network-coding multicast networks with qos guarantees
DESCRIPTION
Network-Coding Multicast Networks With QoS Guarantees. Abdullah Şahin Hasan Saygın Arkan 10.01.2010. Outline. What we are going to present …. Define The Problem …. Solve for Unicast. Convert to Multicast. Introduction. Introduction. - PowerPoint PPT PresentationTRANSCRIPT
NETWORK-CODING MULTICAST NETWORKS WITH QOS
GUARANTEES
Abdullah ŞahinHasan Saygın Arkan
10.01.2010
• Introduction1• Background2• Unicast vs. Multicast3• Numerical Results4
• Conclusion5
Outline
What we are going to present …
Define The Problem …
Solve for Unicast
Convert to Multicast
INTRODUCTION
Introduction• “Network-Coding Multicast Networks
With QoS Guarantees”–Xuan, Y.: Lea, C.-T.– IEEE/ACM Transactions on Networking–30 August 2010
• Related Work• Terms–QoS, Network Coding, unicast, multicast…
UNICAST & MULTICASTCONGESSION
Problem Definition• Admission Control – How?• New QoS Architecture – Non-Blocking
Network! – No admission control
• Low throughput for multicast– Impractical
• Data Transmission– Transmission in Client – Local Server TRIVIAL– Transmission in Backbone PROBLEM!
Problem Definition– Transmission in Backbone PROBLEM!
Internal Rooter
Edge Router
Edge Rouger
Edge Router
Egde Router
Unicast
Data Packet
Data Packet
Multicast
Internal Rooter
Edge Router
Edge Rouger
Edge Router
Egde Router
Data PacketData Packet
Data Packet
Data Packet
Unicast Solution• tij = traffic rate from i edge to j edge
• αi = ingress traffic & βi = egress traffic
• (αi, βi) = (Θ αi’ , Θ βi
’)• Task is maximizing Θ
Edge Router
αi = ingress trafficβi = egress traffic
Unicast Solution
• Σ tij < αi’
• Σ tij < βi’
• Not Applicable on Multicast– α = β for unicast, but not for multicast
Edge Router
Multicast SolutionG = multicast edge group
= { sg, D(g), tg }source, destination set, data rate
Binary Vectors:ϒg(i) = 1, if i = sg δg(j) = 1, if j € D(g)
0, otherwise 0, otherwise
Multicast Solution
• Σ ϒg(i) . tg < αi’ - ingress traffic
• Σ δg(j) . tg < βi’ - egress traffic
• tij = Σ(δg(j) . ϒg(i) . tg)
Optimal Routing
i
j
xije
Optimal Routing
Optimal Routing
Optimal Routing
• For IP networks – Calculation on weights
• MPLS-Type Explicit Routing Networks– Arbitrarily chosen nodes, and calculation of max loaded
link
NUMERICAL RESULTS
Numerical Results
• Constraint-Based Routing Approach• Non-Blocking Based Approach– 15 Nodes, 62 directed links, capacity of 300.
– 10 consecutive rejects = fully loaded
– Number of receivers per multicast flow is random (binomial distribution [2, N-1] , N is total edge
Numerical Results
• Nonblocking Multicast Networks
• b/a ratio, average fan-out = 3, 15 edge nodes
Numerical Results
• Nonblocking Multicast Networks
• b/a ratio, average fan-out = 4, 15 edge nodes
Numerical Results
• Nonblocking vs CBR
• 5 edge nodes, average fan-out = 3
Numerical Results
• Nonblocking vs CBR
• 15 edge nodes, average fan-out = 3
Numerical Results
• Nonblocking vs CBR
• 15 edge nodes, average fan-out = 4
Conclusion
• Better to have admission control at the edge, NOT inside it!
• Non-Blocking removes that need• Main Problem – low throughput• Optimal Paths in Unicast = Optimal Paths in
Multicast Nonblocking with Network Coding
QUESTIONS?