1000-9825/2002/13(12)2251-08 2002 Journal of Software Vol13 No12 ( 100080) E-mail lisa_qiao@sinacom htt//ieliscasaccn QoS(quality of service) P393 A [1] Internet [2] QoS [3] 1 11 m (m>1) 12 [4] 2001-03-09 2001-12-05 (6989625079931000) (1973 ) (1973 ) CSCW (1978 ) GIS (1963 ) (1944 ) CIMS
2252 Journal of Software 200213(12) 1 Sta() 2 Attri() 1 Attri( ) 0 m (m>1) 1 2 m Γ (1) Γ θ θ = a r D v E ( a a 0r r 0 r a D D 0v = 1 Attri( ) = 1 v > 1 Attri( ) = 0 E E E { Attri( ) = 1 } E ( et( ) et( ) et( ) ) = 1 2 m et() j j (j=1 m) jj=1 m r +et() j D { Atrri( ) = 0} E = ( et( ) ) i = 1 v j = 1 i j v m m et ( ) i( i) j(j=1 m) i j j1 j m et() 1j et() 2j et() v j i j1 i v 1 j m r +et() ij D (2) (3) (4) 2 ) 21 3 [56] 4 [6] 5 EA s R EA e R 6 ESP( P R ESP( = Max( r availtime( Max availtime( P Max EA u R R R R EA u )
2253 u=s u=e 7 IES() PE IES( ) = Min ESP( P PE 8 avail( P P Attri()=1 Attri()=0 1 ESP( ) D avail( 0 ESP( ) > D 1 i1 i v ESP( ) D i avail( 0 i1 i v ESP( ) i > D 9 sysavailtime sysavailtime=min P PE (availtime() 10 Se( P P Se( 11 Finishtime( P ESP( ) P Finishtime( ESP( ) ip i i=1 v 22 Attri( ) = 1 Attri( ) = 0 [5] S [78] sysavailtime( ) sysavailtime S P S W W >0 S( = availtime( + W * Se( S S S choose() ESR Γ us Res()
2254 Journal of Software 200213(12) mod(r) R s R mod( R) e R (1) Res()= P S( = Max e { e e PE avail ( e) = 1} S( e) (2) Res() (21) (22) us us Γ us Res( us ) Res()= (1) (23) RR Res() mod(r)=s us us Γ us RR Res() mod( us R)=s (1) (24) P availtime( P ) = Max e { e e PE avail ( e) = 1} availtime( e) (241) r ESR ESR =Max e {e e PE avail(e)=1} availtime(e) Se(P )=Max e {e e PE avail(e)=1} Se(e) (1) (242) r ESR r Max e {e e PE avail(e)=1} availtime(e) Se(P )=Max e {e e PE avail(e)=1} Se(e) (1) (243) P Finishtime( = Min e { e e PE avail( e) = 1Finishtime( 23 (QoS) (QoS) ( ) degrade() degrade() (1) If then (11) (12) If then (13) (11) (12) (2) else exit 24 K d W H (1)
2255 (2) K ( K ) (3) For i=1 to K (or less than K) (31) If i hen degrade( i ) (4) K ( K ) feasible=true feasible=false (5) If (feasible==true) (51) H H()=d +W*IES() (52) H choose() Else (53) (54) H choose( ) (6) (7) (2)~(6) (a) (b) H (c) 3 [5] 31 12 Rand(xy) x y (1) m 1 2 m λλ=(λ 1 λ m )λ 1 λ m λ i+ 1 λi = λi λi 1 i = 2 m 1 (2) β β=λ i+1 /λ I i=1 m 1 β β (3) [5] ) (4) as_p 1 as_p (5) ( ) v (Max_v ) 1 v Rand(1Max _ v) (6) (Min_C Max_C et( ) Rand(Min _ CMax _ C) = λ et ( ) m 1
2256 Journal of Software 200213(12) et ( ) j j=1 1+1 m (7) et( ) Rand(Min _ CMax _ C) et i = ( ) s < et( )( s+ 1) s = 1 v λ et ( ) i m 1 et() i j (i=1 v j=1 1+1 m) et ( ) s < et ( ) ( s+ 1) (s=1 v 1) (8) D D = Rand ( SC(1 + R)* SC) SC R i (9) Use_P Share_P Use_P Share_P 32 200 800 30~60 3 KH WS W P Share_PMax_v 172205 10 1~ 4 5 Scheduling success ratio 95 85 75 65 55 45 0 01 02 03 04 05 06 07 08 09 as flow (as_ 1 11 12 13 14 15 16 17 Processors seed difference New algorithm Myoic algorithm New algorithm Myoic algorithm Scheduling success ratio Fig1 Effect of tas flow (as_ Fig2 Effect of rocessors seed difference on scheduling success ratio on scheduling success ratio 1 (as_ 2 (β) Scheduling success ratio 90 70 50 30 10 0 01 02 03 04 Use_P New algorithm Myoic algorithm Scheduling success ratio 100 90 80 60 70 50 40 20 30 100 90 80 70 60 50 40 0 1 j 001 002 003 New algorithm j 004 R 005 006 007 Myoic algorithm Fig3 Effect of resource usage (Use_ on scheduling success ratio Fig4 Effect of laxity (R) on scheduling success ratio 3 (Use_ 4 (R)
2257 321 (as_ (as_ 1 Use_Pβ R 0511 001 as_p 1 1 2 322 (β) 2 as_puse_p R 0102 001 β β β 1 β<15 20%~30% 323 (Use_ 3 (Use_ as_pβ R 0111001 Use_P Use_P 26% 324 (R) 4 βuse_p as_p 1102 01 R 70% 35% 4 R CPU 4 (QoS) S QoS 30%
2258 Journal of Software 200213(12) References [1] Lehoczy JP Ramos-huel S An otimal algorithm for scheduling soft-aeriodic tass in fixed riority reemtive systems In Proceedings of the 13th IEEE Real-ime System Symosium IEEE Comuter Society Press 1992 110 ~123 [2] Kaneo H Stanovic JA Sen S et al Integrated scheduling of multimedia and hard real-time tass In Proceedings of the 17th IEEE Real-ime System Symosium IEEE Comuter society Press 1996 206~219 [3] Manimaran G Murthy CSR Integrated dynamic scheduling of hard and QoS degradable real-time tass in multirocessor systems In Proceedings of the 5th International Conference on Real-time Comuting Systems and Alications Jaan 1998 [4] Liu JWS Shin WK Liu KJ et al Imrecise comutations Proceedings of the IEEE 1994 82(1)83~94 [5] Ramamritham K Stanovic JA Shiah P-F Efficient scheduling algorithms for real-time multirocessor systems IEEE ransactions on Parallel and Distributed Systems 199011(2)184~194 [6] Manimaran G Murthy CSR An efficient dynamic scheduling algorithm for multirocessor real-time systems IEEE ransactions on Parallel and Distributed Systems 199819(3)312~319 [7] Qiao Ying Wang Hong-an Dai Guo-zhong Develoing a new dynamic scheduling algorithm for real-time multirocessor systems Journal of Software 200213(1)51~58 (in Chinese) [8] Qiao Ying Zou Bing Wang Hong-an et al Develoing a new dynamic scheduling algorithm for real-time heterogeneous systems In Lin Yen-Chun Shen Hong eds Proceedings of the 2nd International Conference on Parallel and Distributed Comuting Alications and echnologies 2001 8~15 [9] an S Wang N Zhao YF et al A constrained finite element method for modeling cloth deformation Visual Comuter 199915(2)90~99 [7] 200213(1)51~58 Design and Evaluation of an Algorithm for Integrated Dynamic Scheduling in Real-ime Heterogeneous Systems QIAO Ying ZOU Bing FANG ing WANG Hong-an DAI Guo-zhong (Intelligence Engineering Laboratory Institute of Software he Chinese Academy of Sciences Beijing 100080 China) E-mail lisa_qiao@sinacom htt//ieliscasaccn Abstract In this aer an efficient algorithm is resented to dynamically schedule the tas sets combining hard and soft real-time tass in heterogeneous systems he roosed algorithm imroves the scheduling success ratio by introducing a new tas assignment olicy and a QoS (quality of service) degradation olicy for soft real-time tass o evaluate the erformance of the new algorithm extensive simulation studies have been done hese simulations aly myoic algorithm to schedule the hard and soft real-time tass in heterogeneous systems and use it as a baseline to comare with the roosed algorithm Simulation results show that the scheduling success ratio of the new algorithm is always higher than that of myoic algorithm in real-time heterogeneous systems for a variety of tas arameters Key words heterogeneous system hard real-time soft real-time dynamic scheduling scheduling success ratio heuristic bactrac Received March 9 2001 acceted December 5 2001 Suorted by the National Natural Science Foundation of China under Grant Nos69896250 79931000