Introduction to Parallel

Computing, Fall 2009

Sinan Kockara
Department of Computer Science
UCA
Network Topologies: Completely Connected and
Star Connected Networks

Example of an 8-node completely connected network.

(a) A completely-connected network of eight nodes; Completely

connected network is static counterpart of crossbar
switching network?
(b) a star connected network of nine nodes.
 Network Topologies:
Star Connected Network

• Every node is connected only to a

common node at the center.
• Distance between any pair of nodes is
O(1). However, the central node becomes
a bottleneck.
• In this sense, star connected networks are
static counterparts of buses.
 Network Topologies:
Linear Arrays, Meshes, and k-d Meshes

• In a linear array, each node has two Linear arrays: (a) with no
neighbors, one to its left and one to its
wraparound links; (b) with
right. If the nodes at either end are
connected, we refer to it as a 1-D torus or wraparound link.
a ring.
• A generalization to 2 dimensions has
nodes with 4 neighbors, to the north,
south, east, and west.
• A further generalization to d dimensions
has nodes with 2d neighbors.
• A special case of a d-dimensional mesh is
a hypercube. Here, d = log p, where p is
the total number of nodes.
• K-d Meshes: one extreme of k-d mesh
family is Linear Arrays, other extreme is
hypercubes
 Network Topologies:
Two- and Three Dimensional Meshes

Two and three dimensional meshes: (a) 2-D mesh with no wraparound;

•2D mesh can be laid out in 2D space , make it attractive from a wiring
standpoint
•Regularly structured computations map very naturally to a 2D mesh

(b) 2-D mesh with wraparound link (2-D torus); and (c) a 3-D mesh with no
wraparound.
 3D mesh
• Each node element except nodes on the
periphery is connected to 6 other nodes
• Physical simulations such as 3D Finite
Element Modeling, 3D weather modeling
naturally fits to 3D mesh topology.
• Therefore, it is commonly used
interconnection network.
 Network Topologies:
Hypercubes and their Construction

Construction of hypercubes from hypercubes of lower

dimension.
 Network Topologies:
Properties of Hypercubes

• Has two nodes along each dimension (e.g.

Number of nodes: 2d)
• The distance between any two nodes is at
most log2 p
• Each node has d=log2 p neighbors.
 Two-dimensional array (mesh)
Computer/
Links processor

Also three-dimensional - used in some large high performance systems.

1b.9
 Three-dimensional hypercube
110 111

100 101

010 011

000 001

The distance between two nodes is given by the

number of bit positions at which the two nodes
differ.
 Four-dimensional hypercube

0110 0111 1110 1111

0100 0101 1100 1101

0010 0011 1010 1011

0000 0001 1000 1001

One of the 3D hypercube has 0 as a prefix

other has 1 as a prefix
5D Hypercube
 Hypercube routing
To go from node x1x2x3 to node y1y2y3
• Each x and y representing bits
– Go to y1x2x3 if y1 and x1 different
– Then go to y1y2x3 if y2 and x2 different
– Then go to y1y2y3 if y3 and x3 different

Example
101 to 010
101 -> 001 ->011-> 010
 Hypercube Routing 2
• This routing scheme is called e-cube
routing algorithm, or left-to-right routing

• Example: What is the route taken from

node # 13 to node # 42?
• 13 (001101)->42(101010)
 Hypercube Routing Example
• Answer:
• 13 (001101)->42(101010)
• 13(001101)->45(101101)->41(101001)->
43(101011)->42(101010)
 Question
• What is the maximum number of steps to
route a message in an n-dimensional
hypercube?

Answer
?
Why?
 Network Criteria: Evaluating
Interconnection networks
• For network evaluation 3 things are
important:
– Communication Latency: time to make
message transfer through network
– Bandwidth: data transmitted per second
(bits/second)
– Cost: number of links in network
 How to Evaluate a network?
• Key issues for evaluation:
– Diameter (Communication Latency measure)
• Minimum number of links between the two farthest nodes (shortest
route is considered) e.g What is diameter for ring network?
• Determines worst case delays
– Bisection width (Connectivity measure)
• Number of links that must be cut to divide the network into two equal
parts
• E.g. think about linear arrays and torus, what are their bisection
width?
– Degree (Cost measure)
• Maximum number of links that a single node has
– Low diameter, low degree, and higher bisection width
are preferred features for a network