Gap-ProcessingTime Analysisof Stall Avoidance Mechanisms forHigh SpeedDownlinkPacket Access withParallel HARQ Schemes

The parallel multi-channel stop-and-wait (SAW) hybrid automatic repeat request (HARQ)mechanismisone of keytechnologies forhighspeeddownlink packet accessinthe wideband codedivision multiple accesssystem. However, this parallel HARQ mechanism may encounter aserious stall problem, resulted fromtheerrorofthenegative acknowledgement (NACK) changing totheacknowledgement (ACK)inthecontrol channel. Inthestall situation, thereceiver waits forapacket thatwill beno longer sentbythetransmitter andstopsdelivering themedium accesscontrol (MAC)layer packets totheupperlayer. Thestall issue seriously degrades thequality ofservice forthehighspeed mobile terminal owingtothehighprobability ofNACK-to-ACK errors. Inthis paper,we present an analytical approach tocompare threestall avoidance schemes: thetimer-based, thewindow- based, andtheindicator-based schemes. Tothisend,we first define a new performance metric - thegap-processing time - theduration fora nonrecoverable gap appearing intheMAC layer re-ordering buffer until itisrecognized. Second, we derive theclosed-form expressions fortheaveragegap-processing time ofthese three stall avoidance schemes. Itwill beshownthat our analytical results matchthesimulations well. Further, byanalysis we demonstrate thattheindicator-based stall avoidance scheme outperforms thetimer-based andthewindow-based schemes. Thedeveloped analytical approaches provide important insight intodetermining a proper numberofprocessesintheparallel SAW HARQ mechanism whenapplying theindicator-based stall avoidance scheme.Moreover, our analysis can facilitate the physical/MAC/radio linkcontrol cross-layer design because the gap-processing timeisclosely related topacket errorrateinthe physical layer, reordering buffer size intheMAC andRLClayer.