analysis and design of a half hypercube interconnection network topology
TRANSCRIPT
1
Analysis and Design of a Half Hypercube Interconnection Network
Shahid Rajaee Teacher Training UniversityFaculty of Computer Engineering
PRESENTED BY:
Amir Masoud Sefidian
2
Todayβs Lecture Content
β’ Introduction
β’ Definition and Properties of the Half Hypercube
β’ Connectivity of Half Hypercube
β’ Routing Algorithm and Diameter of Half Hypercube
β’ Conclusion
3
Todayβs Lecture Content
β’ Introductionβ’ Definition and Properties of the Half Hypercube
β’ Connectivity of Half Hypercube
β’ Routing Algorithm and Diameter of Half Hypercube
β’ Conclusion
4
Introductionβ’ Need to high-performance parallel processing is increasing because modern
application require real-time processing.β’ Parallel processing system can connect thousands of processors with via an
interconnection network.β’ The most common parameters for evaluating the performance of interconnection
networks: β’ Connectivityβ’ Diameter (relevant to message transmission time) β’ Degree (relevant to hardware cost)
β’ Typical parameter to evaluate interconnection network performance:β’ Network cost = degree * diameter
β’ A parallel computer design increase hardware costs of an interconnection network, because of the increased number of processor.
β’ Lower degree can reduce hardware costs but Delay increase and throughput are degraded.
5
Introductionβ’ Hypercube is a typical interconnection network topology:
β’ Is node- and edge-symmetric with a simple routing algorithm.β’ Simple recursive structure. β’ Easily embedded in various types of existing interconnection networks (flexibility).
β’ Hypercube drawback:β’ Increasing network costs associated with the increased degree when the number of nodes increases.
β’ Proposed variations of the hypercube to improve this shortcoming :β’ Multiple Reduced Hypercube β’ Twisted cube β’ Folded hypercube β’ Connected hypercube network β’ Extended hypercube
β’ New variation of the hypercube proposed in 2013 by Jong-Seok Kim, Mi-Hye Kim and Hyeong-Ok Lee:
β’ n-dimensional Half Hypercube π―π―π
β’ reduce the degree by approximately halfβ’ has the same number of nodes as a hypercube.
6
Todayβs Lecture Content
β’ Introduction
β’ Definition and Properties of the Half Hypercubeβ’ Connectivity of Half Hypercube
β’ Routing Algorithm and Diameter of Half Hypercube
β’ Conclusion
7
Definition and Properties of the Half Hypercubeβ’ Denote an n-dimensional half hypercube as π»π»π and represent its node with n binary bits:β’ Node S in π»π»π denoted: π ππ πβ1π πβ2β¦π π β¦π 3π 2π 1 (π β₯ 3).β’ There are two types of edge:
β’ h-edge : connects node π(= π ππ πβ1π πβ2β¦π π β¦π 3π 2π 1) to a node that has the complement in exactly one h position of the bit string of node S(1 β€ β β€ π/2 )
β’ sw(swap)-edge: connects node π(= π ππ πβ1π πβ2β¦π βπ ββ1β¦π 3π 2π 1) to a node in which the π/2 leftmost bits of the bit string of S and the π/2 bits on the right-side of S starting from the π/2 bit are swapped(h= π/2 ).
8
sw-edge Examplewhen n is even:
Connect to
(π ππ πβ1π πβ2β¦ π π/2 +1) and (π π/2 π π/2 β1π π/2 β2β¦ π 3π 2π 1) of S are exchanged.
When n is odd:
Connect to
(π ππ πβ1π πβ2β¦ π π/2 +1) and (π π/2 π π/2 β1π π/2 β2β¦ π 3π 2) of S are exchanged.
π(= π ππ πβ1π πβ2β¦ π π/2 +1 π π/2 π π/2 β1π π/2 β2β¦ π 3π 2π 1)
π π/2 π π/2 β1π π/2 β2β¦ π 3π 2π 1 π ππ πβ1π πβ2β¦ π π/2 +1
π(= π ππ πβ1π πβ2β¦ π π/2 +1 π π/2 π π/2 β1π π/2 β2β¦ π 3π 2π 1)
π π/2 π π/2 β1π π/2 β2β¦ π 3π 2 π ππ πβ1π πβ2β¦ π π/2 +1 π 1
if two parts of βπ/2β bits to exchange in the bit string of node S are the same, the node S connects to oneβs complement of the binary number of node S.
9
Example:
10
Example:
11
Example (Wrong Figure!!):
5-dimensional half hypercube (HH5)
12
Degree of π»π»π:π
2+ 1
number of h-edges (1 β€ β β€π
2)
sw-edge
13
Lemma: An π»π»π graph is expanded with recursive structures.
Proof An π»π»π graph is constructed with the nodes of two (n-1)-dimensional π»π»πβ1
graphs by adding one sw-edge or h-edge.The address of each node in an π»π»π graph is represented as j bit strings with binary
numbers {0, 1}(binary number of length j).
π»π»π0 j +1 position of the address of a node 0
π»π»π1 j +1 position of the address of a node is 1
Case 1 When expanded from an odd-dimension (j) to an even-dimension (k=j+1)
Construct an π»π»π graph by connecting a node of π»π»π0 and a node of π»π»π
1 in π»π»π.
The sw-edges in π»π»π will be replaced with new sw-edges in the expanded π»π»π graph.
A node of π»π»π0 the address bit of which is binary 1 at the n/2 position, will be
connected to a node of π»π»π1 through an sw-edge in π»π»π. When the bit is binary 0,
the node will be connected to a node of π»π»π0.
14
Case 2: When expanded from an even-dimension (k) to an add-dimension (l=k+1):
Construct an π»π»π graph by connecting a node of π»π»π0 and a node of π»π»π
1 in π»π»π.The sw-edges in π»π»π will be replaced with new sw-edges in the expanded π»π»π
graph.
A node of π»π»π0 the address bit of which is binary 1 at the π/2 position, will be
connected to a node of π»π»π1 through an sw-edge in π»π»π. When the bit is binary 0,
the node will be connected to a node of π»π»π0.
15
Lemma : There exist 2 π/2 π/2 -dimensional hypercube structures in an π»π»π.
Proof: node π(= π ππ πβ1π πβ2β¦π β β¦π 3π 2π 1)
β’ The h-edge of π»π»π is equal to the edge of a hypercube.β’ The structure of a partial graph constructed from π»π»π via the h-edge is the same as
the structure of a π/2 -dimensional hypercube.β’ Assume that a partial graph that has the same structure as an π/2 -dimensional
hypercube is π»π»ππ/2
in π»π»π and refer it as a cluster. β’ The address of each node in a cluster is π β β¦π 3π 2π 1.
β’ π»π»π consist 2 π/2 of clusters: β’ because the number of bit strings that can be configured by the bit string
π ππ πβ1π πβ2β¦π β is 2 π/2 .
16
Todayβs Lecture Content
β’ Introduction
β’ Definition and Properties of the Half Hypercube
β’Connectivity of Half Hypercubeβ’ Routing Algorithm and Diameter of Half Hypercube
β’ Conclusion
17
Connectivity of π»π»π
β’ Node (edge) connectivity is the minimum number of nodes (edges) that must be removed to disconnect an interconnection network to two or more parts without duplicate nodes.
β’ if degree = node connectivity said that the interconnection network has maximum fault-tolerance.
β’ Characteristic notation of interconnection network G:
β’ Node connectivity: π(πΊ)
β’ Edge connectivity: π(πΊ)
β’ Degree of interconnection network: d(πΊ)
It has been proven that π(πΊ) β€ π πΊ β€ d(πΊ)
18
Theorem 1 The connectivity of π»π»π
β’ π(π»π»π) = π π»π»π = d(π»π»π)β’ Proof:β’ Letβs prove that π»π»π remains connected even when n nodes are deleted from π»π»π.β’ Characteristic notation of interconnection network G:
β’ Node connectivity: π(πΊ)
β’ Edge connectivity: π(πΊ)
β’ Degree of interconnection network: d(πΊ)
It has been proven that π(πΊ) β€ π πΊ β€ d(πΊ)
From Lemma 2: π»π»π is composed of clusters, all clusters in π»π»π are hypercubes and two arbitrary nodes are connected via an sw-edge.
Assuming that X is a partial graph of π»π»π where |X| = n, it will be proven that π π»π»π β₯π
2+ 1 by
demonstrating that π»π»π remains connected even after the removal of X:Dividing two cases in accordance with the location of X. We denote the π»π»π in which X is deleted as π»π»π β π and a node of HHn as S.
19
Case 1: When X is located in one cluster of π»π»π
β’ Degree of each node in a cluster is π/2 .
If π/2 nodes adjacent to an arbitrary node S of the cluster are the same as the nodes to be deleted from X, π»π»π is divided into two components:
β’ an interconnection network π»π»π β π. β’ a node S.
β’ all nodes in a cluster are linked to other clusters in π»π»π via sw-edges.
β’ 2 π/2 -1 clusters in which X is not located are also connected to other clusters via sw-edges.
β’ Therefore, π»π»π β π is always connected when X is included in only a cluster of π»π»π.
20
Case 1: When X is located in one cluster of π»π»π
21
Case 2: When X is located across two or more clusters of π»π»π
As the nodes of X to be deleted are included across two or more clusters of π»π»π ,the number of nodes to be deleted from a cluster is at most π/2 β 1 .
Even if π/2 β 1 nodes adjacent to an arbitrary node S of a cluster are deleted, thenodes of the cluster in which node S is included remain connected because thedegree of a node in a cluster is π/2 .Although the other node to be removed is located in a different cluster, and not in the cluster that includes node S, it is still clear that π»π»π remains connected.
I π π»π»π β₯ π β₯π
2+ 1 because π»π»π always remains connected after removing X from any clusters in
π»π»π and
(πΌπΌ) π(π»π»π) β€π
2+ 1 because the degree of π»π»π is
π
2+ 1 .
I , (πΌπΌ) β connectivity of π»π»π , π(π»π»π) =π
2+ 1 .
22
Todayβs Lecture Content
β’ Introduction
β’ Definition and Properties of the Half Hypercube
β’ Connectivity of Half Hypercube
β’ Routing Algorithm and Diameter of Half Hypercube
β’ Conclusion
23
Routing Algorithm and Diameter of π»π»π
β’ This section analyzes a simple routing algorithm and diameter of π»π»π.
β’ Assume :β’ Initial node is π = π ππ πβ1π πβ2β¦π π/2β¦π 3π 2π 1β’ Destination node is T = tπ tπβ1tπβ2β¦tπ/2β¦ t3t2t1
β’ Simple routing algorithm can be considered in two cases depending on whether n is an even or an odd number.
24
Case 1 When n is an even number:
Denote S with two π/2 bit strings A(= π ππ πβ1π πβ2β¦ π π/2) and B(= π π2β1β¦π 3π 2π 1) => S=AB
Denote T with two π/2 bit strings C(= tπtπβ1tπβ2β¦ tπ/2) and D(= tπ2β1β¦t3t2t1) => T=CD
S=(0010) => A=00,B=10T=(1101) => C=11,D=01
Corollary 1: When n is even, length of the shortest path is
2 β π/2 + 1 = π + 1by the above algorithm.
25
Case 2 When n is an odd number:
initial node is π = π ππ πβ1π πβ2β¦ π π/2 +1 π π/2 π π/2 β1β¦ π 3π 2π 1destination node is T = tπtπβ1tπβ2β¦π‘ π/2 +1 π‘ π/2 π‘ π/2 β1β¦ π‘3π‘2π‘1
Denote S with two π/2 bit strings: A(= π ππ πβ1π πβ2β¦ π π/2 +1) and B(π π/2 π π/2 β1β¦ π 3π 2)
=> S = ABπ 1Denote T with two π/2 bit strings: C(= π‘ππ‘πβ1π‘πβ2β¦ π‘ π/2 +1) and D(π‘ π/2 π‘ π/2 β1β¦ π‘3π‘2)
=> T = CDπ‘1
26
Simple routing algorithm when n is odd:
S=(00000) => A=00,B=00T=(11111) => C=11,D=11
Corollary 2: When n is odd:
Length of the shortest path is 2 βπ
2+ 2
by the above algorithm.
27
Routing Algorithm and Diameter of π»π»π
Through the proposed simple routing algorithms, we can see an upper bound forthe diameter of π»π»π, thus proving Theorem 2:
28
Conclusion:β’ A new variation of the hypercube (half hypercube interconnection network) that
β’ reduced the degree by approximately halfβ’ has the same number of nodes as a hypercube.
β’ Evaluate the effectiveness of the proposed half hypercube.β’ Analyzed its connectivity and diameter properties. β’ Analyzed a simple routing algorithm of π»π»π and presented an upper bound for the
diameter of π»π»π.β’ These results demonstrate that the proposed half hypercube is an appropriate
interconnection network for implementation in a large-scale system.β’ More works is available on:
β’ βBroadcasting and Embedding Algorithms for a Half Hypercube Interconnection Networkβ
by: Mi-Hye Kim, Jong-Seok Kim and Hyeong-Ok Lee
Reference : http://link.springer.com/chapter/10.1007/978-94-007-6738-6_65
QUESTION??...