• 沒有找到結果。

Fault-tolerant hamiltonian laceability of hypercubes

N/A
N/A
Protected

Academic year: 2021

Share "Fault-tolerant hamiltonian laceability of hypercubes"

Copied!
6
0
0

加載中.... (立即查看全文)

全文

(1)

Fault-tolerant hamiltonian laceability of hypercubes

Chang-Hsiung Tsai, Jimmy J.M. Tan, Tyne Liang, Lih-Hsing Hsu

Department of Computer and Information Science, National Chiao Tung University, Hsinchu, 30050 Taiwan, ROC

Received 16 January 2001; received in revised form 27 December 2001 Communicated by F.Y.L. Chin

Abstract

It is known that every hypercube Qnis a bipartite graph. Assume that n 2 and F is a subset of edges with |F |  n − 2. We prove that there exists a hamiltonian path in Qn− F between any two vertices of different partite sets. Moreover, there exists a path of length 2n− 2 between any two vertices of the same partite set. Assume that n  3 and F is a subset of edges with |F |  n − 3. We prove that there exists a hamiltonian path in Qn− {v} − F between any two vertices in the partite set without v. Furthermore, all bounds are tight.2002 Elsevier Science B.V. All rights reserved.

Keywords: Hamiltonian laceable; Hypercube; Fault tolerance

1. Introduction

In this paper, a network is represented as an undi-rected graph. For the graph definition and notation we follow [1]. G= (V, E) is a graph if V is a fi-nite set and E is a subset of{(a, b) | (a, b) is an or-dered pair of V}. We say that V is the vertex set and

E is the edge set. Two vertices a and b are

adja-cent if (a, b)∈ E. A path is a sequence of adjacent vertices, written asv0, v1, v2, . . . , vm , in which all the vertices v0, v1, . . . , vmare distinct except possibly v0= vm. We also write the path v0, P , vm , where P = v1, . . . , vm−1 . A path is a hamiltonian path if its vertices are distinct and they span on V . A cycle is a path with at least three vertices such that the first

This work was supported in part by the National Science Council of the Republic of China under Contract NSC 89-2115-M-009-020.

* Corresponding author.

E-mail address: [email protected] (L.-H. Hsu).

vertex is the same as the last one. A cycle is a hamil-tonian cycle if it traverses every vertex of G exactly once. A graph is hamiltonian if it has a hamiltonian cycle. A graph G is hamiltonian connected if there exists a hamiltonian path joining any two vertices of

G. A graph G= (V0∪ V1, E) is bipartite if V (G) is

the union of two disjoint sets V0and V1such that each edge consists of one vertex from each set.

As the hamiltonicity of a graph G is concerned, it is an important issue to investigate if G is hamiltonian or hamiltonian connected. However, any hamiltonian bipartite graph G = (V0 ∪ V1, E) satisfies |V0| =

|V1|. Since the colors of the bipartite path alternates, all hamiltonian bipartite graphs are not hamiltonian-connected. Simmons [8] introduces the concept of hamiltonian laceability for those hamiltonian bipartite graphs. A hamiltonian bipartite graph G = (V0

V1, E) is hamiltonian laceable if there is a hamiltonian path between any two vertices x and y with x∈ V0and

y∈ V1. Hsieh et al. [3] further extend this concept into 0020-0190/02/$ – see front matter 2002 Elsevier Science B.V. All rights reserved.

(2)

strongly hamiltonian laceable. A hamiltonian laceable graph G= (V0 ∪ V1, E) is strongly if there is a

simple path of length|V0∪ V1| − 2 between any two vertices of the same partite set. Lewinter et al. [7] also introduce the concept of hyper-hamiltonian laceable. A hamiltonian laceable graph G= (V0∪ V1, E) is

hyper-hamiltonian laceable if for any vertex v∈ Vi, i= 0, 1, there is a hamiltonian path of G − v between

any two vertices of V1−i.

The edge fault-tolerant hamiltonicity proposed by Hsieh, Chen, and Ho [4], measures the performance of the hamiltonian property in the faulty networks. A hamiltonian graph G is k edge fault-tolerant hamil-tonian if G− F remains hamiltonian for every F ⊂

E(G) with |F |  k. The edge fault-tolerant

hamil-tonicity,He(G), is defined to be the maximum integer k such that G is k edge fault-tolerant hamiltonian if G

is hamiltonian, and undefined if otherwise. It is easy to see thatHe(G) δ(G) − 2 for any hamiltonian graph G where

δ(G)= mindeg(v)| v ∈ V (G).

We can further study other fault-hamiltonicity. A ham-iltonian laceable graph G is k edge fault-tolerant hamiltonian laceable if G− F remains hamiltonian laceable for every F⊂ E(G) with |F |  k. The edge fault-tolerant hamiltonian laceability,HLe(G), is

de-fined to be the maximum integer k such that G is

k edge fault-tolerant hamiltonian laceable, and

unde-fined if otherwise. A strongly hamiltonian laceable graph G is k edge fault-tolerant strongly hamiltonian laceable if G− F remains strongly hamiltonian lace-able for every F⊂ E(G) with |F |  k. The edge fault-tolerant strongly hamiltonian laceability,HeSL(G), is

defined to be the maximum integer k such that G is k edge fault-tolerant strongly hamiltonian laceable, and undefined if otherwise. A hyper-hamiltonian laceable graph G is k edge fault-tolerant hyper-hamiltonian laceable if G− F remains hyper-hamiltonian laceable for every F ⊂ E(G) with |F |  k. The edge fault-tolerant hyper-hamiltonian laceability,Heh(G), is

de-fined to be the maximum integer k such that G is

k edge fault-tolerant hyper-hamiltonian laceable, and

undefined if otherwise.

Network topology is usually represented by a graph where nodes represent processors and edges represent links between processors. The binary hypercube, Qn, is one of the most popular topologies [6]. Let u=

un−1un−2. . . u1u0and v= vn−1vn−2. . . v1v0 be two

n-bit binary strings. The Hamming distance h(u, v)

between two vertices u and v is the number of different bits in the corresponding strings of both vertices. The

n-dimensional hypercube consists of all n-bit binary

strings as its vertices and two vertices u and v are adjacent if and only if h(u, v)= 1. Latifi et al. [5] proved thatHe(Qn)= n − 2 if n  2. Harary et al. [2] proved that Qnis strongly hamiltonian laceable if and only if n 2. Lewinter et al. [7] proved that Qn is hyper-hamiltonian laceable if and only if n 3. In this paper, we prove that

HL

e(Qn)= HSLe (Qn)= n − 2 if n  2 and

Hh

e(Qn)= n − 3 if n  3.

Using our approach, we can easily prove thatHe(Qn)

= n − 2 if n  2.

2. Fault-tolerant hamiltonian laceability of hypercubes

Let Qn be the n-dimensional hypercube. In this section, we will prove that HSLe (Qn) n − 2 and Hh

e(Qn) n − 3. It is clear that HSLe (G) n − 2 and Hh

e(G) n − 3, for any n-regular bipartite graph G. So HeSL(Qn)= n − 2, Hhe(Qn)= n − 3, and these results are optimal. Qncan be divided into two copies of Qn−1, denoted by Q0n−1 and Q1n−1. Let Ec be the set of crossing edges, i.e.,

Ec=



(u, u )| (u, u )∈ E(Qn), u∈ V (Q0n−1) and u ∈ V (Q1n−1).

Let F be the set of faulty edges of Qn, F0= F ∩

E(Q0n−1), F1= F ∩E(Q1n−1), and Fc= F ∩Ec. Also let f0= |F0|, f1= |F1|, and fc= |Fc|. Since Qn is edge symmetric, it suffices to consider only the case that fc 1. That means, Qn can be split into Q0n−1 and Q1n−1 using any dimension d , where 1 d  n. Therefore, given a faulty edge set F , Qncan be split into Q0n−1 and Q1n−1such that Fcis not an empty set.

Lemma 1. Q3 is 1 edge fault-tolerant hamiltonian laceable,HLe(Q3)= 1.

(3)

Proof. Let e be a faulty edge in Q3. Q3 can be split into Q02and Q12such that e is neither in E(Q02) nor in E(Q1

2). Suppose that x and y are with different colors. Case 1: x, y∈ V (Q02) or x, y∈ V (Q12). Without

loss of generality, we may assume that x, y∈ V (Q02). Q2 is a cycle of length four, so x and y are adja-cent. Let P = x, x0, x1, y be a path in Q02. Since |F | = 1, there exists an edge, denoted by (u, v), in this path such that (u, u ) and (v, v ) are fault-free

and u , v ∈ Q12. Obviously, u and v are adjacent. Let R= u , w, z, v be a path in Q12. Therefore,

E(P )∪ E(R) − {(u, v)} forms a hamiltonian path in Q3joining x and y.

Case 2: x∈ V (Q02) and y∈ V (Q12), or x∈ V (Q12)

and y∈ V (Q02). Without loss of generality, we may

assume that x∈ V (Q02) and y ∈ V (Q12). Since there

are two vertices in Q02 adjacent to x, we may choose a fault-free edge (u, v) such that u∈ V (Q02), u is

adjacent to x and v∈ V (Q12). Obviously, v and y

are adjacent. Let x, x0, x1, u be a path in Q02 and y, y0, y1, v be a path in Q12, respectively. Combining these two paths, we have a hamiltonian path in Q3 joining x and y.

Lemma 2. Q3is 1 edge fault-tolerant strongly hamil-tonian laceable, i.e.,HSLe (Q3)= 1.

Proof. Let e be a faulty edge in Q3. Q3 can be split into Q02and Q12such that e is neither in E(Q02) nor in E(Q12). Suppose that x and y are with the same color.

In order to prove this lemma, we will construct a fault-free path of length 6 joining x and y.

Case 1: x, y∈ V (Q02) or x, y∈ V (Q12). Without

loss of generality, we may assume that x, y∈ V (Q02).

Letx, u, y be a path in Q02such that (u, u ) is

fault-free and u ∈ Q12. There is a vertex v in {x, y} such that (v, v ) is fault free and v ∈ Q12. Without loss of generality, we assume that v= x. Obviously, u and v are adjacent. Letu , w, z, v be a path in Q12. Therefore,x, v , z, w, u , u, y forms a path of length

6 in Q3joining x and y.

Case 2: x∈ V (Q02) and y∈ V (Q12), or x∈ V (Q12)

and y∈ V (Q02). Without loss of generality, we may

assume that x∈ V (Q02) and y ∈ V (Q12). Let u be

another vertex in V (Q02) having the same color as x.

Also let v∈ V (Q12), v and y have the same color. So

these four vertices x, y, u, and v are all with the same color. And at least one of u and v is not an endpoint of the faulty edge e. We may assume that (u, u ) is

a fault-free edge such that u ∈ V (Q12). Obviously, u

and y are adjacent. Let y, y0, v, u be a path in Q12

andx, x0, u be a path in Q02, respectively. Therefore, x, x0, u, u , v, y0, y forms a path of length 6 joining

x and y.

Lemma 3. The hypercube Qn, n 2, is (n − 2) edge fault-tolerant hamiltonian laceable, i.e., HeL(Qn)= n− 2.

Proof. We prove this lemma by induction on n. First,

we observe that the lemma holds for n = 2. By Lemma 1, the lemma holds if n= 3. For n  4, we assume that the lemma is true for every integer k < n. Let x and y be two vertices with different colors in

Qn, i.e., x and y are in different partite sets. In order to prove that HLe(Qn) n − 2, we must construct a fault-free hamiltonian path joining x and y for any given faulty edge set F with|F | = n−2. Since fc 1, f0 n − 3 and f1 n − 3.

Case 1: x, y∈ V (Q0n−1) or x, y∈ V (Q1n−1). (See

Fig. 1(a).) Without loss of generality, we may as-sume that x and y are in Q0n−1. By induction hypoth-esis, HLe(Q0n−1)= n − 3, there exists a hamiltonian

path x, P0, y with 2n−1− 1 edges. We claim that

there exists an edge (u, v)∈ E(x, P0, y ) such that

both crossing edges (u, u ) and (v, v ) are fault-free.

Since |E(x, P0, y )| = 2n−1− 1, we have 2n−1− 1

choices. If none of the edges of x, P0, y meets the

requirements of such an edge (u, v), then there are at least (2n−1− 1)/2 faults in Fc. (Because a sin-gle fault in Fc eliminates 2 edges ofx, P0, y .) And

(2n−1−1)/2 > n−2 for n  4, this contradicts with the fact that |F |  n − 2. Therefore, we can always find such an edge (u, v). Obviously, u and v are with different colors. SinceHLe(Q1n−1)= n − 3, there

ex-ists a fault-free hamiltonian pathu , P1, v in Q1n−1. Therefore, Ex, P0, y ∪ Eu , P1, v  ∪  (u, u ), (v, v )−(u, v)

forms a hamiltonian path in Qn− F joining x and y. Case 2: x ∈ V (Q0n−1) and y∈ V (Q1n−1), or xV (Q1n−1) and y∈ V (Q0n−1). (See Fig. 1(b).) Without

(4)

Fig. 1. (a) Case 1: x, y∈ V (Q0n−1). (b) Case 2: x∈ V (Q0n−1) and y∈ V (Q1n−1). loss of generality, we assume that x∈ V (Q0n−1) and

y∈ V (Q1n−1). We claim that we can choose a vertex u in Q0n−1 such that u and x are with different colors, and the crossing edge (u, u ) is fault-free,

where u ∈ V (Q1n−1). Since |V (Q0n−1)| = 2n−1, we have 2n−2 choices. (Because there are 2n−2 vertices in Q0n−1 which have different colors from x.) If none of the vertices in Q0n−1 meets the requirements of such vertex u, then there are at least 2n−2 faults in

Fc. This contradicts with the fact that |F |  n − 2 for n 2. Therefore, we can always find such a vertex u. Then, u and y are with different colors. SinceHLe(Q0n−1)= HLe(Q1n−1)= n − 3, there exists

a hamiltonian pathx, P0, u and u , P1, y in Q0n−1 and in Q1n−1, respectively. Therefore,

Ex, P0, u ∪ Eu , P1, y 

∪(u, u )

forms a hamiltonian path in Qn− F joining x and y. This completes the induction. ✷

Given any faulty edge set F with|F | = n − 2, we can chose an edge (u, v) in Qn − F . By the proof above, there exists a hamiltonian path u, P, v in

Qn− F joining u and v. So it is easy to see that Eu, P, v ∪(u, v)

forms a hamiltonian cycle in Qn− F . Hence He(Qn)

= n − 2 if n  2.

Theorem 1. The hypercube Qn, n 2, is (n − 2) edge fault-tolerant strongly hamiltonian laceable, i.e.,

HSL

e (Qn)= n − 2.

Proof. By Lemma 3, we haveHLe(Qn)= n − 2. So all we have to show is the following. Let x and y be

two vertices with the same color in Qn. We must find a fault-free pathx, P, y of length 2n− 2 for any given faulty edge set F with|F | = n − 2. This theorem can be proved using by the same way of Lemma 3 and hence the detail proof is omitted. ✷

Theorem 2. The hypercube Qn, n 3, is (n − 3) edge fault-tolerant hyper-hamiltonian laceable, i.e.,

Hh

e(Qn)= n − 3.

Proof. The proof is again by using induction on n.

Lewinter et al. [7] proved that Qn, n 3, is hyper-hamiltonian laceable. So the lemma holds for the case n= 3. For n  4, we assume the theorem is true for every integer k < n. By induction hypothe-sis, Hhe(Qn−1)= n − 4. Now, we consider Qn. By Lemma 3, we haveHle(Qn)= n − 2. Obviously, Qn is hamiltonian laceable after removing n− 3 edges. In order to prove thatHeh(Qn) n−3, it suffices to show the following. After deleting a given vertex w from

Qn, let x and y be any two vertices in the larger par-tite set of Qn. We must construct a hamiltonian path of Qn− F − w joining x and y for any given faulty edge set F with|F | = n − 3. Without loss of general-ity, we may assume that w∈ V (Q0n−1). Since fc 1, f0 n − 4 and f1 n − 4.

Case 1: x, y∈ V (Q0n−1). (See Fig. 2(a).) By

in-duction hypothesis, Hhe(Q0n−1)= n − 4, there exists

a fault-free hamiltonian pathx, P0, y with 2n−1− 2

edges in Q0n−1 − w. We now show that there ex-ists an edge (u, v) ∈ E(x, P0, y ) such that both

crossing edges (u, u ) and (v, v ) are fault-free. Since

|E(x, P0, y )| = 2n−1− 2, we have 2n−1− 2 choices.

If none of the edges of x, P0, y meets the

(5)

Fig. 2. (a) Case 1: x, y∈ V (Q0n−1). (b) Case 2: x, y∈ V (Q1n−1). (c) Case 3: x∈ V (Q0n−1) and y∈ V (Q1n−1). 2n−2 − 1 faults in Fc. (Because a single fault in Fc

eliminates 2 edges of x, P0, y .) And 2n−2 − 1 > n− 3 for n  4, this contradicts with the fact that

|F |  n − 3. Therefore, we can always find such an edge (u, v). Obviously, u and v are with different col-ors. SinceHLe(Q1n−1)= n − 3, there exists a fault-free

hamiltonian pathu , P1, v in Q1n−1. Therefore,

Ex, P0, y ∪ Eu , P1, v 

∪ 

(u, u ), (v, v )−(u, v)

forms a hamiltonian path in Qn− F − w joining x and y.

Case 2: x, y∈ V (Q1n−1). (See Fig. 2(b).) First, we

will choose a vertex u in Q1n−1 such that u and x are with different colors, and the crossing edge (u, u ) is

fault-free, where u ∈ V (Q0n−1). Since|V (Q1n−1)| =

2n−1, we have 2n−2choices. (Because there are 2n−2 vertices in Q1n−1which have different colors from x.) If none of the vertices in Q1n−1meets the requirements of such vertex u, then there are at least 2n−2 faults in Fc. This contradicts with the fact that|F |  n − 3 for n 2. Therefore, we can always find such a vertex u. SinceHhe(Q1n−1)= n−4, there exists a

fault-free hamiltonian path x, P, y in Q1n−1 − u joining

x and y. We claim that there exists a vertex t such

that the edge (u, t) is fault-free in Q1n−1, the edge

(t, v)∈ E(x, P, y ), and the crossing edge (v, v ) is

fault-free in Qn where v ∈ Q0n−1. Since the number of neighboring vertices of u in Q1n−1is n− 1, we have

n− 1 choices. If none of the vertices in Q1n−1 meets the requirements of such a vertex t , then there are at least n− 1 faults in Fc, making it contradictory to the fact that|F |  n − 3. Therefore, we can always find such a vertex t .

We then divide the pathx, P, y into two sections x, P0, v and t, P1, y , or x, P0, t and v, P1, y .

Without loss of generality, we assume the casex, P0, t

andv, P1, y . Thus, we have two sections as x, P0, u

andv, P1, y . Let (u, u ) and (v, v ) be two crossing

edges incident to vertices u and v, respectively. Then,

u and v are with the same color in Q0n−1. Since u and x are with different colors, u and v are in the larger partite set of Q0n−1− w. By induction hypothe-sis,HLe(Q0n−1)= n−4, there exists a fault-free

hamil-tonian pathu , R, v in Q0n−1. Therefore,

Ex, P0, t ∪ Eu , R, v ∪

Ev, P1, y ∪(u, u ), (v, v ), (t, u)

forms a hamiltonian path in Qn− F − w joining x and y.

Case 3: x ∈ V (Q0n−1) and y∈ V (Q1n−1), or xV (Q1n−1) and y∈ V (Q0n−1). (See Fig. 2(c).) Without

loss of generality, we assume that x∈ V (Q0n−1) and y∈ V (Q1n−1). First, we will choose a vertex u, u= x,

in Q0n−1 such that u and x are with the same color, and the crossing edge (u, u ) is fault-free, where uV (Q1n−1). Since|V (Q0n−1)| = 2n−1, we have 2n−2−1 choices. (Because there are 2n−2−1 fault-free vertices in Q0n−1 which have the same color as x.) If none of the vertices in Q0n−1 meets the requirements of such vertex u, then there are at least 2n−2− 1 faults in

Fc. This contradicts with the fact that|F |  n − 3 for n 3. Therefore, we can always find such a vertex u.

Then, u and y are with different colors. By induction hypothesis, Heh(Q0n−1)= n − 4, there exists a

fault-free hamiltonian path x, P0, u in Q0n−1 joining x and u. Also, sinceHLe(Q1n−1)= n − 3, there exists a

(6)

fault-free hamiltonian pathu , P1, y in Q1n−1joining u and y. Therefore,

Ex, P0, u ∪ Eu , P1, y 

∪(u, u )

forms a hamiltonian path in Qn− F − w joining x and y. The proof is complete.

References

[1] J.A. Bondy, U.S.R. Murty, Graph Theory with Applications, North-Holland, New York, 1980.

[2] F. Harary, M. Lewinter, Hypercubes and other recursively defined hamilton laceable graphs, Congr. Numer. 60 (1987) 81– 84.

[3] S.Y. Hsieh, G.H. Chen, C.W. Ho, Hamiltonian-laceability of star graphs, Networks 36 (2000) 225–232.

[4] S.Y. Hsieh, G.H. Chen, C.W. Ho, Fault-free hamiltonian cycles in faulty arrangement graphs, IEEE Trans. Parallel Distributed Systems 10 (32) (1999) 223–237.

[5] S. Latifi, S. Zheng, N. Bagherzadeh, Optimal ring embedding in hypercubes with faulty links, in: Fault-Tolerant Computing Symp., 1992, pp. 178–184.

[6] F.T. Leighton, Introduction to Parallel Algorithms and Architec-ture: Arrays, Trees, Hypercubes, Morgan Kaufmann, San Ma-teo, CA, 1992.

[7] M. Lewinter, W. Widulski, Hyper-Hamilton laceable and caterpillar-spannable product graphs, Comput. Math. Appl. 34 (1997) 99–104.

[8] G. Simmons, Almost all n-dimensional rectangular lattices are Hamilton laceable, Congr. Numer. 21 (1978) 103–108.

數據

Fig. 1. (a) Case 1: x, y ∈ V (Q 0 n −1 ). (b) Case 2: x ∈ V (Q 0 n −1 ) and y ∈ V (Q 1 n −1 )
Fig. 2. (a) Case 1: x, y ∈ V (Q 0 n −1 ). (b) Case 2: x, y ∈ V (Q 1 n −1 ). (c) Case 3: x ∈ V (Q 0 n −1 ) and y ∈ V (Q 1 n −1 )

參考文獻

相關文件

Define instead the imaginary.. potential, magnetic field, lattice…) Dirac-BdG Hamiltonian:. with small, and matrix

In x 2 we describe a top-down construction approach for which prototype charge- qubit devices have been successfully fabricated (Dzurak et al. Array sites are de­ ned by

• Learn strategies to answer different types of questions.. • Manage the use of time

Lu, Linear-time compression of bounded-genus graphs into information-theoretically optimal number of bits, in Proceedings of the Thirteenth Annual ACM-SIAM Symposium on

 BayesTyping1 can tolerate sequencing errors, wh ich are introduced by the PacBio sequencing tec hnology, and noise reads, which are introduced by false barcode identi cations to

In this work, we will present a new learning algorithm called error tolerant associative memory (ETAM), which enlarges the basins of attraction, centered at the stored patterns,

In this thesis, we develop a multiple-level fault injection tool and verification flow in SystemC design platform.. The user can set the parameters of the fault injection

Singleton,”A methd for computing the fast Fourier Transform with auxiliary memory and limited high-speed storage”, IEEE Trans. Audio