Transcript 下載/瀏覽Download

Centrality-Based Network Coding Node
Selection Mechanism for Improving
Network Throughput
報告人：王姿穎
學號：MA4G0202
Outline
1. Introduction
2. Centrality-Based Network Coding Node
Selection Mechanism
3. Performance Evaluation
4. Conclusion
1.Introduction
What is Network coding?
2. Centrality-Based Network Coding Node Selection Mechanism
Figure 1. Example of an area
2. Centrality-Based Network Coding Node Selection Mechanism
𝑖 = focal node
𝑗 = neighbor nodes
𝑥𝑗𝑖 = If there is traffic connection between node 𝑖 and
node 𝑗 , 𝑥𝑗𝑖 is 1 or not 0.
𝑁 = the number of network nodes
2. Centrality-Based Network Coding Node Selection Mechanism
(2)
𝑖 = focal node
𝑗 = neighbor nodes
𝑤𝑖𝑗 = the value of link bandwidth between node 𝑖 and
node 𝑗 or not 0.
𝑁 = the number of network nodes
2. Centrality-Based Network Coding Node Selection Mechanism
2. Centrality-Based Network Coding Node Selection Mechanism
If this parameter is between 0 and 0.5, output centrality is more
powerful than input centrality. This means if this node turns into NC
node, the volume of output traffic from node 𝑖 to neighbors is larger
and this increase the packet transmission rate.
Whereas if it is between 0.5 and 1, input centrality is more powerful
than output centrality. This means if this node turns into NC node, the
packet
innovativeness of node
𝑖
decoding probability at the receiver.
is higher and this increase the
2. Centrality-Based Network Coding Node Selection Mechanism
(i) Compute centrality: to begin with, CNCNS initializes the number
of NC node in each area (𝑁𝑐 ) (=1) which can be decided by
network administrator. Based on computed degree and weight
using equation (1)-(2), focal node computes the centrality itself
using equation (3)-(4). And connected neighbor nodes to focal
node compute its centrality.
(2)
2. Centrality-Based Network Coding Node Selection Mechanism
(ii) Compare the centrality of focal node and its neighbor nodes:
next, the focal node is one of all network node at a time and
neighbor nodes are connected to focal node in a one hop area. To
compare the centralities of focal node and connected neighbor
nodes, focal node collects the centrality of neighbor nodes. And
then, as much as NC, focal node decides the NC node(s) which
has the maximum value in the area. If selected NC node is the one
of the neighbor nodes, focal node makes it known. Otherwise, focal
node operates NC node.
3. Performance Evaluation
This section evaluates performance of the CNCNS against general
network coding node selection mechanisms such as the all network
nodes operate network coding based on RLNC and the network coding
nodes are randomly selected by using our computer simulator.
The results were obtained with the following assumptions:
single source connect with multiple receivers; and the packet arrival
process follows the Poisson process with rate λ. We use the random
network which consists of 50 nodes with network diameter of 6.
To
put
weight
between
packet
transmission
rate
and
packet
innovativeness, the indicator parameter β is properly chosen within
above defined bounds.
For the precise computation, redundant packets are deleted by receiver,
thus the network throughput is analyzed by decoded packets.
3. Performance Evaluation
隨機線性網絡編碼
隨機線性網絡編碼可以取得更好的組播傳輸速率，較為實用。在實際網絡中，
節點會將來自連入線路的封包緩存起來，當節點需要發送封包時再將緩存的
封包執行網絡編碼，然後發出。
例如節點A有2個上游節點X,Y，X向A發送了封包 ( 2,2,1 , 2𝑥1 + 2𝑥2 +
𝑥3 )(2𝑥1 + 2𝑥2 + 𝑥3 是數據體，(2,2,1)是對數據體執行線性編碼時所用的係
數）,Y向A發送了封包( 1,5,4 , 𝑥1 + 5𝑥2 + 4𝑥3 )。
當A需要發送數據時，便把緩存的這兩個封包取出來，隨機選擇2個係
數（如2和1），獲得新的數據體(2𝑥1 + 2𝑥2 + 𝑥3 )*2+(𝑥1 + 5𝑥2 + 4𝑥3 )*1=
5𝑥1 + 9𝑥2 + 6𝑥3 和新的合成係數(2,2,1)*2+ (1,5,4)*1=(5,9,6)所以A就把
合成後的數據體5𝑥1 + 9𝑥2 + 6𝑥3 連同合成係數(5,9,6)，向下游節點發送
出去。
3. Performance Evaluation
3. Performance Evaluation
Figure 3. Normalized average end-to-end delay versus number of NC nodes
4. Conclusion
 CNCNS improves the network throughput with the minimizing the number of
coding nodes.
 Since CNCNS operates under distributed manner, it is simple while not requiring
centralized information of network.
 Through the control indicator, CNCNS is useful for enhancing a performance of
the network throughput and the network reliability.
 improve the network throughput and end-to-end delay.
Reference
http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=6779083