LSP Preemption Policies for Diff-Serv-aware MPLS Traffic Engineering draft-deoliveira-diff-te-preemption-01.txt - PowerPoint PPT Presentation

Loading...

PPT – LSP Preemption Policies for Diff-Serv-aware MPLS Traffic Engineering draft-deoliveira-diff-te-preemption-01.txt PowerPoint presentation | free to download - id: da114-ZDc1Z



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

LSP Preemption Policies for Diff-Serv-aware MPLS Traffic Engineering draft-deoliveira-diff-te-preemption-01.txt

Description:

RAND: randomly selects LSPs for preemption ... (b(l) r)2 penalizes a choice of an LSP that would result in high bandwidth wastage ... – PowerPoint PPT presentation

Number of Views:11
Avg rating:3.0/5.0
Slides: 10
Provided by: valued84
Learn more at: http://www.ietf.org
Category:

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: LSP Preemption Policies for Diff-Serv-aware MPLS Traffic Engineering draft-deoliveira-diff-te-preemption-01.txt


1
LSP Preemption Policies for Diff-Serv-aware MPLS
Traffic Engineering draft-deoliveira-diff-te-pre
emption-01.txt
  • Jaudelice C. de Oliveira

Broadband and Wireless Networking
Laboratory School of Electrical and Computer
Engineering Georgia Institute of Technology
2
Motivation
  • Study different preemption policies benefits and
    drawbacks
  • Cascading effect
  • Number of preempted LSPs/rerouting
  • Wasted bandwidth
  • Priority of preempted LSPs
  • When rerouting, how much longer (hops) than the
    original path is the new route

3
Preemption Algorithms
  • RAND randomly selects LSPs for preemption
  • P (Priority) selects LSPs purely based on lower
    priority
  • PN (Priority and Number) sorts LSPs in
    increasing priority selects larger LSPs in order
    to preempt a smaller number
  • Preemption Heuristic combines the three
    criteria priority, number, and bandwidth
  • Versatile preemption policy
  • ? ? Priority of preempted LSPs
  • ? ? Number of preempted LSPs
  • ? ? Amount of bandwidth preempted

4
Preemption Heuristic
  • Heuristic cost function
  • H(l) ? (cost of LSP l) ? ? (b(l) r)2
  • ? (cost of LSP l) represents the cost of
    preempting an LSP with a certain priority
  • ? represents the choice of a minimum number of
    LSPs for preemption
  • ? (b(l) r)2 penalizes a choice of an LSP that
    would result in high bandwidth wastage
  • Calculate H(l) for each LSP l
  • Sort H(l) in increasing order and by increasing
    bandwidth
  • Select the LSPs with smaller H that add enough
    bandwidth to accommodate r

5
Example Static Case
  • Preemption in a single link with 16 LSPs
  • New LSP setup request with r155, p0
  • alpha1, beta0, gamma1 LSPs l8, l12 , and l16
  • alpha1, beta1, gamma0 LSPs l7, l12
  • alpha1, beta1, gamma1 LSPs l12,and l15

6
Example Dynamic Case
  • Each link 100 Mbps
  • Random source/ destination request
  • Request bw 2, 4, 6, 8, 10 Mbps
  • 50 priority 7
  • 20 priority 6
  • 6 each, priorities 1-5
  • Interarrival average 2s
  • Holding time average 500s (exponentially
    distributed)

7867 LSP setup requests were created
7
Example Dynamic Case (contd)
Heuristic Heuristic Other Policies Other Policies Other Policies No Preemption
PB BN RAND P PN No Preemption
Accepted Rejected 87.2 12.8 87.2 12.8 87.8 12.2 87.8 12.2 88.4 11.6 82.9 17.1
Preempted Rerouted Destroyed 12 95.4 4.6 15 96.3 3.7 19.5 96.9 3.1 14.2 94.9 5.1 11.2 94 6
Max Cascading Level 1 2 5 1 1
LSPs preempted Worst case 1.1 4 1.8 3 1.3 4 1.4 5 1.2 5
Wasted band (Mbps) Worst case 0.82 8 0.76 8 2.89 8 2.90 8 3.6 8
Priority Worst case 6.90 5 6.61 2 6.48 2 6.90 5 6.92 5
Extra hops (reroute) Worst case 0.8 4 0.6 5 0.5 4 0.7 4 0.8 5
8
Example Dynamic Case (contd)
Heuristic Other Policies Other Policies
PB P PN
Accepted Rejected 87.2 12.8 87.8 12.2 88.4 11.6
Preempted Rerouted Destroyed 12 95.4 4.6 14.2 94.9 5.1 11.2 94 6
Max Cascading Level 1 1 1
LSPs preempted Worst case 1.1 4 1.4 5 1.2 5
Wasted band (Mbps) Worst case 0.82 8 2.90 8 3.6 8
Priority Worst case 6.90 5 6.90 5 6.92 5
9
Conclusions and Future Work
  • When preemption is based on priority, cascading
    is not critical
  • When bandwidth is taken into account, the wasted
    bandwidth is low, and more LSPs are successfully
    rerouted
  • The heuristic combines the above features
  • Future work investigate further benefits when
    using the Russian Dolls model
About PowerShow.com