all 1 ≤ 𝑖 ≤ 𝑛.
𝑥 = 𝑓(0, . . . , 0, 𝑥, ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ 0, . . . , 0),
𝑖−1 𝑚−𝑖
(13) Elements 𝑥1 , 𝑥2 , . . . , 𝑥𝑛−1 are called left cancellable in
an 𝑛-ary semigroup (R, 𝑔) if
for all 𝑥 ∈ R and 1 ≤ 𝑖 ≤ 𝑚. We call 0 as an identity element
of (𝑚, 𝑛)-semiring (R, 𝑓, 𝑔). 𝑔 (𝑥1𝑛−1 , 𝑎) = 𝑔 (𝑥1𝑛−1 , 𝑏) 󳨐⇒ 𝑎 = 𝑏, (22)
Similarly, 𝑛-ary semigroup (R, 𝑔) has an identity ele-
ment 1 if for all 𝑥1 , 𝑥2 , . . . , 𝑥𝑛−1 , 𝑎, 𝑏 ∈ R.
(R, 𝑓, 𝑔) is called multiplicatively left cancellative if
𝑦 = 𝑔(1, . . . , 1, 𝑦, ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ 1, . . . , 1), elements 𝑥1 , 𝑥2 , . . . , 𝑥𝑛−1 ∈ R\{0} are multiplicatively
(14)
𝑗−1 𝑛−𝑗
left cancellable in 𝑛-ary semigroup (R, 𝑔).
for all 𝑦 ∈ R and 1 ≤ 𝑗 ≤ 𝑛.
We call 1 as an identity element of (𝑚, 𝑛)-semiring Theorem 11. Let (R, 𝑓, 𝑔) be an (𝑚, 𝑛)-semiring with 𝑓-
(R, 𝑓, 𝑔). identity 0.
We therefore call 0 the 𝑓-identity, and 1 the 𝑔-identity. (i) If elements 𝑥1 , 𝑥2 , . . . , 𝑥𝑛−1 ∈ R are multiplicatively
left cancellable, then elements 𝑥1 ,𝑥2 ,. . .,𝑥𝑛−1 are not left
Remark 9. In an (𝑚, 𝑛)-semiring (R, 𝑓, 𝑔), placing 0 and 1, divisors.
(𝑚−2) and (𝑛−2) times, respectively, we obtain the following
(ii) If the (𝑚, 𝑛)-semiring (R, 𝑓, 𝑔) is multiplicatively left
binary operations:
cancellative, then it is zero-divisor free.
𝑥 + 𝑦 = 𝑓(𝑥, 0, . . . , 0, 𝑦),
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
𝑚−2 We have generalized Theorem 11 from Theorem 4.4 of
Hebisch and Weinert [28].
𝑥 × 𝑦 = 𝑔(𝑥, ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
1, . . . , 1, 𝑦), (15) We have generalized the definition of idempotents of
𝑛−2
semirings given by Bourne [29] and Hebisch and Weinert
∀𝑥, 𝑦 ∈ R. [28]), as follows.

Definition 10. Let (R, 𝑓, 𝑔) be an (𝑚, 𝑛)-semiring with an 𝑓- Definition 12. Let (R, 𝑓, 𝑔) be an (𝑚, 𝑛)-semiring. Then,
identity element 0 and 𝑔-identity element 1. Then, (i) it is called additively idempotent if (R, 𝑓) is an
(i) 0 is said to be multiplicatively absorbing if it is absorb- idempotent 𝑚-ary semigroup, that is, if
ing in (R, 𝑔), that is, if 𝑓(𝑥, 𝑥,⏟⏟⏟
⏟⏟⏟⏟⏟⏟⏟⏟ ⏟⏟) = 𝑥,
. .⏟⏟⏟⏟⏟⏟
.,𝑥 (23)
𝑚
𝑔 (0, 𝑥1𝑛−1 ) = 𝑔 (𝑥1𝑛−1 , 0) = 0, (16)
for all 𝑥 ∈ R;
for all 𝑥1 , 𝑥2 , . . . , 𝑥𝑛−1 ∈ R. (ii) it is called multiplicatively idempotent if (R, 𝑔) is an
(ii) (R, 𝑓, 𝑔) is called zero-divisor free if idempotent 𝑛-ary semigroup, that is, if
𝑔(𝑦, 𝑦, . . . , 𝑦) = 𝑦,
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
𝑔 (𝑥1 , 𝑥2 , . . . , 𝑥𝑛 ) = 0 (17) 𝑛
(24)

always implies 𝑥1 = 0 or 𝑥2 = 0 or ⋅ ⋅ ⋅ or 𝑥𝑛 = 0. for all 𝑦 ∈ R, 𝑦 ≠ 0.


4 Journal of Applied Mathematics

Theorem 13. An (𝑚, 𝑛)-semiring (R, 𝑓, 𝑔) having at least Theorem 15. Let (R, 𝑓, 𝑔) be an (𝑚, 𝑛)-semiring and the
two multiplicatively idempotent elements in the center is not relation 𝜎 be a congruence relation on (R, 𝑓, 𝑔). Then, the
multiplicatively cancellative. quotient (R/𝜎, 𝐹, 𝐺) is an (𝑚, 𝑛)-semiring under 𝐹((𝑥1 )/𝜎,
. . . , (𝑥𝑚 )/𝜎) = 𝑓(𝑥1𝑚 )/𝜎 and 𝐺((𝑦1)/𝜎, . . . , (𝑦𝑛)/𝜎) = 𝑔(𝑦1𝑛 )/𝜎,
Proof. Let 𝑎 and 𝑏 be two multiplicatively idempotent ele- for all 𝑥1 , 𝑥2 , . . . , 𝑥𝑚 and 𝑦1 , 𝑦2 , . . . , 𝑦𝑛 in R.
ments in the center, 𝑎 ≠ 𝑏. Then,
Proof. Omitted as obvious.
𝑔(1, . . . , 1, 𝑎, 𝑏) = 𝑔(1,
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ . . . , 1, 𝑏, 𝑎),
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ (25)
𝑛−2 𝑛−2
Definition 16. We define homomorphism, isomorphism, and
which can be written as follows: a product of two mappings as follows.
(𝑛) (𝑛) (i) A mapping 𝜑 : R → S from (𝑚, 𝑛)-semiring
𝑔(1, . . . , 1, 𝑔 (𝑎), 𝑏) = 𝑔(1,
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ . . . , 1, 𝑔 (𝑏), 𝑎),
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ (26) (R, 𝑓, 𝑔) into (𝑚, 𝑛)-semiring (S, 𝑓󸀠 , 𝑔󸀠 ) is called a
𝑛−2 𝑛−2 homomorphism if
which is represented as 𝜑 (𝑓 (𝑥1𝑚 )) = 𝑓󸀠 (𝜑 (𝑥1 ) , 𝜑 (𝑥2 ) , . . . , 𝜑 (𝑥𝑚 )) ,
(𝑛−1) (30)
𝑔(1, . . . , 1, 𝑔(1, 𝑎 ), 𝑎, 𝑏)
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ 𝜑 (𝑔 (𝑦1𝑛 )) = 𝑔󸀠 (𝜑 (𝑦1 ) , 𝜑 (𝑦2 ) , . . . , 𝜑 (𝑦𝑛 )) ,
𝑛−3
(27) for all 𝑥1 , 𝑥2 , . . . , 𝑥𝑚 , 𝑦1 , 𝑦2 , . . . , 𝑦𝑛 ∈ R.
(𝑛−1)
= 𝑔(1, . . . , 1, 𝑔(1, 𝑏 ), 𝑏, 𝑎).
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ (ii) The (𝑚, 𝑛)-semirings (R, 𝑓, 𝑔) and (S, 𝑓󸀠 , 𝑔󸀠 ) are
𝑛−3 called isomorphic if there exists one-to-one homo-
If the (𝑚, 𝑛)-semiring (R, 𝑓, 𝑔) is multiplicatively can- morphism from R onto S. One-to-one homomor-
cellative, then the following holds true: phism is called isomorphism.
(iii) If we apply mapping 𝜑 : R → S and then 𝜓 : S →
(𝑛−1)
𝑔(1, . . . , 1, 𝑔(1, 𝑎 ), 1, 1)
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ T on 𝑥, we get the mapping (𝜓 ∘ 𝜑)(𝑥) which is equal
𝑛−3 to 𝜓(𝜑(𝑥)), where 𝑥 ∈ R. It is called the product of 𝜓
and 𝜑 [28].
(𝑛−1)
= 𝑔(1, . . . , 1, 𝑔(1, 𝑏 ), 1, 1),
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ (28)
𝑛−3
We have generalized Definition 16 from Definition 2 of
Allen [31].
(𝑛−1) (𝑛−1) We have generalized the following theorem from Theo-
𝑔(1, 𝑎 ) = 𝑔(1, 𝑏 ),
rem 3.3 given by Hebisch and Weinert [28].
which implies that 𝑎 = 𝑏, which is a contradiction to the
assumption that 𝑎 ≠ 𝑏; therefore, (R, 𝑓, 𝑔) is not multiplica- Theorem 17. Let (R, 𝑓, 𝑔), (S, 𝑓󸀠 , 𝑔󸀠 ), and (T, 𝑓󸀠󸀠 , 𝑔󸀠󸀠 ) be
tively cancellative. (𝑚, 𝑛)-semirings. Then, if the following mappings 𝜑 :
(R, 𝑓, 𝑔) → (S, 𝑓󸀠 , 𝑔󸀠 ) and
We have generalized Exercise 2.7 in Chapter I of Hebisch
𝜓 : (S, 𝑓󸀠 , 𝑔󸀠 ) → (T, 𝑓󸀠󸀠 , 𝑔󸀠󸀠 ) are homomorphisms;
and Weinert [28] to get the following.
then,
Definition 14. Let (R, 𝑓, 𝑔) be an (𝑚, 𝑛)-semiring and 𝜎 an 𝜓 ∘ 𝜑 : (R, 𝑓, 𝑔) → (T, 𝑓󸀠󸀠 , 𝑔󸀠󸀠 ) is also a homomor-
equivalence relation on R. phism.
(i) Then, 𝜎 is called a congruence relation or a congruence Proof. Let 𝑥1 , 𝑥2 , . . . , 𝑥𝑚 and 𝑦1 , 𝑦2 , . . . , 𝑦𝑛 be in R. Then
of (R, 𝑓, 𝑔), if it satisfies the following properties for
all 1 ≤ 𝑖 ≤ 𝑚 and 1 ≤ 𝑗 ≤ 𝑛: (𝜓 ∘ 𝜑) (𝑓 (𝑥1𝑚 )) = 𝜓 (𝜑 (𝑓 (𝑥1 , 𝑥2 , . . . , 𝑥𝑚 )))

(a) if 𝑥𝑖 𝜎𝑦𝑖 then 𝑓(𝑥1𝑚 )𝜎𝑓(𝑦1𝑚 ), = 𝜓 (𝑓󸀠 (𝜑 (𝑥1 ) , 𝜑 (𝑥2 ) , . . . , 𝜑 (𝑥𝑚 )))
(b) if 𝑧𝑗 𝜎𝑢𝑗 then 𝑔(𝑧1𝑛 )𝜎𝑔(𝑢1𝑛 ), = 𝑓󸀠󸀠 (𝜓 (𝜑 (𝑥1 )) , 𝜓 (𝜑 (𝑥2 )) , . . . , 𝜓 (𝜑 (𝑥𝑚 )))
for all 𝑥1 , 𝑥2 , . . . , 𝑥𝑚 , 𝑦1 , 𝑦2 , . . . , 𝑦𝑚 , 𝑧1 , 𝑧2 , . . . ,
𝑧𝑛 , 𝑢1 , 𝑢2 , . . . , 𝑢𝑛 ∈ R. = 𝑓󸀠󸀠 ((𝜓 ∘ 𝜑) (𝑥1 ) , (𝜓 ∘ 𝜑) (𝑥2 ) , . . . , (𝜓 ∘ 𝜑) (𝑥𝑚 )) .
(31)
(ii) Let 𝜎 be a congruence on an algebra R. Then, the quo-
tient of R by 𝜎, written as R/𝜎, is the algebra whose In a similar manner, we can deduce that
universe is R/𝜎 and whose fundamental operation (𝜓 ∘ 𝜑) (𝑔 (𝑦1𝑛 ))
satisfies
= 𝑔󸀠󸀠 ((𝜓 ∘ 𝜑) (𝑦1 ) , (𝜓 ∘ 𝜑) (𝑦2 ) , . . . , (𝜓 ∘ 𝜑) (𝑦𝑛 )) .
𝑓R (𝑥1 , 𝑥2 , . . . , 𝑥𝑚 )
𝑓R/𝜎 (𝑥1 , 𝑥2 , . . . , 𝑥𝑚 ) = , (29) (32)
𝜎
Thus, it is evident that 𝜓 ∘ 𝜑 is a homomorphism from
where 𝑥1 , 𝑥2 , . . . , 𝑥𝑚 ∈ R [30]. R → T.
Journal of Applied Mathematics 5

This proof is similar to that of Theorem 6.5 given by Burris ..


.
and Sankappanavar [30].
= 𝑓(𝑓(𝑥1 , 𝑥2 , . . . , 𝑥𝑚−1 , 0), 𝑥𝑚 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
0, . . . , 0)
Definition 18. Let (R, 𝑓, 𝑔) and (S, 𝑓󸀠 , 𝑔󸀠 ) be (𝑚, 𝑛)- 𝑚−2
semirings and 𝜑 : R → S a homomorphism. Then, the
kernel of 𝜑, written as ker 𝜑, is defined as follows: = 𝑓(𝑓(𝑥1 , 𝑥2 , . . . , 𝑥𝑚−1 , 𝑥𝑚 ), ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
0, . . . , 0)
𝑚−1

ker 𝜑 = {(𝑎, 𝑏) ∈ R × R | 𝜑 (𝑎) = 𝜑 (𝑏)} . (33) = 𝑓 (𝑥1 , 𝑥2 , . . . , 𝑥𝑚 ) .


(35)
Generalization of Burris and Sankappanavar [30].
(ii) Similar to part (i).
Theorem 19. Let (R, 𝑓, 𝑔) and (S, 𝑓󸀠 , 𝑔󸀠 ) be (𝑚, 𝑛)-semirings
and 𝜑 : R → S a homomorphism. Then, ker 𝜑 is a con-
4. Partial Ordering on Fault Tolerance
gruence relation on R, and there exists a unique one-to-one
homomorphism 𝜓 from R/ ker 𝜑 into S. In this sections we use 𝑥𝑖 , 𝑦𝑖 , and so forth, where 𝑖 ∈ Z+
to denote individual system components that are assumed
Proof. Omitted as obvious. to be atomic at the level of discussion; that is, they have no
components or subsystems of their own. We use component
Corollary 20. Let (R, 𝑓, 𝑔) be an (𝑚, 𝑛)-semiring and 𝜌 and
to refer to such an atomic part of a system, and subsystem to
𝜎 congruence relations on R, with 𝜌 ⊆ 𝜎. Then, 𝜎/𝜌 =
refer to a part of a system that is not necessarily atomic. We
{𝜌(𝑥), 𝜌(𝑦) | (𝑥, 𝑦) ∈ 𝜎} is a congruence relation on R/𝜌,
assume that components and subsystems are disjoint, in the
and (R/𝜌)/(𝜎/𝜌) ≅ R/𝜎.
sense that if they fail, they fail independently and do not affect
Lemma 21. Let 𝑥1 , 𝑥2 , . . . , 𝑥𝑚 , 𝑦1 , 𝑦2 , . . . , 𝑦𝑛 ∈ R. Then, the functioning of other components.
Let U be a universal set of all systems in the domain of
discourse as given by Rao [15], and let 𝑓 be a mapping 𝑓 :
𝑓(𝑓( . . . 𝑓(𝑓(𝑥1 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
(i) ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ 0, . . . , 0), 𝑥2 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
0, . . . , 0), . . . ),
𝑚 𝑚−1 𝑚−2
U𝑚 → U, that is, 𝑓 is an 𝑚-ary operation. Likewise, let 𝑔 be
𝑥𝑚 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
0, . . . , 0) = 𝑓(𝑥1 , 𝑥2 , . . . , 𝑥𝑚 ), an 𝑛-ary operation.
𝑚−2
Definition 22. We define 𝑓 and 𝑔 operations for systems as
𝑔(𝑔( . . . 𝑔(𝑔(𝑦1 , 1,
(ii) ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ . . . , 1), 𝑦2 , 1,
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ . . . , 1), . . . ),
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ follows.
𝑛 𝑛−1 𝑛−2
𝑦𝑛 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
1, . . . , 1) = 𝑔(𝑦1 , 𝑦2 , . . . , 𝑦𝑛 ). (i) 𝑓 is an 𝑚-ary operation which applies on systems
𝑛−2 made up of 𝑚 components or subsystems, where if
any one of the components or subsystems fails, then
Proof. (i) the whole system fails.
If a system made up of 𝑚 components 𝑥1 , 𝑥2 , . . . , 𝑥𝑚 ,
𝑓(𝑓( . . . 𝑓(𝑓(𝑥1 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ 0, . . . , 0), 𝑥2 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
0, . . . , 0), . . . ), 𝑥𝑚 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
0, . . . , 0). then, the system over operation 𝑓 is represented
𝑚 𝑚−1 𝑚−2 𝑚−2
as 𝑓(𝑥1 , 𝑥2 , . . . , 𝑥𝑚 ) for all 𝑥1 , 𝑥2 , . . . , 𝑥𝑚 ∈ U.
(34)
The system 𝑓(𝑥1 , 𝑥2 , . . . , 𝑥𝑚 ) fails when any of the
components 𝑥1 , 𝑥2 , . . . , 𝑥𝑚 fails.
By associativity (Definition 2 (i)), (34) is equal to
(ii) 𝑔 is an 𝑛-ary operation which applies on a system
consisting of 𝑛 components or subsystems, which fails
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ 0, . . . , 0), 𝑥1 , 𝑥2 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
𝑓(𝑓( . . . 𝑓(𝑓(0, ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ 0, . . . , 0), . . . ), 𝑥𝑚 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
0, . . . , 0)
𝑚 𝑚−1 𝑚−3 𝑚−2
if all the components or subsystems fail; otherwise
it continues working even if a single component or
= ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
𝑓(𝑓( . . . 𝑓(𝑓(0, 𝑥1 , 𝑥2 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
0, . . . , 0), 𝑥3 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
0, . . . , 0), . . . ), subsystem is working properly.
𝑚−1 𝑚−3 𝑚−2
Let a system consist of 𝑛 components 𝑥1 , 𝑥2 , . . . , 𝑥𝑛 ,
𝑥𝑚 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
0, . . . , 0) then, the system over operation 𝑔 is represented
𝑚−2 as 𝑔(𝑥1 , 𝑥2 , . . . , 𝑥𝑛 ) for all 𝑥1 , 𝑥2 , . . . , 𝑥𝑛 ∈ U. The
system 𝑔(𝑥1 , 𝑥2 , . . . , 𝑥𝑛 ) fails when all the components
= ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
𝑓(𝑓( . . . 𝑓(𝑓(0, . . . , 0), 𝑥1 , 𝑥2 , 𝑥3 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ 0, . . . , 0), . . . ), 𝑥1 , 𝑥2 , . . . , 𝑥𝑛 fail.
𝑚−1 𝑚 𝑚−4

Consider a partial ordering relation ≼ on U, such that


𝑥𝑚 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
0, . . . , 0)
𝑚−2
(U, ≼) is a partially ordered set (poset). This is a fault-
tolerance partial ordering where 𝑓(𝑥1𝑚 ) ≼ 𝑓(𝑦1𝑚 ) means
= ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
𝑓(𝑓( . . . 𝑓(𝑓(0, 𝑥1 , 𝑥2 , 𝑥3 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
0, . . . , 0), 𝑥4 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
0, . . . , 0), . . . ), that 𝑓(𝑥1𝑚 ) has a lower measure of some fault metric than
𝑚−2 𝑚−4 𝑚−2 𝑓(𝑦1𝑚 ) and 𝑓(𝑥1𝑚 ) has a better fault tolerance than 𝑓(𝑦1𝑚 ), for
𝑥𝑚 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
0, . . . , 0) all 𝑓(𝑥1𝑚 ), 𝑓(𝑦1𝑚 ) ∈ U (see Rao [32] for more details) and
𝑚−2 𝑥1 , 𝑥2 , . . . , 𝑥𝑚 , 𝑦1 , 𝑦2 , . . . , 𝑦𝑚 are disjoint components.
6 Journal of Applied Mathematics

Assume that 0 represents the atomic system “which is Lemma 27. If ≼ is a fault-tolerance partial order and
always up” and 1 represents the system “which is always 𝑥1 , 𝑥2 , . . . , 𝑥𝑚 , 𝑦1 , 𝑦2 , . . . , 𝑦𝑚 , 𝑧1 , 𝑧2 , . . . , 𝑧𝑛 , 𝑢1 , 𝑢2 , . . . , 𝑢𝑛 are
down” (see Rao [32]). disjoint components, which are in U, where 𝑚, 𝑛 ∈ Z+ , then
for all 1 ≤ 𝑖 ≤ 𝑚 and 1 ≤ 𝑗 ≤ 𝑛 the following holds true:
Observation 23. We observe the following for all disjoint
components 𝑥1 , 𝑥2 , . . . , 𝑥𝑚 , 𝑦1 , 𝑦2 , . . . , 𝑦𝑛 , which are in U. (i) if 𝑥𝑖 ≼ 𝑦𝑖 , then 𝑓(𝑥1𝑚 ) ≼ 𝑓(𝑦1𝑚 ),
𝑗−1 𝑛 (ii) if 𝑧𝑗 ≼ 𝑢𝑗 , then 𝑔(𝑧1𝑛 ) ≼ 𝑔(𝑢1𝑛 ).
(i) 𝑔(𝑦1 , 0, 𝑦𝑗+1 ) = 0 for all 1 ≤ 𝑗 ≤ 𝑛.
This is so since 0 represents the component or system Proof. (i) Since 𝑥𝑖 ≼ 𝑦𝑖 for all 1 ≤ 𝑖 ≤ 𝑚, we have
which never fails, and as per the definition of 𝑔,
𝑥1 ≼ 𝑦1 , (36)
the system as a whole fails if all the components
fail, and otherwise it continues working even if a which is represented as follows:
single component is working properly. In a system
𝑗−1 𝑗−1
𝑔(𝑦1 , 0, 𝑦𝑗+1𝑛
), even if all other components 𝑦1 and 𝑓(0, . . . , 0, 𝑥1 ) ≼ 𝑓(0,
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ . . . , 0, 𝑦1 ),
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ (37)
𝑛 𝑚−1 𝑚−1
𝑦𝑗+1 fail even then 0 is up and the system is always up.
𝑓(0, . . . , 0, 𝑥2 ) ≼ 𝑓(0,
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ . . . , 0, 𝑦2 ).
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
(ii) 𝑓(𝑥1𝑖−1 , 1, 𝑥𝑖+1
𝑚
) = 1 for all 1 ≤ 𝑖 ≤ 𝑚. (38)
𝑚−1 𝑚−1
This is so since 1 represents the component or system
which is always down, and as per the definition of By 𝑓 operation on both sides of (37) with 𝑦2 , we get
𝑓 if either of the component fails, then the whole 𝑓(𝑓(0, . . . , 0, 𝑥1 ), 𝑦2 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
0, . . . , 0)
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
system fails. Thus, even though all other components 𝑚−1 𝑚−2
are working properly but due to the component 1 the (39)
system is always down. ≼ 𝑓(𝑓(0, . . . , 0, 𝑦1 ), 𝑦2 , 0,
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ . . . , 0).
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
𝑚−1 𝑚−2
Definition 24. If (U, 𝑓, 𝑔) is an (𝑚, 𝑛)-semiring and (U, ≼) is
By 𝑓 operation on both sides of (38) with 𝑥1
a poset, then (U, 𝑓, 𝑔, ≼) is a partially ordered (𝑚, 𝑛)-semiring
if the following conditions are satisfied for all 𝑥1 , 𝑥2 , . . . , 𝑥𝑚 , 𝑓(𝑓(0, . . . , 0, 𝑥2 ), 𝑥1 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ 0, . . . , 0)
𝑦1 , 𝑦2 , . . . , 𝑦𝑛 , 𝑎, 𝑏 ∈ U and 1 ≤ 𝑖 ≤ 𝑚, 1 ≤ 𝑗 ≤ 𝑛. 𝑚−1 𝑚−2
(40)
(i) If 𝑎 ≼ 𝑏, then 𝑓(𝑥1𝑖−1 , 𝑎, 𝑥𝑖+1
𝑚
) ≼ 𝑓(𝑥1𝑖−1 , 𝑏, 𝑥𝑖+1
𝑚
). ≼ 𝑓(𝑓(0, . . . , 0, 𝑦2 ), 𝑥1 , 0,
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ . . . , 0).
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
𝑗−1 𝑗−1 𝑚−1 𝑚−2
𝑛 𝑛
(ii) If 𝑎 ≼ 𝑏, then 𝑔(𝑦1 , 𝑎, 𝑦𝑗+1 ) ≼ 𝑔(𝑦1 , 𝑏, 𝑦𝑗+1 ).
From (39) and (40), we get
Remark 25. As it is assumed that 0 is the system which is
always up, it is more fault tolerant than any of the other 𝑓(𝑓(0, . . . , 0, 𝑥1 ), 𝑦2 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ 0, . . . , 0)
𝑚−1 𝑚−2
systems or components. Therefore 0 ≼ 𝑎, for all 𝑎 ∈ U. (41)
Similarly, 𝑎 ≼ 1 because 1 is the system that always fails, and ≼ 𝑓(𝑓(0, . . . , 0, 𝑦1 ), 𝑦2 , 0,
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ . . . , 0).
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
therefore, it is the least fault tolerant; every other system is 𝑚−1 𝑚−2
more fault tolerant than it.
Similarly, we find for 𝑚 terms
Observation 26. The following are obtained for all disjoint 𝑓( . . . (𝑓(𝑓(0, . . . , 0, 𝑥1 ), 𝑥2 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
0, . . . , 0), . . . ), 𝑥𝑚 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
0, . . . , 0)
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
components 𝑟, 𝑠, 𝑥𝑖 , 𝑦𝑗 , 𝑎𝑖 , 𝑏𝑗 , which are in U, where 1 ≤ 𝑖 ≤ 𝑚, 𝑚 𝑚−1 𝑚−2 𝑚−2
1 ≤ 𝑗 ≤ 𝑛.
≼ ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
𝑓( . . . (𝑓(𝑓(0, . . . , 0, 𝑦1 ), 𝑦2 , 0,
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ . . . , 0), . . . ), 𝑦𝑚 , 0,
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ . . . , 0).
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
(i) 0 ≼ 𝑓(𝑥1𝑖−1 , 𝑟, 𝑥𝑖+1
𝑚
) ≼ 1. 𝑚 𝑚−1 𝑚−2 𝑚−2
𝑗−1 𝑛 (42)
(ii) 0 ≼ 𝑔(𝑦1 , 𝑠, 𝑦𝑗+1 ) ≼ 1.
𝑗−1 From Lemma 21, (42) may be represented as
(iii) 0 ≼ 𝑔(𝑦1 , 𝑓(𝑎1𝑚 ), 𝑦𝑗+1
𝑛
) ≼ 1.
𝑓 (𝑥1 , 𝑥2 , . . . , 𝑥𝑚 ) ≼ 𝑓 (𝑦1 , 𝑦2 , . . . , 𝑦𝑚 ) (43)
(iv) 0 ≼ 𝑓(𝑥1𝑖−1 , 𝑔(𝑏1𝑛 ), 𝑥𝑖+1
𝑚
) ≼ 1.
so
From the above description of 0 and 1, the observa-
tion is quite obvious. Case (i) shows that 0 is less faulty 𝑓 (𝑥1𝑚 ) ≼ 𝑓 (𝑦1𝑚 ) . (44)
than 𝑓(𝑥1𝑖−1 , 𝑟, 𝑥𝑖+1
𝑚
), and 𝑓(𝑥1𝑖−1 , 𝑟, 𝑥𝑖+1
𝑚
) is less faulty than 1.
Similarly, case (ii) shows that 0 is more fault tolerant than (ii) Since 𝑧𝑗 ≼ 𝑦𝑗 , for all 1 ≤ 𝑗 ≤ 𝑛
𝑗−1 𝑛 𝑗−1 𝑛
𝑔(𝑦1 , 𝑠, 𝑦𝑗+1 ) and 𝑔(𝑦1 , 𝑠, 𝑦𝑗+1 ) is more fault tolerant than 𝑔(1, . . . , 1, 𝑧1 ) ≼ 𝑔(1,
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ . . . , 1, 𝑢1 )
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
𝑛 𝑗−1
1. Likewise, case (iii) shows the operation 𝑔 over 𝑦1 , 𝑦𝑗+1 𝑛−1 𝑛−1
(45)
and 𝑓 of 𝑎1𝑚 to be less faulty than 1 and more faulty than 0, 𝑔(1, . . . , 1, 𝑧2 ) ≼ 𝑔(1,
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ . . . , 1, 𝑢2 ).
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
and a similar interpretation is made for (iv). 𝑛−1 𝑛−1
Journal of Applied Mathematics 7

After following similar steps as seen in part (i), we use the 𝑔 Lemma 29. If ≼ is a fault-tolerance partial order and 𝑥𝑖 , 𝑦𝑗
operation for 𝑛 terms, are disjoint components which are in U, where 1 ≤ 𝑖 ≤ 𝑚 and
1 ≤ 𝑗 ≤ 𝑛, one gets the following:
𝑔( . . . (𝑔(𝑔(1,
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ . . . , 1, 𝑧1 ), 𝑧2 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ 1, . . . , 1), . . . ), 𝑧𝑛 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
1, . . . , 1)
𝑛 𝑛−1 𝑛−2 𝑛−2 (i) 𝑥𝑖 ≼ 𝑓(𝑥1 , 𝑥2 , . . . , 𝑥𝑚 ),

≼ 𝑔( . . . (𝑔(𝑔(1, . . . , 1, 𝑢1 ), 𝑢2 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ 1, . . . , 1), . . . ), 𝑢𝑛 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
1, . . . , 1), (ii) 𝑔(𝑦1 , 𝑦2 , . . . , 𝑦𝑛 ) ≼ 𝑦𝑗 .
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
𝑛 𝑛−1 𝑛−2 𝑛−2
(46) Proof. (i) As

which is represented as 0 ≼ 𝑥1 , (55)

by 𝑓 operation on both sides of (55) with 𝑥𝑖 , we get


𝑔 (𝑧1 , 𝑧2 , . . . , 𝑧𝑛 ) ≼ 𝑔 (𝑢1 , 𝑢2 , . . . , 𝑢𝑛 ) , (47)
𝑓(0, 𝑥𝑖 , 0, . . . , 0) ≼ 𝑓(𝑥1 , 𝑥𝑖 , 0,
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ . . . , 0).
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ (56)
and so 𝑚−2 𝑚−2

𝑔 (𝑧1𝑛 ) ≼ 𝑔 (𝑢1𝑛 ) . (48) Therefore,


𝑥𝑖 ≼ 𝑓(𝑥1 , 𝑥𝑖 , 0, . . . , 0).
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ (57)
Theorem 28. If ≼ is a fault-tolerance partial order and given 𝑚−2

disjoint components 𝑎𝑖 , 𝑐𝑗 , 𝑏𝑖 , 𝑑𝑗 in U, where 1 ≤ 𝑖 ≤ 𝑚, 1 ≤ Similarly, we obtain


𝑗 ≤ 𝑛 and 1 ≤ 𝑘 ≤ 𝑚, the following obtain.
𝑥𝑖 ≼ 𝑓(𝑥1 , 𝑥𝑖 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
0, . . . , 0) ≼ ⋅ ⋅ ⋅ ≼ 𝑓 (𝑥1 , 𝑥2 , 𝑥𝑖 , . . . , 𝑥𝑚−1 , 0)
(i) If 𝑎𝑖 ≼ 𝑏𝑖 , then 𝑚−2

𝑗−1 𝑗−1 ≼ 𝑓 (𝑥1 , 𝑥2 , . . . , 𝑥𝑚 ) .


𝑔 (𝑦1 , 𝑓 (𝑎1𝑚 ) , 𝑦𝑗+1
𝑛
) ≼ 𝑔 (𝑦1 , 𝑓 (𝑏1𝑚 ) , 𝑦𝑗+1
𝑛
)
(49) (58)
∀𝑦1 , 𝑦2 , . . . , 𝑦𝑛 ∈ U.
Hence,
(ii) If 𝑐𝑗 ≼ 𝑑𝑗 , then 𝑥𝑖 ≼ 𝑓 (𝑥1 , 𝑥2 , . . . , 𝑥𝑚 ) , (59)

𝑓 (𝑥1𝑘−1 , 𝑔 (𝑐1𝑛 ) , 𝑥𝑘+1


𝑚
) ≼ 𝑓 (𝑥1𝑘−1 , 𝑔 (𝑑1𝑛 ) , 𝑥𝑘+1
𝑚
) for all 1 ≤ 𝑖 ≤ 𝑚.
(50) (ii) As
∀𝑥1 , 𝑥2 , . . . , 𝑥𝑚 ∈ U.
𝑦1 ≼ 1, (60)
Proof. (i) Since 𝑎𝑖 ≼ 𝑏𝑖 , for all 1 ≤ 𝑖 ≤ 𝑚. by 𝑔 operation on both sides of (60) with 𝑦𝑗 , we get
Therefore, from Lemma 27 (i)
𝑔(𝑦1 , 𝑦𝑗 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
1, . . . , 1) ≼ 𝑦𝑗 . (61)
𝑓 (𝑎1𝑚 ) ≼ 𝑓 (𝑏1𝑚 ) , ∀𝑎1 , 𝑎2 , . . . , 𝑎𝑚 , 𝑏1 , 𝑏2 , . . . , 𝑏𝑚 ∈ U. 𝑛−2
(51)
Similarly, we obtain
From Definition 24 of a partially ordered (𝑚, 𝑛)-semiring, we
deduce that 𝑔 (𝑦1 , 𝑦2 , . . . , 𝑦𝑛 ) ≼ 𝑔 (𝑦1 , 𝑦2 , 𝑦𝑗 , . . . , 𝑦𝑛−1 , 1)
(62)
𝑗−1
𝑔 (𝑦1 , 𝑓 (𝑎1𝑚 ) , 𝑦𝑗+1
𝑛 𝑗−1
) ≼ 𝑔 (𝑦1 , 𝑓 (𝑏1𝑚 ) , 𝑦𝑗+1
𝑛
), (52) ≼ ⋅ ⋅ ⋅ ≼ 𝑔(𝑦1 , 𝑦𝑗 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
1, . . . , 1) ≼ 𝑦𝑗 .
𝑛−2

for all 1 ≤ 𝑗 ≤ 𝑛. Hence,


(ii) Since 𝑐𝑗 ≼ 𝑑𝑗 , for all 1 ≤ 𝑗 ≤ 𝑛, from Lemma 27 (ii),
we find that 𝑔 (𝑦1 , 𝑦2 , . . . , 𝑦𝑛 ) ≼ 𝑦𝑗 , (63)

for all 1 ≤ 𝑗 ≤ 𝑛.
𝑔 (𝑐1𝑛 ) ≼ 𝑔 (𝑑1𝑛 ) , ∀𝑐1 , 𝑐2 , . . . , 𝑐𝑛 , 𝑑1 , 𝑑2 , . . . , 𝑑𝑛 ∈ U.
(53) Corollary 30. If ≼ is a fault-tolerance partial order, then the
following hold for all disjoint components 𝑥𝑖 , 𝑦𝑗 which are
From Definition 24 of a partially ordered (𝑚, 𝑛)-semiring, we elements of U, where 1 ≤ 𝑖 ≤ 𝑚, 1 ≤ 𝑗 ≤ 𝑛 and 𝑘, 𝑡 ∈ Z+ :
deduce that
0, . . . , 0) ≼ 𝑓(𝑥1𝑚 ), where 𝑘 < 𝑚,
(i) 𝑓(𝑥1 , 𝑥2 , . . . , 𝑥𝑘 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
𝑓 (𝑥1𝑘−1 , 𝑔 (𝑐1𝑛 ) , 𝑥𝑘+1
𝑚
) ≼ 𝑓 (𝑥1𝑘−1 , 𝑔 (𝑑1𝑛 ) , 𝑥𝑘+1
𝑚
), (54) 𝑚−𝑘

(ii) 𝑔(𝑦1𝑛 ) ≼ 𝑔(𝑦1 , 𝑦2 , . . . , 𝑦𝑡 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟


1, . . . , 1), where 𝑡 < 𝑛.
for all 1 ≤ 𝑘 ≤ 𝑚. 𝑛−𝑡
8 Journal of Applied Mathematics

Proof. (i) From (58), we deduce that Theorem 34. Let ≼ be a fault-tolerance partial order and 𝑥𝑖 ≼
𝑦𝑖 and 𝑧𝑗 ≼ 𝑢𝑗 for all 𝑥𝑖 , 𝑦𝑖 , 𝑧𝑗 , 𝑢𝑗 ∈ U, where 1 ≤ 𝑖 ≤ 𝑚 and
𝑓(𝑥1 , . . . , 𝑥𝑘 , 0, . . . , 0) ≼ 𝑓(𝑥1 , . . . , 𝑥𝑘+1 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ 0, . . . , 0) 1 ≤ 𝑗 ≤ 𝑛. Then, the following obtain:
𝑚−𝑘 𝑚−𝑘−1
(64)
(𝑚) (𝑚)
≼ ⋅ ⋅ ⋅ ≼ 𝑓 (𝑥1 , 𝑥2 , . . . , 𝑥𝑚 ) . (i) 𝑓(𝑓(𝑥1𝑚 )) ≼ 𝑓(𝑓(𝑦1𝑚 )),
(𝑛) (𝑛)
Therefore,
(ii) 𝑔(𝑔(𝑧1𝑛 )) ≼ 𝑔(𝑔(𝑢1𝑛 )),
0, . . . , 0) ≼ 𝑓 (𝑥1𝑚 ) .
𝑓(𝑥1 , . . . , 𝑥𝑘 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ (65) (𝑚) (𝑚)
𝑚−𝑘 (iii) 𝑓(𝑔(𝑧1𝑛 )) ≼ 𝑓(𝑔(𝑢1𝑛 )),
(ii) As in part (i), we deduce from (62) that (𝑛) (𝑛)
(iv) 𝑔(𝑓(𝑥1𝑚 )) ≼ 𝑔(𝑓(𝑦1𝑚 )).
𝑔 (𝑦1𝑛 ) ≼ 𝑔(𝑦1 , 𝑦2 , . . . , 𝑦𝑡 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
1, . . . , 1). (66)
𝑛−𝑡 Proof. (i) As

(𝑚) 𝑥𝑖 ≼ 𝑦𝑖 , 1 ≤ 𝑖 ≤ 𝑚, (69)
𝑓(𝑓(𝑎1𝑚 )) represents the system which is obtained after
applying the 𝑓 operation on 𝑚 repeated 𝑓(𝑎1𝑚 ) systems or from Lemma 27 (i), we get
(𝑛)
subsystems. Similarly, 𝑔(𝑔(𝑏1𝑛 ))
represents the system which 𝑓 (𝑥1𝑚 ) ≼ 𝑓 (𝑦1𝑚 ) . (70)
is obtained after applying the 𝑔 operation on 𝑛 repeated 𝑔(𝑏1𝑛 )
systems or subsystems. This is written as

Theorem 31. If ≼ is a fault-tolerance partial order, and com- 𝑓(0, . . . , 0, 𝑓(𝑥1𝑚 )) ≼ 𝑓(0,
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ . . . , 0, 𝑓(𝑦1𝑚 )).
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ (71)
𝑚−1 𝑚−1
ponents 𝑥1 , 𝑥2 , . . . , 𝑥𝑚 , 𝑦1 , 𝑦2 , . . . , 𝑦𝑛 are disjoint components
and are in U, then So by 𝑓 operation on both sides of (71) with 𝑓(𝑥1𝑚 ), we get
(𝑚)
(i) 𝑓(𝑥1𝑚 ) ≼ 𝑓(𝑓(𝑥1𝑚 )), 𝑓(𝑓(0, . . . , 0, 𝑓(𝑥1𝑚 )), 𝑓(𝑥1𝑚 ), ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ 0, . . . , 0)
𝑚−1 𝑚−2
(𝑛)
(72)
(ii) 𝑔(𝑔(𝑦1𝑛 )) ≼ 𝑔(𝑦1𝑛 ). ≼ 𝑓(𝑓(0, . . . , 0, 𝑓(𝑥1𝑚 )), 𝑓(𝑦1𝑚 ), ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ 0, . . . , 0).
𝑚−1 𝑚−2
Corollary 32. The following hold for all disjoint components
𝑥1 , . . . , 𝑥𝑚 , 𝑧1 , . . . , 𝑧𝑛 , 𝑦1 , . . . , 𝑦𝑚 , 𝑢1 , . . . , 𝑢𝑛 , which are ele- So by 𝑓 operation on both sides of (71) with 𝑓(𝑦1𝑚 ), we get
ments of U, where 𝑚, 𝑛 ∈ Z+ .
(𝑚)
𝑓(𝑓(0, . . . , 0, 𝑓(𝑥1𝑚 )), 𝑓(𝑦1𝑚 ), ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ 0, . . . , 0)
(i) If 𝑓(𝑓(𝑥1𝑚 )) ≼ 𝑓(𝑦1𝑚 ), then 𝑚−1 𝑚−2
(73)
𝑓 (𝑥1𝑚 ) ≼ 𝑓 (𝑦1𝑚 ) . ≼ 𝑓(𝑓(0, . . . , 0, 𝑓(𝑦1𝑚 )), ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ 0, . . . , 0, 𝑓(𝑦1𝑚 )).
(67)
𝑚−1 𝑚−2

(𝑛) From (72) and (73), we get


(ii) If 𝑔(𝑧1𝑛 ) ≼ 𝑔(𝑔(𝑢1𝑛 )), then
(2) (2)
𝑔 (𝑧1𝑛 ) ≼ 𝑔 (𝑢1𝑛 ) . (68) 𝑓(0, . . . , 0, 𝑓(𝑥1𝑚 ))≼ 𝑓(0,
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ . . . , 0, 𝑓(𝑦1𝑚 )) .
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ (74)
𝑚−2 𝑚−2
(𝑚)
Proof. (i) 𝑓(𝑓(𝑥1𝑚 )) ≼ 𝑓(𝑦1𝑚 ) and from Theorem 31, 𝑓(𝑥1𝑚 ) ≼ Similarly, we get for 𝑚 terms
(𝑚)
𝑓(𝑓(𝑥1𝑚 )). Therefore, 𝑓(𝑥1𝑚 ) ≼ 𝑓(𝑦1𝑚 ). (𝑚) (𝑚)
(ii) The proof is very similar to that of part (i). 𝑓 (𝑓(𝑥1𝑚 ))≼ 𝑓 (𝑓(𝑦1𝑚 )) . (75)

Corollary 33. Let 𝑘 and 𝑡 be positive integers and 𝑘 < 𝑚, (ii) We know that
𝑡 < 𝑛. Given disjoint components 𝑥1 , . . . , 𝑥𝑚 , 𝑦1 , 𝑦2 , . . . , 𝑦𝑚 ,
𝑧𝑗 ≼ 𝑢𝑗 , 1 ≤ 𝑗 ≤ 𝑛. (76)
𝑧1 , 𝑧2 , . . . , 𝑧𝑛 , 𝑢1 , 𝑢2 , . . . , 𝑢𝑛 that are in U, the following hold:
(𝑘) From Lemma 27 (ii), we get
(i) If 𝑓(0, . . . , 0, 𝑓(𝑥1𝑚 )) ≼ 𝑓(𝑦1𝑚 ), then 𝑓(𝑥1𝑚 ) ≼ 𝑓(𝑦1𝑚 ).
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
𝑚−𝑘 𝑔 (𝑧1𝑛 ) ≼ 𝑔 (𝑢1𝑛 ) . (77)
(𝑡)
(ii) If 𝑔(𝑧1𝑛 ) ≼ 𝑔(1, . . . , 1, 𝑔(𝑢1𝑛 )),
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ then 𝑔(𝑧1𝑛 ) ≼ 𝑔(𝑢1𝑛 ). Which is represented as follows
𝑛−𝑡
𝑔(1, . . . , 1, 𝑔(𝑧1𝑛 )) ≼ 𝑔(1,
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ . . . , 1, 𝑔(𝑢1𝑛 )).
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ (78)
Proof. Similar to Corollary 32. 𝑛−1 𝑛−1
Journal of Applied Mathematics 9

Now by 𝑔 operation on both sides of (78) with 𝑔(𝑧1𝑛 ), we get Proof. (i) Proof is similar to that of Theorem 34 (i). We find
the 𝑓 operation of 𝑘 terms where ∀𝑘 ∈ Z+ , and 𝑘 < 𝑚.
(2)
𝑔 ( 𝑔(𝑧1𝑛 ), 1, . . . , 1) ≼ 𝑔(𝑔(𝑧1𝑛 ), 𝑔(𝑢1𝑛 ), ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ 1, . . . , 1). (79) (ii) Proof is similar to that of Theorem 34 (ii). We find the
𝑛−2 𝑛−2 𝑔 operation of 𝑡 terms where ∀𝑡 ∈ Z+ , and 𝑡 < 𝑛.
So by 𝑔 operation on both sides of (78) with 𝑔(𝑢1𝑛 ), we get We propose the following theorem for very complex
systems.
(2)
𝑔(1, . . . , 1, 𝑔(𝑧1𝑛 ), 𝑔(𝑢1𝑛 )) ≼ 𝑔(1,
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ . . . , 1, 𝑔(𝑢1𝑛 )) .
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ (80)
𝑛−2 𝑛−2
Theorem 36. If ≼ is a fault-tolerance partial order, disjoint
components 𝑎𝑖 , 𝑏𝑖 , 𝑐𝑗 , 𝑑𝑗 , 𝑥𝑘 , 𝑦𝑘 , 𝑧𝑡 , 𝑢𝑡 are in U and 𝑎𝑖 ≼ 𝑏𝑖 ,
So now from (79) and (80), we get 𝑐𝑗 ≼ 𝑑𝑗 , 𝑥𝑘 ≼ 𝑦𝑘 and 𝑧𝑡 ≼ 𝑢𝑡 , where 1 ≤ 𝑖 ≤ 𝑚, 1 ≤ 𝑗 ≤ 𝑛,
(2) (2)
1 ≤ 𝑘 ≤ 𝑚 and 1 ≤ 𝑡 ≤ 𝑛, then
𝑔(1, . . . , 1, 𝑔(𝑧1𝑛 ))≼ 𝑔(1,
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ . . . , 1, 𝑓(𝑢1𝑛 )) .
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ (81)
𝑛−2 𝑛−2
(i) 𝑓(𝑥1𝑘−1 , 𝑓(𝑎1𝑚 ), 𝑥𝑘+1
𝑚
) ≼ 𝑓(𝑦1𝑘−1 , 𝑓(𝑏1𝑚 ), 𝑦𝑘+1
𝑚
), for all
1 ≤ 𝑘 ≤ 𝑚;
Similarly, we find for 𝑛 terms
(ii) 𝑓(𝑥1𝑘−1 , 𝑔(𝑐1𝑛 ), 𝑥𝑘+1
𝑚
) ≼ 𝑓(𝑦1𝑘−1 , 𝑔(𝑑1𝑛 ), 𝑦𝑘+1
𝑚
), for all 1 ≤
(𝑛) (𝑛) 𝑘 ≤ 𝑚;
𝑔 (𝑔 (𝑧1𝑛 ))≼ 𝑔 (𝑔 (𝑢1𝑛 )) . (82)
(iii) 𝑔(𝑧1𝑡−1 , 𝑓(𝑎1𝑚 ), 𝑧𝑡+1
𝑛
) ≼ 𝑔(𝑢1𝑡−1 , 𝑓(𝑏1𝑚 ), 𝑢𝑡+1
𝑛
), for all 1 ≤
(iii) From Lemma 27 (ii) 𝑡 ≤ 𝑛; and
𝑔 (𝑧1𝑛 ) ≼ 𝑔 (𝑢1𝑛 ) . (83) (iv) 𝑔(𝑧1𝑡−1 , 𝑔(𝑐1𝑛 ), 𝑧𝑡+1
𝑛
) ≼ 𝑔(𝑢1𝑡−1 , 𝑔(𝑑1𝑛 ), 𝑢𝑡+1
𝑛
), for all 1 ≤
𝑡 ≤ 𝑛.
Similar to part (i), we find 𝑓 operation of 𝑚 terms and get
𝑓(𝑔(𝑧 𝑛 𝑛 𝑛 Proof. (i) From Lemma 27 (i), if 𝑎𝑖 ≼ 𝑏𝑖 , then 𝑓(𝑎1𝑚 ) ≼ 𝑓(𝑏1𝑚 )
1 ), 𝑔(𝑧1 ), . . . , 𝑔(𝑧1 ))
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
𝑚 for all 1 ≤ 𝑖 ≤ 𝑚.
𝑛 𝑛 𝑛 We prove in a similar manner as Lemma 27 (i) that
≼ 𝑓(𝑔(𝑢 1 ), 𝑔(𝑢1 ), . . . , 𝑔(𝑢1 )),
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ (84)
𝑚 𝑓(𝑓(𝑎1𝑚 ), 𝑥1 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
0, . . . , 0) ≼ 𝑓(𝑓(𝑏1𝑚 ), 𝑦1 , ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
0, . . . , 0). (89)
(𝑚) (𝑚) 𝑚−2 𝑚−2
𝑓 (𝑔 (𝑧1𝑛 ))≼ 𝑓 (𝑔 (𝑢1𝑛 )) .
Similarly, we get
(iv) We know that
𝑓 (𝑓 (𝑎1𝑚 ) , 𝑥1𝑘−1 , 𝑥𝑘+1
𝑚
) ≼ 𝑓 (𝑓 (𝑏1𝑚 ) , 𝑦1𝑘−1 , 𝑦𝑘+1
𝑚
). (90)
𝑥𝑖 ≼ 𝑦𝑖 , 1 ≤ 𝑖 ≤ 𝑚, (85)
so from Lemma 27 (i), we get Thus,

𝑓 (𝑥1𝑚 ) ≼ 𝑓 (𝑦1𝑚 ) . (86) 𝑓 (𝑥1𝑘−1 , 𝑓 (𝑎1𝑚 ) , 𝑥𝑘+1


𝑚
) ≼ 𝑓 (𝑦1𝑘−1 , 𝑓 (𝑏1𝑚 ) , 𝑦𝑘+1
𝑚
). (91)

As proved in part (ii), we find 𝑔 operations of 𝑛 terms and get Similar to the above, we can prove (ii), (iii), and (iv).
𝑚 𝑚 𝑚
𝑔(𝑓(𝑥 1 ), 𝑓(𝑥1 ), . . . , 𝑓(𝑥1 ))
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
𝑛 Acknowledgment
(87)
𝑚 𝑚 𝑚
≼ 𝑔(𝑓(𝑦 1 ), 𝑓(𝑦1 ), . . . , 𝑓(𝑦1 )).
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ The work was done while the first author was a masters stu-
𝑛 dent at the International Institute of Information Technology
Thus, we get (IIIT)-Bangalore.
(𝑛) (𝑛)
𝑔 (𝑓 (𝑥1𝑚 ))≼ 𝑔 (𝑓 (𝑦1𝑚 )) . (88) References
[1] F. Cristian, B. Dancey, and J. Dehn, “Fault-tolerance in air traffic
Corollary 35. If ≼ is a fault-tolerance partial order and 𝑘 < control systems,” ACM Transactions on Computer Systems, vol.
𝑚, 𝑡 < 𝑛 where 𝑘, 𝑡 ∈ Z+ , if 𝑥𝑖 ≼ 𝑦𝑖 , 𝑧𝑗 ≼ 𝑢𝑗 for all disjoint 14, no. 3, pp. 265–286, 1996.
components 𝑥𝑖 , 𝑧𝑗 , 𝑦𝑖 , 𝑢𝑗 , which are in U, where 1 ≤ 𝑖 ≤ 𝑚 and [2] D. Brière and P. Traverse, “AIRBUS A320/A330/A340 electrical
1 ≤ 𝑗 ≤ 𝑛, then flight controls: a family of fault-tolerant systems,” in Proceedings
of the 23rd Annual International Symposium on Fault-Tolerant
(𝑘) (𝑘) Computing (FTCS ’93), pp. 616–623, Toulouse, France, 1993.
(i) 𝑓(0, . . . , 0, 𝑓(𝑥1𝑚 )) ≼ 𝑓(0,
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ . . . , 0, 𝑓(𝑦1𝑚 )),
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
𝑚−𝑘 𝑚−𝑘 [3] T. Ayav, P. Fradet, and A. Girault, “Implementing fault-tolerance
(𝑡) (𝑡)
in real-time programs by automatic program transformations,”
(ii) 𝑔(1, ...,1
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ 𝑔(𝑧1𝑛 )) ≼ 𝑔(1, . . . , 1, 𝑔(𝑢1𝑛 )).
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟ Transactions on Embedded Computing Systems, vol. 7, no. 4,
𝑛−𝑡 𝑛−𝑡 article 45, 2008.
10 Journal of Applied Mathematics

[4] C. Ferrell, “Failure recognition and fault tolerance of an auton- [23] B. Davvaz, W. A. Dudek, and S. Mirvakili, “Neutral elements,
omous robot,” Adaptive Behavior, vol. 2, pp. 375–398, 1994. fundamental relations and 𝑛-ary hypersemigroups,” Interna-
[5] B. Lussier, M. Gallien, J. Guiochet, F. Ingrand, M. O. Killijian, tional Journal of Algebra and Computation, vol. 19, no. 4, pp.
and D. Powell, “Fault tolerant planning for critical robots,” in 567–583, 2009.
Proceedings of the 37th Annual IEEE/IFIP International Confer- [24] W. A. Dudek, “Idempotents in 𝑛-ary semigroups,” Southeast
ence on Dependable Systems and Networks (DSN ’07), pp. 144– Asian Bulletin of Mathematics, vol. 25, no. 1, pp. 97–104, 2001.
153, Edinburgh, UK, June 2007.
[25] L. M. Gluskin, “Fault tolerant planning for critical robots,”
[6] A. Avižienis, “The dependability problem: Introduction and Matematicheskii Sbornik, vol. 68, pp. 444–472, 1965.
verification of fault tolerance for a very complex system,” in
Proceedings of the Fall Joint Computer Conference on Exploring [26] W. A. Dudek and V. V. Mukhin, “On topological 𝑛-ary semi-
Technology: Today and Tomorrow (ACM ’87), pp. 89–93, IEEE groups,” Quasigroups and Related Systems, vol. 3, pp. 73–88,
Computer Society Press, Los Alamitos, Calif, USA, 1987. 1996.
[7] F. Cristian, B. Dancey, and J. Dehn, “Fault-tolerance in the [27] W. A. Dudek, “On distributive 𝑛-ary groups,” Quasigroups and
advanced automation system,” in Proceedings of the 4th Work- Related Systems, vol. 2, no. 1, pp. 132–151, 1995.
shop on ACM SIGOPS European Workshop, pp. 6–17, ACM, [28] U. Hebisch and H. J. Weinert, Semirings: Algebraic Theory and
Bologna, Italy, 1990. Applications in Computer Science, vol. 5 of Series in Algebra,
[8] C. De Capua, A. Battaglia, A. Meduri, and R. Morello, “A World Scientific Publishing, River Edge, NJ, USA, 1998.
patient-adaptive ECG measurement system for fault-tolerant [29] S. Bourne, “On multiplicative idempotents of a potent semir-
diagnoses of heart abnormalities,” in Proceedings of the IEEE ing,” Proceedings of the National Academy of Sciences of the
Instrumentation and Measurement Technology (IMTC ’07), pol, United States of America, vol. 42, no. 9, pp. 632–638, 1956.
May 2007.
[30] S. Burris and H. P. Sankappanavar, A Course in Universal
[9] T. Perraju, S. Rana, and S. Sarkar, “Specifying fault tolerance in Algebra, vol. 78 of Graduate Texts in Mathematics, Springer, New
mission critical systems,” in Proceedings of the High-Assurance York, NY, USA, 1981.
Systems Engineering Workshop, pp. 24–31, IEEE, October 1996.
[31] P. J. Allen, “A fundamental theorem of homomorphisms for
[10] P. E. Beckmann, Fault-tolerant computation using algebraic semirings,” Proceedings of the American Mathematical Society,
homomorphisms [Ph.D. thesis], Department of Electrical Engi- vol. 21, no. 2, pp. 412–416, 1969.
neering and Computer Science, Massachusetts Institute of
Technology, Cambridge, Mass, USA, 1992. [32] S. Rao, “A systems algebra and its applications,” in Proceed-
ings of the IEEE International Systems Conference Proceedings
[11] C. N. Hadjicostis, Fault-tolerant computation in semigroups and
(SysCon ’08), pp. 28–35, Montreal, Canada, April 2008.
semirings [MIT Master of Engineering thesis], 1995.
[12] C. J. Monico, Semirings and semigroup actions in public-key
cryptography [Ph.D. thesis], Department of Mathematics, Grad-
uate School of the University of Notre Dame, Notre Dame, Ind,
USA, 2002.
[13] T. J. Green, G. Karvounarakis, and V. Tannen, “Prove-
nance semirings,” in Proceedings of the 26th ACM SIGMOD-
SIGACT-SIGART Symposium on Principles of Database Systems
(PODS ’07), pp. 31–40, Beijing, China, June 2007.
[14] J. S. Golan, Semirings and Their Applications, Kluwer Academic
Publishers, Dordrecht, The Netherlands, 1999.
[15] S. Rao, “An algebra of fault tolerance,” Journal of Algebra and
Discrete Structures, vol. 6, no. 3, pp. 161–180, 2008.
[16] J. Timm, Kommutative n-Gruppen [Ph.D. thesis], University of
Hamburg, Hamburg, Germany, 1967.
[17] G. Crombez, “On (𝑛, 𝑚)-rings,” Abhandlungen aus dem Math-
ematischen Seminar der Universität Hamburg, vol. 37, pp. 180–
199, 1972.
[18] G. Crombez and J. Timm, “On (𝑛, 𝑚)-quotient rings,” Abhand-
lungen aus dem Mathematischen Seminar der Universität Ham-
burg, vol. 37, pp. 200–203, 1972.
[19] J. J. Leeson and A. T. Butson, “Equationally complete (𝑚, 𝑛)
rings,” Algebra Universalis, vol. 11, no. 1, pp. 28–41, 1980.
[20] J. J. Leeson and A. T. Butson, “On the general theory of (𝑚, 𝑛)
rings,” Algebra Universalis, vol. 11, no. 1, pp. 42–76, 1980.
[21] W. A. Dudek, “On the divisibility theory in (𝑚, 𝑛)-rings,” Dem-
onstratio Mathematica, vol. 14, no. 1, pp. 19–32, 1981.
[22] B. Davvaz, W. A. Dudek, and T. Vougiouklis, “A generalization
of 𝑛-ary algebraic systems,” Communications in Algebra, vol. 37,
no. 4, pp. 1248–1263, 2009.

Você também pode gostar