Hybrid Virtual Network

Embedding Algorithm with k-core

decomposition using Path-Splitting
OUTLINE
Network Virtualization
Virtual Network Embedding
Problem Definition
Network Model
Hybrid Virtual Network Embedding
Simulation Results
Conclusion
Network Virtualization
Network virtualization is an efficient approach of solving
the ossification problem.
Network virtualization allows multiple heterogeneous
virtual networks (VNs) to run simultaneously on a shared
substrate networks (SNs).
The environment of network virtualization consists of
infrastructure providers (IPs) and service providers (SPs).
Network Virtualization

Network virtualization model

Virtual network Embedding

Virtual network embedding (VNE) is the key area in

network virtualization.
Mapping of virtual network onto substrate network
known as virtual network embedding problem.
The main objective of VN embedding is to map the VNs
to the substrate network to make efficient use of the
substrate network resources, when the VN requests arrive
and depart over time.
Virtual network Embedding Example
Virtual network Embedding
Virtual network embedding is done by InPs.
VNE algorithm can be divided in two different stages:
node and link mapping.
Due to the combination of node and link constraints, the
VNE problem is NP-hard.
VNE algorithms can be divided into two categories: one-
stage VNE algorithm and two-stage VNE algorithm.
Problem Definition

“Developing and simulating Hybrid

Virtual Network Embedding
Algorithm with k-core
decomposition using path splitting.”
NETWORK MODEL AND PROBLEM DESCRIPTION

Substrate network- An undirected graph G S  ( N S , LS , C NS , C LS )

Virtual request- An undirected graph G V  ( N V , LV , C NV , C LV )
Embedding- M (GV ) : (GV , C NV , C LV )  (G S , C NS , C LS )

Node mapping- M N ( N V ) : ( N V , C NV )  ( N S , C NS )
Link mapping- M L ( LV ) : ( LV , C LV )  ( LS , C LS )
Substrate node stress S N (n S )   )
c ( nV

nV  n S

Substrate link stress S E (e S )   )

b ( eV

eV  E S
Residual or the available capacity of a RN ( n S )  c ( n )  S N ( n S )
N
S

substrate Node
Residual capacity of a substrate link RE (e S )  bN (e S )  S E (e S )
 NETWORK MODEL AND PROBLEM
DESCRIPTION
Available bandwidth capacity of a substrate path P  P S
is given by
RE ( P )  min e S P RE (e S )
Cost calculate by
C (G )    f eV
eS
  c ( nV
)
e E V e S E S
V
nV N V

V - allocated CPU capacity for the virtual node

n
eV
f eS
- allocated bandwidth capacity for the virtual link
e V on
eS
Revenue
R (G V )   b(eV )   c ( n V
)
e v E V n v N V
HYBRID VIRTUAL NETWORK
EMBEDDING

Improve the performance of VNE algorithm using k-core

decomposition.
Network Pruning and Classification
Hybrid Embedding Process
Core network mapping phase
Edge network mapping phase
 Path Splitting
Network Pruning and Classification
K-core decomposition is an important tool for
visualization of complex networks and consists in
identifying particular subsets of the graph.

With k-core decomposition algorithm we can disentangle

the hierarchical structure of VNR by recursively pruning
the nodes whose degree is equal to 1 and the
corresponding link until no further removal is possible or
the number of residual nodes in the VNR is 2.
Network Pruning Flow-chart
Network Pruning Example
Network Pruning Example
Network Pruning Example
Path Splitting
Path splitting enables better resource utilization by
allowing the substrate to accept more VN requests.
 It splits the available bandwidth of the path into small
bandwidth to satisfie the resource constraints.
Path-splitting Example
Hybrid Virtual Network Embedding
algorithm using Path splitting
Algorithm1 Hybrid VNE algorithm using Path Splitting

1. Sort the request according to their revenue in decreasing order.

2. If no request left, stop.
3. Select the request with largest revenue.
4. For each request
4.1 Sort the nodes of selected request according to their degree in decreasing
order.
4.2 Apply k-core decomposition algorithm and record k core nodes.
4.3 Map core network to substrate network using algorithm 2.
4.4 For each k core node do
4.4.1 Map edge network using one-stage VNE algorithm.
4.5 if edge network mapping failed then
Break.
4.6 End if.
Hybrid Virtual Network Embedding
algorithm using Path splitting
Algorithm 2 Node and Link Mapping
Node Mapping
Select virtual node with the largest degree and then
compute the amount of available bandwidth of node.
Select substrate node that satisfy the resource and
bandwidth requirement.
If more than one substrate node satisfies then select
substrate node with largest degree and mapped to that
substrate node.
Repeat step 1 to 3 until no node left.
 Link Mapping
6. Select the virtual link with largest bandwidth. Let denote c (u )
the bandwidth of virtual link l (v ) .
7. Assume mn( x(v))  i ( s ) and mn( y (v ))  j ( s ). Algorithm
constructs a graph between two nodes i (s ) and j (s ) that
satisfy bandwidth requirement.
i (s )
8. Algorithm finds shortest path between nodes and j (s )
c(u )  c(u ) / 2
9. If no substrate path exits then compute and repeat
step 6 to 9.
Al ( s )  Al ( s )  
10. Compute bandwidth .
11. Mapped virtual link to substrate link.
12. Repeat step 6 to 11 until no link left.
Node Mapping
Link Mapping
Hybrid Virtual Network Embedding
Example

Revenue of VN request 1 =377 Revenue of VN request 2 =238

Hybrid Virtual Network Embedding
Example

Substrate Network
Hybrid Virtual Network Embedding
Example
Simulation Result
Virtual and Substrate network generated using NS2
Simulator.
Performance Metrics
Revenue
Acceptance Ratio
Performance Metrics

Acceptance ratio
The acceptance ratio of an algorithm measures the
percentage of total VN requests accepted by that
algorithm in a given period.

Revenue
The revenue of an algorithm is the summation of virtual
nodes and virtual links mapped onto substrate network.
Simulation Parameters
Simulator NS2

Nodes/Links in Substrate Network 30/50, located at integer coordinates

with at least one substrate link
between nodes.
Capacity Real number between 0-50 assigned
randomly.
Arrival Rate 1 VN per 100 time units
By using Poisson process.
Lifetime of VN Requests Exponentially distributed with average
500 time units
Delay between successive VNRs 100 units of time

Probability of VN connectivity 0.5

Improvement in Revenue
Improvement in Revenue
Time Revenue of Revenue of
HVNE HVNE with
Path splitting

10.483257430 335 355

20.999385611 553 583
30.938629574 653 705.5
35.64296593 758 810.5
40.48396285 778 850.5
45.125500305 808 880.5
Improvement in Acceptance Ratio
Improvement in Acceptance Ratio
Time Acceptance Acceptance
Ratio of Ratio of HVNE
HVNE with Path
splitting

10.483257430 0.82 0.90

20.999385611 0.84 0.94
30.938629574 0.82 0.91
40.48396285 0.84 0.92
45.125500305 0.75 0.89
CONCLUSION

Network virtualization has been proposed as an efficient way

to overcome the current ossification of the Internet by allowing
heterogeneous VNs to coexist on a shared substrate network.
A major challenge in network virtualization is the VN mapping
problem that deals with efficient assignment of virtual nodes
and links onto the substrate network resources.
We proposed a hybrid VNE algorithm using path splitting to
increase the acceptance ratio of VNR and revenue.
Path splitting is also supported to enable better resource
utilization. With the help of path splitting substrate accept more
VN request.