10.1.1.24.7784
TRANSCRIPT
-
7/28/2019 10.1.1.24.7784
1/8
Parallel Routing in Hypercube Networks with Faulty Nodes
Eunseuk Oh Jianer ChenDepartment of Computer Science, Texas A&M University
College Station, TX 77843-3112, USA
eunseuko, chen
@cs.tamu.edu
Abstract
The concept of strong fault-tolerance was introduced to
characterize the property of parallel routing [15]. A net-
work of degree is said strongly fault-tolerant if with at
most
faulty nodes, any two nodes
and
in
are
connected by
node-disjoint paths,
where
and
are the numbers of non-faulty
neighbors of the nodes and in , respectively. We show
that the hypercube networks are strongly fault-tolerant and
develop an algorithm that constructs the maximum number
of node-disjoint paths in a hypercube network with faults.
Our algorithm is optimal in terms of time and length of
node-disjoint paths.
1. Introduction
Parallel routing on large size networks with faults is an
important issue in the study of computer interconnection
networks, which allows networks to have alternative routes
to tolerate faulty nodes. The new concept of the network
strong fault tolerance was introduced to measure fault toler-
ance for interconnection networks and has been studied for
the star networks [15]. In the current paper, we continue the
study of the strong fault tolerance of interconnection net-
works, particularly, of the popular hypercube networks.
The
-dimensional hypercube
has been studied ex-
tensively by many researchers as an interconnection net-
work topology for multicomputer systems. Parallel routing
on hypercube networks without faulty nodes was first stud-
ied in [18]. An algorithmthat constructsnode-disjoint paths
between disjoint source-destination pairs was proposed in
[8, 14]. The problem of determining the diameter of hyper-
cube networks with faults was considered in [10, 11]. Many
This work is supported in part by the National Science Foundation
under Grant CCR-0000206.
fault-tolerant communication algorithms concentrating on
one-to-one routing or broadcasting in hypercube networks
have been proposed [2, 4, 6, 7, 12, 13, 16, 17].
A network of degree is said strongly fault-tolerant
[15] if with at most faulty nodes, any two nodes
and
in
are connected by
node-disjoint paths, where
and
are the numbers
of non-faulty neighbors of the nodes and in , respec-
tively.
The study of strong fault tolerance in the star networks
showed that node-disjoint paths can be constructed effi-
ciently based on the orthogonal partition of the star net-
works with faults, which decomposes the
-star network
into -dimensional substar networks and an
independent set of nodes [3]. Roughly speak-
ing, a path from a non-faulty neighbor of the source node
to a non-faulty neighbor of the destination node is con-
structed in a separated -dimensional substar, and theindependent set
helps the paths to enter the substar from
a proper node. The algorithm proposed in [15] constructs
the maximum number of node-disjoint paths of nearly opti-
mal length in the
-star networks with at most
faulty
nodes.
We observe that the techniques used in the previous
studying for star networks [15] are not applicable to the
case for hypercube networks. Specifically, the hypercube
networks do not seem to have similar orthogonal decompo-
sition structure. Parallel routing in the
-dimensional hy-
percube networks may require constructing node-disjoint
paths, while an -dimensional hypercubes can be decom-
posed into at most
-dimensional subcubes. There-
fore, there may be no extra nodes available that can help to
distribute the paths into the subcubes.
We develop new techniques that construct node-disjoint
paths between pairs of neighbors of the source node
and
the destination node . First, a prematching process pairs
non-faulty neighbors of and in
. For given pairs
-
7/28/2019 10.1.1.24.7784
2/8
of neighbors of and , we introduce three procedures to
construct paths by permutations of edge sequences between
them. Node-disjoint paths are constructed by searching
properpaths, ensuring that each node in a path is notused by
other paths. Our algorithm constructs node-disjointpaths in
optimal time and the length of paths is also optimal in the
hypercube network
: For any two non-faulty nodes and
in
, the algorithm constructs min
node-disjoint paths of minimum length plus 4 between
and in time .
The paper is organized as follows. Notations and termi-
nology are introduced in section 2. In section 3, we dis-
cuss parallel paths between two non-faulty nodes. In case
there are no faulty neighbors for both sourceand destination
nodes, we pre-pair the neighbors of the source and the des-
tination nodes by a process, called Prematch-I. There is a
special situation that may block all possible sets of parallel
paths between two neighbors of the source and the destina-
tion nodes induced from Prematch-I. In this situation, we
use a different process, called Prematch-II instead. The
third process Prematch-III covers the case in which there
is at least one faulty neighbor of the source or the destina-
tion. The algorithm is presented and discussed in section 4
and the final section concludes the paper.
2. Preliminaries
An
-dimensional hypercube
is an undirected graph
consisting of nodes represented by binary numbers from
to
, and
edges connecting nodes whose bi-
nary representations differ in exactly one bit. An edge is
called an -edge if two nodes connected by it differ in the
th bit (the first bit is the leftmost bit). The Hamming dis-
tance between two nodes and , is the length
of the shortest path from to . Actually, is the
number of bits in which binary representations of
and
differ. Since the hypercube
is vertex-symmetric, a
set of node-disjoint paths from a node to a node can
be mapped to a set of node-disjoint paths from the node
to the node in a straightforward way,
where . Therefore, we will concentrate on
the construction of node-disjoint paths from the node
to
the node in
.
The node connected from the node by an -edge is
denoted by
, and the node connected from the node
by a -edge is denoted by
. A path from the node
to the node can be uniquely specified
by a sequence of labels of the edges on
in the order of
traversal. In particular, a path from the node to the node
that uses an
-edge, an
-edge, , an
-edge, in that or-
der, will be denoted by
. For example, for
the nodes and , spec-
ifies the path
. We extend this notation for a single permutation to
a set of permutations, as follows. Let be a set of permu-
tations, then the notation
denotes the set of paths:
is a permutation in
For example, suppose , ,
, , then consists of four paths
from
to
:
,
,
, and
.
We say that an edge
does not lead to a shortest
path to a node
if
. The
following fact can be easily verified.
Fact 2.1 If an edge
in
does not lead to a short-
est path to
, then
. In
general, if in a path from a node to a node , thereare exactly
edges that do not lead to a shortest path to
,
then the length of the path is equal to
.
It is known [18] that for any two nodes
and
in
, there exist node-disjoint paths such that
of these paths are of length , and the remaining
paths are of length
.
3. Parallel paths between two non-faulty nodes
In this section, we show how a set of paths between two
non-faulty nodes
and
in the hypercube network
canbe constructed.
Our parallel routing algorithm is based on an effective
pairing of the neighbors of the node
and
. We first assume that the nodes and have no
faulty neighbors. We pair the neighbors of
and
by the
following strategy.
Prematch-I
Assumption: and have no faulty neighbors.
1. pair
with
for 1 ;
2. pair
with
for
;
Under the pairing given by Prematch-I, we constructparallel paths between the paired neighborsof
and
using
the following procedure.
1The calculation for indices between 1 and
can be given by a rather
lengthy formula based on modular operation. For simplicity, we only need
to remember the following three special cases: Let be an index between
1 and . (1) for , is interpreted as and is interpreted as
; (2) for , is interpreted as ; and (3) for , is
interpreted as 1.
-
7/28/2019 10.1.1.24.7784
3/8
u = 1110
1010
1000
1011
1001 v = 0000
1111
0110
0111
0101
0001
0100
0010 0011
1100 1101
2
4
3
Figure 1. Parallel paths between and
in
with two faulty(dark) nodes.
Procedure-I
1. for , and the paired neighbors
and
, we construct node-disjoint
paths between
and
, which consist of
paths of the form
(1)
where
is the set ofall cyclicpermutations
of the sequence ,
and of paths of the form
(2)
for all
,
.
2. for , and the paired neighbors
and
, we construct
node disjoint
paths between and , which consist of
paths of the form
(3)
where
is the set ofall cyclicpermutations
of the sequence , and of
paths of the form
(4)
for all , and .
The paths constructed by cyclic permutations of a se-
quence are pairwisely disjoint(see, for example [18]). It is
easy to verify that for each pair of neighbors of
and
,
the paths constructed between them are pairwisely disjoint.
Figure 1 shows an example of parallel paths in
with two
faulty(dark) nodes. For a path
from
to
, we define
as the node on the path
starting from
and
following the edge labels in .
Lemma 3.1 Let
and
be two pairs given by
Prematch-I. Then, there is at most one path in the path set
constructed by Procedure-I for the pair
that share
common nodes with a path in the path set constructed by
Procedure-I for the pair
.
PROOF. For two paths
and
such that
is for
the pair
and
is for the pair
,
,
assume that
and
have a common node. Then, the
same set of bits in
and
are different
from those of
. Since
th bit and
th bit in the common
node are different from those of , it must be of the form
. We show below that
this node must have the form
. Thus, the path
is uniquely determined by
and
.
Case 1. Suppose
and
.
Suppose the common node is of the form
, then
must
be
(if
),
(if
), or
for
some . If
then the node
has
th bit identical to that of
while the node
has th bit different from that of .
If
and then there is no node of form
. Thus, the index
does not exist, and
must be of the form
. If
then
must be of the form
.
In that case, the sequences in
and
are constructed
by cyclic permutations of a sequence except
the index . It has been known that paths constructed
by cyclic permutations of a sequence are disjoint. This
property still holds when the index is added such as
. It contradicts the assumption that
and
has a common node. Thus, the index
does not
exist.
Case 2. Suppose
and
, or
and .
First assume
and
. The se-
quence in the path
must be of the form
for some since . Since is the only index larger
than
in this sequence and
, we must have
.
Thus, the path
must be of the form
.
The case and can be proved by
symmetry.
Case 3. and .
The sequences in
and
cannot be a permutation of
because , . Thus,
must be of the form
and
is
for some
. In that case,
should be
and
should be
because and are the only indices larger than . Thus,
the path
is of the form
.
-
7/28/2019 10.1.1.24.7784
4/8
Combining all cases, we complete the proof.
Fact 3.1 For a pair
,
given by
Prematch-I, a path of the form
has no common nodes with any other paths con-
structed by Procedure-I.
We have shown that for each paired nodes by Prematch-
I, the algorithm Procedure-I constructs at least dis-
joint paths between them. Since there may be up to
faulty nodes, in the worst case, there can be a pair
of nodes by Prematch-I, for which all paths con-
structed by Procedure-I are blocked. In this case, we pair
the neighbors of
and
by the following rule:
Prematch-II
Assumption: there is a pair
,
given by Prematch-I such that all
paths constructed by Procedure-I for
are blocked by faulty nodes.
1.
is paired with
;
2.
is paired with
;
3.
is paired with
;
4. forother neighborsof and , use Prematch-I
In Prematch-II, operations on indices between 1 and
are by mod
. For each pair given by Prematch-II, we con-
struct a path as follows.
Procedure-II
1. For a pair of form
, the path is
;
2. For a pair of form
, the path is
;
3. For a pair of form
, the path is
;
4. For other pairs, use Procedure-I to con-
struct paths between them: For pair
,
,
, the
path is
;
For pair
,
, the
path is
if
, and
if .
Lemma 3.2 Under the conditions ofPrematch-II, the al-
gorithm Procedure-II constructs fault-free parallel paths
of length
from
to
.
PROOF. It easy to see that Paths constructed by
Procedure-II have length bounded by . Ex-
cept paths of form
or
,
, whose length is
, other paths have length
.
Now we show that all
pathsconstructed by Procedure-
II are fault-free. After that, we show that these paths
are disjoint. Recall that all possible paths constructed by
Procedure-I for the pair
are blocked by faulty
nodes. Denote these paths by
.
The path
and
only share a node
because every node in
has its th bit identical to that of while nodes ex-
cept
in
has
th bit different that of
. Since
the node
is non-faulty, the path
is fault-free. The
path
and
have no common nodes because
th bits in nodes in
and
are not identical. Thus,
is fault-free. The
path
and
only share a node
because nodes except
(=
) have th bit identical to that of
while nodes in
have th bit different that of . A
path of form
,
has no common nodes with any other paths
constructed by Procedure-I by fact 3.1. Since
, the
path
is fault-free. Finally, consider a path
constructed
for a pair
, . If , all faulty
nodes are in paths between
and
and
is of the form
. Sincea node
in the path
can be identical to only a node of form
,
by lemma 3.1, the path
and
haveno common nodes.
In case , all faulty nodes are in paths between
and
and
is of the form
. Similarly, we can
prove that
is fault-free sincea node
in
can
be identical to only a node of form
.
Therefore, all paths constructed by Procedure-II are
fault-free.
Now we show that paths constructed by Procedure-II
are disjoint.
It is easy to see that
and
have no common nodes
because of an index . Similarly,
and
have no
commonnodes because of an index . Also,
and
have no common nodes because nodes except
in
have th bit identical to that of while nodes in
have th
bit different that of , and the node
is not included in
. Thus, paths
,
, and
are disjoint. Each path
of the form
, is
disjoint by fact 3.1. To show that paths constructed for pairs
,
are disjoint with paths
,
, and
, suppose
is the sequence of the
path
. Since ,
becomes and the path
-
7/28/2019 10.1.1.24.7784
5/8
1010
1000
1011
1001 v = 0000
0110
0111
0101
0001
0100
0010 0011
1100 1101
4
1110
u = 1111
3
2
1
Figure 2. Paths constructed by Procedure-II
between and in
.
cannot have common nodes with the path
. Suppose
is the sequenceof
, then since ,
becomes
and
and
have no common nodes.
Also if
is the sequence of
, then since
,
becomes
when
and
becomes
when
. For both cases, it is easy to see that
and
have no common nodes. Finally, each path of
the form
or
,
is disjoint with other paths because of a unique index .
Therefore, all paths constructed by Procedure-II are
pairwisely disjoint. Figure 2 shows paths constructed by
Procedure-II between
and
with two
faulty(dark) nodes in
.
So far, we have assume that all neighbors of the source
node and the destination node are non-faulty. Now we
relax such a restriction to deal with any faulty neighbors
of two nodes
and
. We introduce Prematch-III to pairthe edges incident on the neighbors of the nodes
and
,
instead of the neighbors of
and
.
Prematch-III
Assumption: or have at least one faulty
neighbor, and edges are paired only when all
nodes on them are non-faulty.
for each edge
where both
and
are non-faulty do
1. if and , then pair
with the edge
;
2. if and , then pair
with the edge
;
3. otherwise, pair
with
,
where the indices and are such that
Prematch-I pairs the node
with
, and
the node
with
.
Note that it is possible that an edge
with both
and
non-faulty is not paired with any edge because the
corresponding edge in Prematch-III contains faulty nodes.
For each pair of edges givenby Prematch-III, we construct
a path between them by thealgorithmcalled Procedure-III.
Basically, sequences in paths constructed in Procedure-III
follow Procedure-I if there is no comment for that. We
assume that for edges
, , their
pairs are given in the increasing order of
, and operations
on indices between 1 and are by mod .
Procedure-III
1. for and and paired
edges
,
, construct
the path
;
2. for and and
paired edges
,
,
construct a path by flipping
and
in the
path
;
3. otherwise, for paired edges
,
, if , construct a path by flip-
ping
and
in the path
; if
, construct a path by flipping and
in the path
.
Notice that Procedure-III forces the sequence between
a pair induced by Prematch-III to be obtained from a path
constructed by Procedure-I such that is of the form
, where
or when
.Thus, consider paths of the form
constructed
by Procedure-I such that or when .
The following fact shows that any two paths of the forms
and
as described above have no
nodes in common. It comes directly from lemma 3.1.
Fact 3.2 Let
be a path of the form
such
that
or
when
as given in Procedure-I.
For a path
of the form
, ,
has no
nodes in common.
Lemma 3.3 For a non-faulty node
, Procedure-III con-
structs at most fault-free one-to-many disjoint paths
from the node
to all non-faulty nodes
,
.
PROOF. For all non-faulty edges of form
,
, consider edges paired by Prematch-III. Sup-
pose . For ,
is paired with
when , and
-
7/28/2019 10.1.1.24.7784
6/8
when . Otherwise,
is paired with
such that
is paired with
by Prematch-I.
Since
is uniquely paired with
and
by Prematch-I,
there are at most non-faulty edge pairs constructed
by Prematch-III. Suppose . Since for
, a non-faulty node
,
is uniquely
paired with a node
by Prematch-I, there are at most
non-faulty edge pairs constructed by Prematch-III.
Now, consider paths between above edge pairs. All paths
constructed by Procedure-III are obtained from paths con-
structed by Procedure-I such that these paths are of the
form
, or when , flipping
first two indices or last two indices. We have shown that
such paths of the form
,
or
when
, are disjoint in fact 3.2. Thus, flipping first two
indices on such paths make paths share the node
, and
flipping last two indices on these path make paths have all
different entering nodes to which are given by Prematch-
III. Thus, Procedure-III constructs at most fault-free
one-to-many disjoint paths from
to all non-faulty nodes
, .
Further, the following fact can be easily verified from the
above discussion.
Fact 3.3 For any two edge pairs
and
given by Prematch-III, paths
constructed by Procedure-III for these pairs have common
nodes of form
, where
and
.
4. Parallel routing algorithm on faulty hyper-cube networks
First, consider the lower bound of the length of the
min
node-disjoint paths from a node
to a node
in hypercube
, where
. Suppose a neighbor node of
,
be non-faulty, and we want to find a path from to
via
. Assume that all neighbors of
are faulty except
two nodes
and
. Then, a
fault-free path of the form from to has
length at least
. Thus, the length of the
min
disjoint paths from to is at least
.
We now ready to present our main algorithm for par-
allel routing in the hypercube networks
with at most
faulty nodes. For two non-faulty nodes
and
in
, our algorithm constructs
node-disjoint fault-free paths from
to such that the length of the paths is bounded by
Parallel-Routing
Input: non-faulty nodes
and
in
with at most
faulty nodes.
Output:
parallel fault-free paths of
length
from
to
.
1. case 1.
and
have no faulty neighbors
for each pair
given by Prematch-I do
1.1 if all paths for
by Procedure-I include faultynodes
then use Prematch-II and Procedure-II to construct
parallel paths from to ; STOP.
1.2 if there is a fault-free unused path from
to
by
Procedure-I
then mark the path as used by
;
1.3 ifall fault-free paths constructed for
include
used nodes
then pick any fault-free path for
, and for the
pair
that uses a node on , find a new
path;
2. case 2. and have at least one faulty neighbor
for each edge pair
given by
Prematch-III do
2.1 ifthere is a fault-free unused path from
to
by
Procedure-III
then mark the path as used by the pair
,
,
,
;
2.2 ifall fault-free paths constructed for the pair include
used nodes
then pick any fault-free path
for the edge pair, and
for the edge pair that uses a node on
, find a new
path;
Figure 3. Parallel routing on the hypercube
network with faulty nodes
. The algorithm called Parallel-Routing is
given in Figure 3.
Lemma 3.2 guarantees that step 1.1 of the algorithm
Parallel-Routing constructs
fault-free parallel paths oflength
from
to
. Step 1.3 of the algo-
rithm requires further explanation. In particular, we need to
show that for the pair
, we can always construct
a new fault-free path from
to
in which no nodes
are used by other paths. This is ensured by the following
lemma.
Lemma 4.1 Let
and
be two pairs given
by Prematch-I such that two paths constructed for
and
share a node. Then the algorithm Parallel-
Routing can always find fault-free paths for
and
, in which no nodes are used by other paths.
PROOF. We assume that we will search a fault-free and
unused path for each pair given by Prematch-I in order of
cyclic permutations as given in Procedure-I in the follow-
ing discussion. For example,
,
,
,
.
Suppose all fault-free paths constructed for
in-
clude used nodes, and one of fault-free path is picked for
-
7/28/2019 10.1.1.24.7784
7/8
, which includes a node used for
, .
Then we want to show that we can always find a fault-free
unused path for
.
Since all fault-free paths constructed for
are
used, all unused paths must be blocked by faulty nodes. By
lemma 3.1, a node of form
can be identical to
a node of form
. Thus, all paths leaving with
from
are unused, and there are
such paths. Remain paths of form
,
are used or faulty. Since a used path means
that unused paths previously searched before it also should
be blocked by faulty nodes, when we search new path for
among paths unused by other pairs, there exists
at least one unused fault-free path which can be used for
. From the above discussion, we can easily show
that for a pair
, , there exist at least one unused
fault-free path between them. Therefore, step 1.3 of the
algorithm Parallel-Routing occurs once during the whole
execution.
A similar analysis shows that step 2.2 of the algorithm
Parallel-Routing can always construct a new fault-free
path without nodes used by other paths.
Lemma 4.2 Let
,
and
,
, be edge pairs given by Prematch-III such that
two paths constructed for them share a node. Then the al-
gorithm Parallel-Routing always find fault-free paths be-
tween them, in which no nodes are used by other paths.
We summarize all these discussions in the following the-orem.
Theorem 4.3 If the hypercube network
has at most
faulty nodes, then for each pair of non-faulty nodes
and in
, in time the algorithm Parallel-Routing
constructs
node-disjoint fault-free
paths of length bounded by from to .
PROOF. We first discuss the length of
,
node-disjoint fault-free paths. It is easy to see
that paths constructed by Procedure-I is length of at most
. If paths are constructed by Procedure-II or
Procedure-III, the length is still at most be-
cause all paths constructed by Procedure-II or Procedure-
III are constructedbasedon Procedure-I, only flipping first
or last two indices in paths.
We now discuss the time complexity the algorithm
Parallel-Routing.
For each pair given by Prematch-I, a path is constructed
by the algorithm by searching a proper path in a set of paths
1010
1000
1011
1001 v = 0000
0110
0111
0101
0001
0100
0010 0011
1100 1101
4
11110
u = 1111
3
2
Figure 4. Paths constructed by Parallel-
Routing between and in
between them, which takes time
, where
is the number of faulty nodes in the set of paths for the pair
. If we find a fault-free and unused path of the form
for a pair
, , then mark a node
as an used node. In such a way, we can detect usedpaths in time
since thereare at most
used paths for
. If all fault-free paths for the pair
include
used nodes, we pick any fault-free path for
, and
for the pair
that uses a node on
, find a new path.
As we have discussed in previous lemma, this happens once
during the whole execution. Thus, the time complexity is
bounded by
since the
number
is bounded by .
If for a pair
given by Prematch-I, all possible
paths are blocked by faulty nodes, we simply ignore all
paths constructed for other pairs
, , and ap-
ply Procedure-II. Thus, it takes additional
time to
construct paths for pairs given by Prematch-II.
For pairs given by Prematch-III, paths are constructed
in the similar way to construct paths for pairs given by
Prematch-I. Thus, without detail explanations, we con-
clude that the time complexity for constructing paths be-
tween non-faulty neighbors of
and
is bounded by
. Figure 4 shows paths constructed by Parallel-
Routing between and in
.
5 Conclusion
Network strong fault tolerance is a natural extension of
the study of network fault tolerance and network parallel
routing. In particular, it studies the fault tolerance of large
size networks with faulty nodes. In this paper, we have
studied the strong fault tolerance of the popular hypercube
networks, and shown that hypercube networks are strong
fault tolerant. We developed an time algorithm that
for two given non-faulty nodes and in a -dimensional
-
7/28/2019 10.1.1.24.7784
8/8
hypercube
with at most faulty nodes, constructs
node-disjoint fault-free paths from
to
such that the length of the paths is bounded by
. The time complexity of our algorithm is opti-
mal since each path from to may have length as large as
, and there can be as many as
node-disjoint paths from
to . Thus, even printing these paths should take time
. The length of the paths constructed by our algo-
rithm is also optimal, as we can construct pairs of nodes
and in the hypercube
with faulty nodes for which
any set of parallel paths connecting and has at least
one path of length
. Finally, our algorithm
does not require prior knowledge of the failures.
Strong fault tolerance for networks with bounded de-
gree, such as ring networks, mesh networks, and butterfly
networks, are relatively easier. On the other hand, strong
fault tolerance for unbounded degree networks, such as net-
works based on Cayley graphs, seems much more difficult.
The hypercube networks and the star networks are the first
two such classes of networks whose strongly faulty toler-
ant have been proved. For star networks, the strong fault
tolerance was proved based on the orthogonal partition of
the star networks, while for hypercube networks, the strong
fault tolerance was proved by careful pre-matching of the
neighbors of the source and destination nodes. It will be
interesting to study the strong fault tolerance of other hier-
archical networks with unbounded degree.
References
[1] G. Birkhoff and S. MacLane, A Survey of Modern
Algebra, The Macmillan Company, New York, 1965.
[2] G. -M. Chiu and S. -P. Wu, A fault-tolerant routing
strategy in hypercube multicomputers, IEEE Trans.
Computers 45, 1996, pp. 143-154.
[3] C. C. Chen and J. Chen, Nearly optimal one-to-many
parallel routing in star networks, IEEE Trans. Paral-
lel, Distrib. Syst. 8, 1997, pp. 1196-1202.
[4] M. -S. Chen and K. G. Shin, Adaptive fault-tolerant
routing in hypercube multi-computers, IEEE Trans.
Computers 39, 1990, pp. 1406-1416.
[5] K. Day and A. Tripathi, A comparative study of
topological properties of hypercubes and star graphs,
IEEE Trans. Parallel, Distrib. Syst. 5, 1994, pp. 31-38.
[6] P. Fraigniaud, Asymptotically optimal broadcasting
and gossiping in faulty hypercube multicomputers,
IEEE Trans. Computers 41, 1992, pp. 1410-1419.
[7] Q. -P. Gu and S. Peng, Optimal algorithms for node-
to-node fault tolerant routing in hypercubes, The
Computer Journal 39, 1996, pp. 626-629.
[8] Q. -P. Gu and S. Peng, An efficient algorithm for
the -pairwise disjoint paths problem in hypercubes,
J. Parallel and Distributed Computing 60, 2000, pp.764-774.
[9] P. Hall, On representatives of subsets, J. London
Math. Soc. 10, 1935, pp. 26-30.
[10] M. S. Krishnamoorthy and B. Krishnamurthy, Fault
diameter of interconnection networks, Comput.
Math. Appl., 13, 1987, pp. 577-582.
[11] S. Latifi, Combinatorial analysis of the fault-diameter
of the -cube, IEEE Trans. Computers 42, 1993, pp.
27-33.
[12] S. Lee and K. Shin, Interleaved all-to-all reliable
broadcast on meshes and hypercubes, Proc. Intl
Conf. Parallel Processing, 1990, pp.III-110-113.
[13] T. C. Lee and J. P. Hayes, Routing and broadcast-
ing in faulty hypercube computers, Proc. third Conf.
Hypercube Concurrent Computers and Applications,
1988, pp. 625-630.
[14] S. Madhavapeddy and I. H. Sudborough, A topolog-
ical property of hypercubes: node disjoint paths, 2nd
IEEE Symposium on Parallel and Distributed Process-
ing, 1990, pp. 532-539.
[15] E. Oh and J. Chen, Strong fault-tolerance: paral-
lel routing in star networks with faults, Tech. Re-
port, Dept. Computer Science, Texas A M Univer-
sity, 2001.
[16] S. Park and B. Bose, All-to-all broadcasting in faulty
hypercubes, IEEE Trans. computers 46, 1997, pp.
749-755.
[17] P. Ramanathanand K. Shin, Reliable broadcast in hy-
percube multicomputers, IEEE Trans. Computers 37,
1988, pp. 1654-1657.
[18] Y. Saad and M. H. Schultz, Topological properties of
hypercubes, IEEE Trans. Computers, 37, 1988, pp.
867-872.