Performance Evaluation of a QoS-Aware Handover Mechanism

Transcription

1 Performance Evaluation of a QoS-Aware Handover Mechanism

2 1.Introduction

3 Background Requirements in mobile communication Seamless mobility support Guarantee certain levels of QoS Mobile communications over IP QoS support is not clear yet Handover from one base station to another Base station may not have sufficient resources Base station may not have sufficient quality Base stations may be overlapped

4 Technical requirements Ensuring the data flow to the new station Already solved by Mobile IP How to make an IP-based mobility solution QoS aware Signaling using IP Guarantee handover only when resources are available

5 Objective: QoS-aware Handover Add QoS information in Binding Update Provide quantitative results using simulation Previous work QoS-Conditionalized Handoff in MIP6 Only mechanism, no reults Compare with a QoS unaware situation Introduce additional refinements

6 2.Related work

7 Mobile IPv6 Mobile Node Identified with HoA & CoA Home Agent Located in MN s home network Intercept packets while MN is away from home High signaling load Long handover latencies

8 Hierarchical MIPv6 Overcome MIP6 drawbacks for local handovers Mobility Anchor Point (MAP) Proxy for HA in a foreign network Mobile node at foreign link with MAP Regional CoA (RCoA) On-link CoA (LCoA) Local binding update is sent to MAP Global binding update is sent to HA and CN

9 QoS Support for Mobile IP Extension of RSVP with flow transparency Identifying the flow address with HoA Extension of RSVP with mobility proxy Placed at edge of a MAP domain Use RCoA and LCoA to identify session RSVP based approach Latency problems for QoS signaling Signaling overhead

10 QoS Support for Mobile IP (2) Framework for QoS Support in MIP6 QoS option (hop by hop option) in BU Can check QoS resources while handover Problems No feedback to MN Handover always takes place Draft no longer exists QoS-Conditionalized Handoff in MIP6 Covers the problems above

11 3.Proposed mechanism

12 Overview Based on HMIPv6 Short handover latency Low signaling load Establish QoS flows during handover Use Binding Update message Ensure QoS requirements are fulfilled while handover

13 Assumptions Routers Responsible for maintenance QoS resources Defined as QoS entities APs and MAP QoS entities Receive QoS requests Bandwidth of the link Buffer space of the router CPU resources of the router

14 Capabilities of QoS entities Accept a request Reserve the required resources Reject a request Offer a lower QoS level Release a reservation

15 Protocol Operation QoS option Attached as IP Option Header Allows process with Binding Update MN sends BU+QoS to the MAP Each entity between the two checks their resources Entities reply BA+QoS if failed to surve else reserve resources BU is processed if reached at MAP

16 Protocol overview Old route Internet CN MAP BU+QoS router BA+QoS router New route AP1 AP2 MN BU+QoS

17 Problems Releasing reservations Send explicit messages after handover Give lifetime to reservations Limitation in resources MNs can t perform handovers Short lifetime in reservations More signaling overhead

18 4.Evaluation

19 Topology HA MAP Internet CN

20 Scenarios Movement model MN moves between APs randomly at constant speed RA from APs is used for movement detection Load model 400Kbps data from MN to CN, vice versa 70MN on three 10Mbps links

21 Protocols used for MN Application is not QoS sensitive (BE-traffic) HMIPv6 Application is QoS sensitive (QoS traffic) QoS conditionalized Handoff in MIP6 Extension of HMIPv6 QoS conditionalized Handoff in MIP6 with R-flag Allow release of superfluous reservations witch the MN has no link-layer connectivity

22 Metrics used Handover Starts when IP connection is lost When radio connectivity to an AP is lost Ends when BA is received Movement detection Router advertisement If not three consecutive RAs, assume movement Interval is 100ms, therefore ms needed Propagation delay and processing time is 5-13ms

23 Metric used (2) Packet Success Rate (PSR) Metric to show efficiency of the network PSR = (received packets at CN) / (packets sent at MN) Maximal PSR is 98.75% Scenarios and protocols were implemented OMNeT++ simulater

24 Parameters in Evaluation Advertisement interval of AP = 100ms Speed of MN = 6.5UL/sec Position update = 600ms interval Packet length = 8000bit Packet interval from MN = 20ms Queue length in routers = bit Lifetime of QoS reject = 1sec Dimension of the simulation field = 300x300UL

25 Calculation of maximal PSR Maximum PSR of 98.75% in the simulation 3 handovers / min 250ms for best handover(mean) 20ms interval for each packet sent For a minute total downtime for handoff = 750ms Number of packet loss = 750 / 20 = 37.5 Total number of packets sent = / 20 = 3000 ( ) / 3000 = 98.75

26 Number of HO in a min 6.5UL/sec x 60sec = 390UL Radio range is either 71UL or 100 UL MN will at least go through 3 handoff

27 Handover Latency Most handovers: 200ms 300ms Mean Value for handoffs Best Effort: 260.3ms QoS: ms Why? QoS handovers can t connect to a new AP when there are no resource available

28 What is Load Capacity? For each MN: 50packets / sec (20ms interval) Packet size = 8000bit 8000bit/packet * 50 packet/sec = 400Kbits/sec With IP header: 425Kbits/sec APs are connected w/ 10Mbits/sec line 23.5MNs / each line 3 lines for APs: 23.5 * 3 = 70.5MNs 70MNs = 100% Load Capacity!!!

29 Result of Simulation#1 QoS Lifetime: 10s, Radio Range: 71UL Max PSR: Best Effort(92.8%) at Load Capacity of 80% Best Effort & QoS w/ Rflag has approximately the same performance QoS w/o Rflag is lower due to superfluous reservations

30 Result of Simulation#2 QoS Lifetime: 10s, Radio Range: 100UL Max PSR: both QoS(97.8%) at Load Capacity of 80% Best Effort is lower at due to overload situation(mns use same AP) QoS has more APs to choose from better performance

31 Result of Simulation#3 QoS Lifetime: 50s, Radio Range: 71UL Best Effort and QoS w/ Rflag has same result w/ when QoS Life time was 10s QoS w/o Rflag collapses due to superfluous reservations(no timeout = no resource)

32 Result of Simulation#4 QoS Lifetime: 50s, Radio Range: 100UL QoS w/o Rflag is almost equal to QoS w/ Rflag until 95% No difference to Best Effort Not much difference to QoS w/ Rflag as well

33 Result of Simulation#5 Radio Range varies from 71UL 150UL PSR: no increase from 110UL Handover Latency: 100UL for the best performance

34 Conclusion For the best performance Choose right radio range QoS is useful when uplink is narrow Radio Range influence w/ QoS Too small: low PSR, longer HO time Too big: high PSR, longer HO time FIND the BEST RADIO RANGE for optimal performance!!!