Combined effects and simulation results

In this section, we present the simulation results of IARSVP, MRSVP and HMRSVP to illustrate the combined effects of dynamic FP locating and common route resource sharing. The simulation model adopts an 8× 8 wrapped-around mesh topology as shown in Fig. 5.6 to simulate a mobile computing environment with an unbounded number of regions. Without loss of generality, we configure a two-level hierarchical infrastructure with access routers (ARs) in the lower level and gateway routers (GRs) in the higher level. The ARs and GRs represent the MAs and GMAs, respectively, in HMRSVP. However, in order to demonstrate the effects of alternative routes on resource reservations, we adopt the multi-homing concept in our simulation.

AR07

Figure 5.6: The 8× 8 mesh simulation models

Therefore, all 64 ARs connect to two GRs and can access the Internet through either GR. (HMRSVP selects only one of the two GRs as the GMA of a region to serve all 64 MAs simultaneously.)

According to the wrapped-around mesh topology, when an MN moves right and away from the cell served by AR₇₀, an inter-region handover occurs and the MN will enter the cell served by AR₀₀ of a new region. Similarly, when the MN moves down and away from the cell served by AR₀₇, an inter-region handover occurs and the MN will enter the cell served by AR₀₀ of a new region. As mentioned previously, both MRSVP and HMRSVP designate a specific node as the FP of a data stream for a session. Therefore, MRSVP may select R₁ as the anchor point, whereas HMRSVP may select either GR₁ or GR₂, statically, as the GMA.

According to the mesh topology, each MN can make at most four resource reser-vations (one active reservation and at most three passive reserreser-vations). However, we assume that the bandwidth of a passive reservation in a cell, which is made by an MN in a neighbor cell, can be borrowed by another MN that is making a handoff to the cell.

We compare IARSVP with MRSVP and HMRSVP in terms of the forced ter-mination, reservation blocking, and session completion probabilities with respect to system loads. The system load is denoted as ρ and is equal to ^λ_μ, where the λ is the arrival rate and the μ is service completion rate of resource reservations. In this paper, we assume the inter-arrival time (¹_λ) and the holding-time (_μ¹) of re-source reservations are both an exponential distribution with a mean of ¹_λ and ¹_μ,

IJĶ ijı ijĶ

őųŰţŢţŪŭŪŵźġĩĦĪ

ŊłœŔŗő ŎœŔŗő ʼnŎœŔŗő

ı Ķ IJı

ıįIJ ıįij ıįĴ ıįĵ ıįĶ ıįķ ıįĸ ıįĹ ıįĺ IJ

ŇŰųŦťġŕŦųŮŪůŢŵŪŰůġő

ŔźŴŵŦŮġŐŧŧŦųŦťġōŰŢťġĩĦĪ

Figure 5.7: Forced termination probabilities respectively.

Fig. 5.7 shows the forced termination probabilities for the three schemes under discussion. In general, the forced termination probabilities increase in all schemes when the offered load increases. The forced termination probability of IARSVP is lower than both MRSVP and HMRSVP, with a difference of up to 16%. This is because that IARSVP can dynamically relocate FPs and reserve resources on a more optimized route. As a consequence, it can reuse the resource of an active reservation for passive reservation and thus avoid making unnecessary resources reservation on a long-winded routing path, from a CN to an anticipated location of an MN. Therefore IARSVP can spare more bandwidth for fulfilling more resource reservation requests made by other connections. On the contrary, MRSVP and HMRSVP select FPs statically as mentioned above. In particularly, HMRSVP must initially assign a unique GMA to a region, and any packet destined for the region must travel through the GMA since the GMA hides the local movements of the MN from the nodes outside the region.

Furthermore, unlike HMRSVP, both MRSVP and IARSVP schemes can make use of the multi-homing characteristics and configure resource reservation paths through AR₁ (GMA₁) or AR₂ to the Internet. However, MRSVP makes end-to-end resource reservation for each path independently, either passive or active, from the

ı Ķ IJı IJĶ ijı ijĶ

ıįIJ ıįij ıįĴ ıįĵ ıįĶ ıįķ ıįĸ ıįĹ ıįĺ IJ

œŦŴŦųŷŢŵŪŰůġŃŭŰŤŬŪůŨġőųŰţŢţŪŭŪŵźġĩĦĪ

ŔźŴŵŦŮġŐŧŧŦųŦťġōŰŢťġĩĦĪ

ŊłœŔŗő ŎœŔŗő ʼnŎœŔŗő

Figure 5.8: Reservation blocking probabilities

access router of a CN or the MN’s HA to the current or neighbor proxies of an MN. Therefore, MRSVP may waste bandwidth on redundant reservation. On the contrary, IARSVP can make resource reservation locally from an MN’s passive FP to the proxy of an anticipated location that the MN may visit next.

Fig. 5.8 presents the reservation blocking probabilities for the three schemes under discussion. It is clear that when the offered load increases, the reservation blocking probabilities increases in all schemes. According to the same reasons men-tioned in the descriptions of Fig. 5.7, we can observe that the reservation blocking probability of IARSVP is lower than both MRSVP and HMRSVP, with a maximum difference of about 13%. Finally, Fig. 5.9 depicts the session completion probabil-ities for the three resource reservations schemes. We can further observe that the IARSVP also outperforms HMRSVP and MRSVP in terms of session completion probability due to the same reasons.

From the performance results mentioned above, we could conclude that the ca-pability of autonomous operation of MIAs makes IARSVP better in locating FP dynamically. As a consequence, IARSVP is much superior to both the MRSVP and HMRSVP in supporting QoS-guaranteed resource reservations for real-time appli-cation in Mobile IP networks.

ĹĶ ĺı ĺĶ IJıı

ġőųŰţŢţŪŭŪŵźġĩĦĪ

ŊłœŔŗő ŎœŔŗő ʼnŎœŔŗő

ĸı ĸĶ Ĺı ĹĶ

ıįIJ ıįij ıįĴ ıįĵ ıįĶ ıįķ ıįĸ ıįĹ ıįĺ IJ

ŔŦŴŴŪŰůġńŰŮűŭŦŵŪŰů

ŔźŵŦŮġŐŧŧŦųŦťġōŰŢťġĩĦĪ

Figure 5.9: Session completion probabilities

5.1.5 Summary

we present an IARSVP scheme to enable the resource reservations and smooth handoff for Mobile IP networks. In IARSVP, MIAs can perform binding updates on behalf of MNs away from their home networks. Furthermore, the capability of au-tonomous operation of MIAs makes IARSVP more effective in resource reservation because MIA can adjust FPs dynamically to avoid redundant resource reservation, support route optimization and regional registration, and find the alternative routes in accordance with the network topology and resource usages. Simulation results show that the proposed approach, compared with MRSVP and HMRSVP, can sig-nificantly improve link utilization and reduce disconnection rate for MNs. In the future, we plan to propose a qualitative analysis model to evaluate the performance of IARSVP.

5.2 Mobile Bandwidth-Aggregation Reservation