1. draft-deoliveira-diff-te-preemption-02.txt J. C. de Oliveira, JP Vasseur, L. Chen, C. Scoglio Updates: –Co-author: JP Vasseur –New preemption criterion added Minimize blocking probability –Simulation results based on periodic link fault events Jaudelice C. de Oliveira Drexel University jau@ece.drexel.edu JP Vasseur Cisco Systems, Inc jpv@cisco.com

2. IETF’58draft-deoliveira-diff-te-preemption-02.txt2 Preemption Criteria Preempt the LSPs that have the least priority (preemption priority). –The QoS of high priority traffics would be better satisfied and the cascading effect would be limited. Preempt the least number of LSPs. –The number of LSPs that need to be rerouted would be lower. Preempt the least amount of bandwidth that still satisfies the request. –Resource utilization is improved. Preempt LSPs that minimize the blocking probability –Minimize risk that preempted TE LSP cannot be rerouted.

3. IETF’58draft-deoliveira-diff-te-preemption-02.txt3 Preemption Cascading Preemption Cascading effect : –The decision of preempting an LSP may cause other preemptions in the network. –Different cascading levels may be achieved by the preemption of a single LSP. Cascading levels: –When an LSP is preempted and rerouted without causing any further preemption, the cascading is said to be of level 0. –When a preempted LSP is rerouted and in order to be setup in the new route it also causes the preemption of other LSPs, the cascading is said to be of level 1, and so on. Preemption cascading is not desirable and therefore policies that minimize it are of interest. Typically, this can result in severe network instabilities.

4. IETF’58draft-deoliveira-diff-te-preemption-02.txt4 Preemption Policies Preemption Policies with preemption criteria considered in a given order of importance: –P: Sorts candidate LSPs by priority only. –PN: Sorts the LSPs by priority, and for cases in which the priority is the same, orders those by decreasing bandwidth selects larger LSPs for preemption in order to minimize number of preempted LSPs –PB: Sorts the LSPs by priority, and for LSPs with the same priority, sort those by crescent bandwidth selects smaller lower priority LSPs in order to reduce bandwidth wastage

5. IETF’58draft-deoliveira-diff-te-preemption-02.txt5 Preemption Policies Preemption Policies with preemption criteria considered (Heuristic) in a customized weighted balance: –HPB: The heuristic with priority and bandwidth wastage as the most important criteria. –HBlock: The heuristic considering the minimization of blocking probability and priority. –HNB: The heuristic with number of LSPs preempted and wasted bandwidth in consideration.

6. IETF’58draft-deoliveira-diff-te-preemption-02.txt6 Simulations Random topology (150 nodes): –Average connectivity: 3 –10% of the nodes have connectivity 6 –10% of the links are OC3, 70% are OC48, and 20% are OC192 –T: total number of TE LSPs in the network (T = 18,306 LSPs) Two classes of TE LSPs: Voice/AToM:Data(Internet/VPN): Number20% of T 80% of T 4%8% 20%40% Priority0, 1, and 234567 Size as uniform distribution between: 30M and 50M20M and 50M15M and 40M10M and 20M1M and 20M1K and 1M Priority and size distributions for Voice/AToM and Internet/VPN LSPs.

7. IETF’58draft-deoliveira-diff-te-preemption-02.txt7 Simulation results Simulation results for constant network failure: 1 random failure every hour. HeuristicOther algorithms HPBHBlockHNBPPNPB Need to be Rerouted532 Preempted612483619504477598 Rerouted4670.763410.734750.773470.693350.74360.73 Rejected1450.241300.271440.231570.311420.31620.27 Maximum Cascading4.5252.752 Wasted Bandwidth Average (Mbps)663863636479824789556832 Worst Case (Mbps)353212601036809285013140623449 Preempted Priority Average66.55.86.6 Worst Case1.53.81.23.8 Extra Hops in new Route Average0.230.250.220.25 0.23 Worst Case3.253 332.75

8. IETF’58draft-deoliveira-diff-te-preemption-02.txt8 Conclusions Cascading is not an issue when priority is taken into account by preemption policy Heuristic configured to minimize blocking probability (HBlock) –Small LSPs with low priority were selected for preemption on normally loaded links Several LSPs were selected for preemption, but the rate of LSPs that were not successfully rerouted was the lowest. Since the LSPs are “small,” it is easier to find a new route in the network –Fewer (larger) LSPs with low priority were selected on congested links fewer larger LSPs are selected improving load balance –The resource wastage was the overall lowest The heuristic is very flexible and can be configured according to the network provider’s best interest regarding the considered criteria