Lecture 05 - Chapter 3 - Models of parallel computers and interconnections

Parallel and Distributed Computing
Chapter 3: Models of Parallel Computers and
Interconnections
1
Muhammad Haroon
mr.harunahmad2014@gmail.com
Cell# +92300-7327761
Department of Computer Science
Hitec University
Taxila Cantt
Pakistan

3.1a: Architecture of Theoretical Parallel
Computer
2
 Parallel Random Access Machine (PRAM) is a
theoretical model of parallel computer, with
1.) p identical processors
2.) a global memory of unbounded size
3.) memory is uniformly accessible to all processors
 Processors share a common clock
 They may execute different instructions in each cycle
 There are four subclasses of PRAM, based on the
memory access protocols

3.1b: Illustration of the PRAM Model
3

3.1c: PRAM Subclasses
4
 Exclusive-read, exclusive-write (EREW) PRAM: Access to a
memory location is exclusive. No concurrent read or write
operations are allowed
The weakest PRAM model, affording minimum concurrency in
memory access
 Concurrent-read, exclusive-write (CREW) PRAM: Multiple read
accesses to a memory location is allowed. Multiple write
accesses to a memory location is serialized
 Exclusive-read, concurrent-write (ERCW) PRAM: Multiple write
accesses are allowed to a memory location. Multiple read
accesses are serialized
 Concurrent-read, concurrent-write (CRCW) PRAM: Both
multiple read and multiple write accesses to a memory location
are allowed
This is the most powerful PRAM model

3.1d: PRAM Semantics
5
 Concurrent read access to a memory location by all
processors is OK.
 Concurrent write access to a memory location
presents semantic discrepancy and requires
arbitration
 The most frequently used arbitration protocols are:
 Common: Concurrent write is allowed if all the writing
processors have the same value
 Arbitrary: An arbitrary processor is allowed to
proceed with the write operation, and the rest fail
 Priority: Processors are prioritized a priori, the
processor with the highest priority writes and others
fail
 Sum: The sum of all the quantities is written

3.2: Processor Granularity
6
 Coarse-grained: Few powerful processors
 Fine-grained: Many less powerful processors
 Medium-grained: between the above two
 The granularity definition is relative
 Another definition of granularity is with respect to the
relative rates of communication to computation
Fine-grained: shorter duration between
communication
Coarse-grained: longer duration between
communication

3.3a: Interconnection Networks
7
 Static networks: Processing nodes are
connected by point-to-point communication
links (direct networks)
Mostly used for message-passing computers
 Dynamic networks: Communication links
are connected to one another dynamically by
the switches to establish paths among
processing nodes and memory banks
(indirect networks)
 Mostly used for shared memory computers

3.3c: Examples of Static Interconnections
Full-y connected
8
2D mesh with wraparound

3.3d: Switch Board Connections
9

3.3e: Dynamic Interconnection
10

3.3f: Multistage Dynamic Interconnection
11

3.3g: Switch Functionalities
12
 A single switch has a set of input ports and a set of
output ports
 The switch functionalities include:
1) a mapping from input to output ports (basic)
2) additional support for
internal buffering (when the requested output port
is busy)
routing (to alleviate network congestion), and
multicasting (same output on multiple ports)
 The degree of a switch is the total number of ports

3.3h: Cost of a Switch
13
 The cost of a switch is influenced by the cost
of mapping hardware, the peripheral
hardware, and packaging costs
 The mapping hardware typically grows as the
square of the degree of the switch
 The peripheral hardware grows linearly as
the degree
 The packaging costs grow linearly as the
number of pins

3.3h: Network Interface (Static)
14
 Network interface is to handle the connectivity
between the node and the network
 Network interface has input and output ports that
pipe data into and out of the network
 Its functionalities include:
1) packetizing data
2) computing routing information
3) buffering incoming and outgoing data
4) error checking

3.3i: Approximation of Network Costs
15
 For dynamic interconnection networks: Its
cost is proportional to the number of switches
used in the network
 For static Interconnection networks: Its cost
is proportional to the number of links

3.4a: Network Topologies
16
 Multiple processors must be working together to
solve a common task
 They must communicate during the course of
solving the task
 The communication is provided by the
interconnection networks
 How to connect multiple processors in a parallel
system --
This is a trade-off between cost and scalability with
performance

3.4b: Bus-Based Networks
17
 A bus-based network consists of a shared medium
that connects all nodes
 The cost of a bus-based network scales linearly with
respect to the number of processors p
 The distance between any two nodes in the network
is constant O(1)
 Ideal for broadcasting information
 Disadvantage: bounded bandwidth & blocking
Performance is not scalable with respect to the
number of processors p

3.6b: Bus-Based Interconnect with Cache
18

3.6c: Crossbar Network
19
 A crossbar network uses a grid of switches or
switching nodes to connect p processors to b
memory banks
 It is a non-blocking network
 The total number of switching nodes is Θ(pb)
 In many cases, b is at least on the order of p, the
complexity of the crossbar network is Ω(p*p)
 Disadvantage: Switch complexity is difficult to realize
at high data rates
 Scalable in terms of performance, but not scalable
in terms of cost

3.6e: Crossbar Network (II)
21

3.6f: Multistage Networks
22
 To balance the scalability between performance and
costs
 Allowing multiple stages between processors and
memory banks
 Switches are installed at each stage
 It is more scalable than bus-based networks in
terms of performance
 It is more scalable than the crossbar networks in
terms of costs
 A special multistage interconnection network is the
omega network

3.7: Completely-Connected Network
 Each node has a direct
communication link to
every other node in the
network (non-blocking)
 How many
communication links
are needed?
 Scalable in terms of
performance, not
scalable in terms of
cost
23

3.8: Star-Connected Network
 One processor acts as the
central processor
 Every other processor has a
communication link with this
processor
 Congestion may happen at
the central processor
 This is a blocking network
 Scalable in terms of cost,
not scalable in terms of
performance
24

3.9: LinearArray and Ring Networks
Scalable in terms of costs, not scalable in terms
of performance
25

3.10: 2D Mesh Networks
2D mesh network
26
2D mesh network with
wraparound

3.11: 3D Mesh Network
Many physical simulations can be mapped naturally to a
3D network topology. 3D mesh interconnection is common
27

3.13a: Tree-Based Networks
28
 There is only one path between any pair of
nodes
 Linear array and star-connected networks are
special cases of tree networks
 Tree networks can be static or dynamic
 In case of dynamic interconnection, the
intermediate level processors are switching
nodes and the leaf nodes are processing
elements

3.13b: What is This Tree Network?
29

3.13c: Static and Dynamic Tree Networks
30

3.13d: Communication in Tree Networks
31
 Messages from one half tree to another half
tree are routed through the top level nodes
 Communication bottleneck forms at higher
levels of the trees
 The solution is to increase the number of
communication links and switching nodes at
the higher levels
 The fat tree is suitable for dynamic networks

3.14a: Evaluating Static Interconnection
Networks
32
 There are several criteria to characterize the
cost and performance of static
interconnection networks
 Diameter
 Connectivity
 Bisection Width
 Bisection Bandwidth
 Cost

3.14b: Diameter of a Network
 The diameter of a
network is the the
maximum distance
between any two
processing nodes in the
network
 The distance between
two processing nodes
is defined as the
shortest path between
them
33

3.14c: Diameters of Mesh Networks
34

3.15a: Connectivity of Networks
35
 The connectivity of a network is a measure
of the multiplicity of paths between any two
processing nodes
 The arc connectivity is the minimum number
of arcs that must be removed from the
network to break it into two disconnected
networks
 A network with high connectivity has lower
contentions for communication resources

3.15b: Connectivity of Mesh Array
36

3.16a: Bisection Width & Channel Width
37
 The bisection width is the minimum number
of communication links that must be removed
to partition the network into two equal halves
 The channel width is the number of bits that
can be communicated simultaneously over a
link connecting two nodes
 Channel width is equal to the number of
physical wires in each communication link

3.16b: Channel Rate &Channel
Bandwidth
38
 The peak rate a single physical wire can deliver bits
is called channel rate
 The channel bandwidth is the peak rate at which
data can be communicated between the ends of a
communication link
 Channel bandwidth is the product of channel rate
and channel width
 The bisection bandwidth is the minimum volume of
communication allowed between any two halves of
the network
 It is the product of bisection width and channel
bandwidth

3.16c: Characteristics of Static Networks
Network Diameter Bisection
Width
Arc
Connect.
Number
of Links
Fully conn-ted 1 p2
/4 p-1 p(p-1)/2
Star 2 1 1 p-1
Binary tree 2log((p 1)/ 2) 1 1 p-1
Linear array p-1 1 1 p-1
Ring |p-2| 2 2 p
2D mesh 2( p 1) p 2 2( p p)
2D meshwrap
2

p / 2
 2 p 4 2p
Hypercube log p p/2 log p (p log p) / 2
39

3.17: Cost of Static Interconnection
Networks
40
 The cost of a static network can be defined in
proportion to the number of communication links or
the number of wires required by the network
 Another measure is the bisection bandwidth of a
network
a lower bound on the area in a 2D packaging or the
volume in a 3D packaging
Definition is in terms of the order of magnitudes
 Completely connected and hypercube networks are
more expensive than others

3.18: Evaluating Dynamic Interconnection
Networks
41
 Need consider both processing nodes and switching
units
 Criteria similar to those used with the static
interconnection networks can be defined
 The diameter is defined as the maximum distance
between any two nodes in the network
 The connectivity is the maximum number of nodes
(or edges) that must fail to break the network
 The cost of a dynamic network is determined by the
number of switching nodes in the network

Lecture 05 - Chapter 3 - Models of parallel computers and interconnections

Recommended

More Related Content

What's hot (20)

Similar to Lecture 05 - Chapter 3 - Models of parallel computers and interconnections (20)

More from National College of Business Administration & Economics ( NCBA&E) (20)

Recently uploaded (20)

Lecture 05 - Chapter 3 - Models of parallel computers and interconnections