1. Introduction and Preliminaries
The accessibility of low-cost, efficient microprocessors and memory chips has recently encouraged researchers to create and work on multiprocessor interconnection networks. Interconnection networks are used to exchange data among the various processors in multistage networks. Thus, the implementation of a system involving multiprocessors depends upon interconnection networks. The key parameters driving the performance of the interconnection network are symmetry and simplicity of its structure and efficiency in routing and broadcasting messages. The fault tolerance in connectivity is also an important factor that improves the working capacity of an interconnection network. Chen et al. [
1] introduced a unique addressing scheme for the nodes of hexagonal meshes by wrapping these meshes systematically, which facilitates routing and broadcasting. The same systematic wrapping was used by Olson et al. [
2] for fault tolerant routing in torus and hexagonal meshes. In [
3], new addressing, routing and broadcasting algorithms were proposed for honeycomb networks of higher dimensions. In [
4], the routing algorithm partitions the network into subnets which are further divided into segments providing more freedom in placing turn restrictions as compared to other algorithms, which ensures deadlock freedom and better connectivity among the subnets. In [
5], region-based routing puts the destinations into subnets allowing a significant reduction in the sizes of routing tables. Recently, Azhar et al. [
6] has discussed the use of fault tolerant addressing schemes in mesh-related networks. Bossard et al. [
7] discussed the occurrence of faults as clusters in the larger networks and focused on the topological properties of the interconnection networks. For further study on addressing, routing and broadcasting see [
8,
9,
10,
11].
Silicates are produced by the chemical reaction of sand with metal oxides and metal carbonates. The basic building block of these silicates is the tetrahedron structure
arranged in a different symmetrical manner. The end nodes in a
structure depict oxygen atoms and the node at the center represents the silicon atoms which are called oxygen nodes and silicon nodes, respectively. Deleting silicon atoms from silicate networks gives different types of oxide networks. The oxide networks under discussion possess all the aforementioned characteristics of the interconnection networks. The
interconnection networks have order (number of nodes)
and
edges, where
m is the number of edges in a row line. The
interconnection networks have order
and
edges with
n corner vertices. The
interconnection networks of order
and
edges with
n number of row lines. This shows that the order of these networks is at most quadratic functions whereas in hypercube we get exponential function. The symmetry in these structures helps to divide nodes into certain clusters or regions. The fault tolerant addressing schemes assign unique code (representation) to each node of the network in terms of the distances from other subnets (regions) that even if one of the subnets is not accessible then other subnets can still uniquely identify the nodes in terms of distances from the remaining subnets. The current research trends in metric-related graph parameters and characteristics of these oxide networks motivated us to study region-based fault tolerant addressing schemes for interconnection networks. The most recent research on graph theoretic properties of different types of oxide, silicate and triangulene interconnection networks can be seen in [
12,
13,
14,
15]. In our current study, we have calculated the partition and fault tolerant partition dimensions of
,
and
interconnection networks.
A subcollection of nodes with minimal size is called a metric basis if each node of the network is at a specific distance from nodes in the subcollection. The uniqueness of metric basis motivated, Slater [
16] and Melter et al. [
17] independently introduced this notion in 1975 and 1976. These sets were used for navigation of robots [
18] and location of functional groups in chemistry [
19] in 1996 and 2000, respectively. In 2015, Garey et al. [
20] established the fact that the computation of metric basis in a network is NP-hard. A resolving partition of a network is the partition of the node set having each node of the network uniquely classified in terms of the distances from sets in the partition. The minimal cardinality of sets in a resolving partition is termed the partition dimension of the network. This notion was presented in 2000 by Chartrand et al. [
21].
Consider a connected network
W with node set,
and edge set,
. We denote the distance of a node
p from the node
q of the network by
. The distance of a node
q from a subset
A of
is
and neighborhood of a node
p in
W is
. Let
be an ordered set of nodes, the vector
is called the representation of the node
q with respect to
. The set
is called the resolving set of
W if each node of the network has a distinct representation. The metric dimension of a network is the minimal size of a resolving set, symbolized as
. Let
be an ordered
partition of the node set of the network
W and
be the partition representation of node
q with respect to
. An
partition is called a resolving partition of a network if the representations of all nodes of the network are distinct whereas its least cardinality is called the partition dimension of
W, symbolized as
. Chartrand et al. [
21] gave the basic rules interrelating these two parameters of the networks which are given in the following proposition.
Proposition 1 ([
21]).
If W is a connected network of order , then;
W is path if and only if ;
W is the complete network if and only if .
In 2015, the partition dimension of certain classes of tree graphs was studied in [
22]. In 2017, the partition dimension of lollipop and Jahangir graphs was computed in [
23]. In 2019, the partition dimensions were studied for fullerene and cycle books graphs in [
24] and [
25], respectively. Additional research on partition dimension can be found in [
26,
27].
In 2020, the notion of
partition dimension of networks was presented by Moreno et al. [
28]. If the representation vectors are distinct for each node of the network, in at least
places, then the partition is known as the
partition generator of the network. A generator of minimal size is known as
partition basis. The minimal size of
is known as the
partition dimension of the network, represented by
. The
is termed as a fault tolerant partition dimension. The
was computed for some important graphs in [
6,
29,
30,
31,
32,
33,
34]. Furthermore, the following lemma characterizes the graphs with a fault tolerant partition dimension bounded below by 4 and will be used in computing the fault tolerant partition dimension of
,
and
interconnection networks in the forthcoming subsections.
Lemma 2 ([
33]).
Let W be a graph of order . If W has a node of degree at least 4, then . Main Results
The research conducted in this manuscript leads to the following results.
For and ,
and .
For ,
and .
For ,
and .
In
Section 2, the partition dimension of
,
and
interconnection networks is computed. The fault tolerant partition dimension of these interconnection networks is computed in
Section 3. In
Section 4, we apply the partition and fault tolerant basis to create a novel addressing scheme for region-based routing in the networks. The results are summarized and an open problem is proposed in
Section 5.
4. Application
The applications of partition and fault tolerant addressing schemes have recently been discussed for routing optimization in [
29], supply chain optimization in [
31] and sensors deployment in [
6]. In this section, an application of the fault tolerant addressing scheme in the context of optimal data flow is included.
The performance of a network depends on the network topology and routing algorithm. Unique addressing schemes have been proposed in the literature for the interconnection networks for efficient routing. One way to assign unique addresses to nodes of the networks is to distribute nodes into different regions or subnets which allows efficient implementation of routing algorithms and also reduces the size of routing tables (see [
4,
5]). In this context, resolving partitions divide nodes into subnets and assign unique addresses to all the nodes in terms of distances from the subnets. A fault tolerant partition generator works in the presence of fault as if one of the subnets is not accessible then other subnets can still uniquely identify the nodes in terms of distances from the remaining subnets. We can choose suitable sizes for both resolving partitions and fault tolerant partition generators according to our needs but attaining the minimal size is NP-hard for general networks see [
33].
An interconnection network is equipped with many devices such as laptops, printers, switches, repeaters, routers, bridges and hubs. The aim is to overcome memory requirements, delays in data transfer and power consumption which increase with the increasing size of networks. As an explanatory case, consider an interconnection network in the form of
in
Figure 6, the devices are nodes and connections among these are edges of the network. We can split
Figure 6 into 3 and 4 four regions according to Theorem 2.3 and Theorem 3.4, respectively. The 3 and 4 - dimensional unique addresses can be constructed from the tables given in these theorems.
Fault tolerant basis for : , ,
4—dimensional fault tolerant addresses for :
, , , ,
It is evident from the addresses of nodes that the whole network is split into four regions and addresses are unique in at least two places which can reduce the routing tables considerably and can improve the efficiency of routing and broadcasting algorithms. Suppose in
Figure 6 if we want to pass the message from
to
which are in the same region then we can follow the path
but if the connection between
and
is faulty then
can pass the message to
which is in the other region and can pass message to
.