ࡱ> s@=ī|A ')0g5~Hd v<xx] TՕ>M=:: jc34a,)7h/ft1 UYB%: d\5n&)ND5#;ܹ^3󬵧ܟsϹ^v j.ѿ8tDc'p""L4Q iGnN5!Rf&Gwd6+ς˵RT+>4T| xxyT~MSXN?L}H<@gIAR菉k,w*P /K_1JRTSSCtgرci4i$6mВ%Kh4g6[oR^}v=,w+eQJTHddA9PTyrQ@E'7fA9PTy, ʁ "㴇4_TA@PPYPTA@PYPTAA*>&C]m(V12U JJcxϟ'C>a|z%ԫŝ0 ۜ`.:*[>[5#ͣVsp%>7?5: W6;(Nzq]8q+}H?wnL37?:@=k:ܜ{2nq?Wc?o'h;mNZ82:U*[q}%kF?V^?B>@"CO>96;gws䑠o㜧^u'k_}CTM&CB?n~ lDn,`̶Zi}-Nn!}vvaӑo՜.1g265dt#OWJ1D纏qca7OK755bjHY'?U+WGǵiqoDu2 ei!6vUOer=t p[;EXqg;|K=1鏏17 |o15 =>v4EK۬4{33< qwzR7:>?nr(7 ~b[ |1fFE!t7xwOwe2' '?NF40YuPdu@s6Io~M^_ߩ1j26lcvI-}􊖮w<^/ח^^- KH߭AwR*C7}dwPu{C4txGJ؇q CC(C lϕ+W:\a0atӜ:\a_9v'6tҢ%:L/S~͵Jl35/|V'_ 9퐪 9~xS3#Cިs1mx+ju܎.ܽ9-Jj|-x9rzQ96@'гڿ{IQͣ6㺟 \qq!|>?Dcۍp#‡Dn"|^?DzǦN\w9{~?觓Zt|};EǗo#}7Nw$֥6Fb T߻ ;n~Nj<9-y{A]}<~49t-#ߌ -w{뼶ݿs}$Tt99Y&Qta=<#P]_jӅx vIE6]x߁pJ.j? 3Ÿ%TaӅEal}=a?P]~gT`Ö^̐y\&!V+YQϏ)|u|qKEBjIO9F;^z@O~jӦi'ꝴO~j֦'WdY@]`$m,7~;~w9o#8;AO~yM?^'s#$꧞ï?,$꧞̈́=[zo󻑋P.+FH^#KJ]Wv}UQ zpmTλ~e2k#%Kɚp\?^S5YӰT)炕ѳ>,k!:쑽q3y]xy ǂ $H=ojBG,a6uE!r+׶JuA'FJh8 ld [ዥ}D-bD2*auiHfIǘQm3vH[{#Hf#ȭ=Xh;-]*e#O%ltL-FJ0O71Dnl3,h#NY%l4ql'Sa#%K/"_T?ۂOIzg/o?7w6ow M|N;.-Zv a}>;~WV7L|%7|'1_l&>߂/_(6^_u7|Ųe%o:{-/Lco!I[wڏco"I؛_PUjocm7+Y6&ȭML.m'X0'E,yC%΍C&CD|C}S}y<~W`xyߝUtSb7DJ +@3iXm!Fsրçh-$x:4hk5J z|#ںlqU3¸NԖcI%Eð\BR}V?pӿ))ǖF.2kTXOWkdO{Ip2V] >Se; 3 ,we=.e*cuIz}ORb_ۥou{eJj )kU5WZReͲ蕤2qM7!c[1Bael 6gȽ'&,roB"&YluUX^)~:1bm{!|ccȻ3!|VexΘ20c$0;K2禖ǙKB[-cSOOq? 㨶~h߲C:י7uU:׹:ww\[3|1jmNfc܆kt6h>3i䚱Nftsߡ ~Lowje~s֦LSzn>3Nt4pGӜAMdhrH!uz6u+3?hhv3Yds]3h$DlȾu]m,-~gy韨ΙsfOw]= ΟaYR %\مn0sJGX!i{yq jqz_vOGΑ9w?yWߐAa#,> Tvf +.:<ζ:TY" zN"GbG <V.W7SgΏ\{v78* _f[n3leC6e9|[hv-Xg?QҲjyCe(g4s:?4ID>tB3],Q}譐k>Y 1po>ȿ9u78sŮenEk=QrZߟ5;XUFٕTq ProblemGiven a workflow and a Grid, how to allocate resources and schedule the workflow in order to optimize performance. Which resources should be provisioned When they should be provisioned How much capacity should be provisioned Scheduling on provisioned resources*sZZs Previous WorkNo allocation in case of time sharing systems. Combined allocation and scheduling in case of queuing based systems. Scheduling of data is coupled with scheduling of computations. Operations Research (inventory control, MRP)  Formal Model An allocation plan consisting of allocation requests and cost AP = {ar1, ar2, , arn} ar = <s,t,n,d,h> CA(AP) = f(ar1, ar2, arn) e.g. utilization, wait time A policy, S, for scheduling partial workflows and associated cost, CSCH costs can be makespan, lateness, reliability etc. The goal is to minimize the total cost of allocation and scheduling. CT = .CA(AP) + .CSCH>rI3F> JJ JJ JJJDJ3BFB B  J  B  J  B  J V  ti Relation to previous workThe AP defines the Virtual Grid. Can be used to explain previous scheduling approaches based on the appropriate cost formulations. Dedicated resources imply constant allocation cost Others imply the cost of allocation is the wait time associated with it. &}} Approaches to AllocationDeterministic (using reservations) Stochastic (using predictions) A combination of both The space of possible allocation plans is likely to be exponential even with reservations. Search based heuristics might be useful.    Preliminary WorkJava based Grid Simulator A parametric task graph generator 4 Grid sites with background load Backfill based FCFS scheduling policy for Grid resources. Resources can be queried for earliest start time of a request.  Preliminary Work_Preliminary results show good results with very simple allocation strategy and small workflows.`` Current Research$Modeling resource availability in Grids Heuristics for creating an allocation plan Greedy approach Search based metaheuristics Experiments to evaluate the performance. Implementation to be integrated with current planning and execution systems such as Pegasus. <S,S,p< ` ffff!` 33` 3fff` 3PPf` fff̙3f` 3f33` ff3fff` fff̙` ̙>?" dZ@,|?" dd@   " @ ` n?" dd@   @@``PT    @ ` `4p>> P8H81107(  03T  0 " 0  hBC"DEFJ.. e%/5BEMebebur7p_Uz?dM<H E"5 %% "$@               `" b b@ 0C# n"l (I  0 xBSCuDEFSuS @`"b 0 xBCbDEFbv1b @`"~@ 0 xBC1DEF1 @`"5p 0 2BCDETF~UUaSHfup`R@@&^F pG` Mr_&x669D>2}uynpbaN@N@BJ 6%:| <@@                            `"Z  L b ) $  0# "    0B BCDEF00.%mR 9#-< HQXh  &-($!}tkc[!R$J)B.4=*P%g#~#&).bd@`"}  l  0B .BCDEF::mpv!4Li}unf\RH=1% !+5?IS[dlt{o_O. ~wmvx@`") /O d  0B &BCDEdFn^i s{2Not^<1& .?Op[Sa.^48@`"y[ t  0B 6BuCDElFvK$./B>TPfgtuxsklZcJY:N-C$: EOYbimqrK8<@`"9&   0B BCMDE4F>  ,< M5% @`"!? 6v Nb  0C# *J"O $ 0 NcFBC4DExF ,Hc}+E=lSj 4gWpKQD5?8,)+D2 ,>@@`" 0 cFB/CDE<FF7e-//-g)G$4"  $@`"= A, 0 cFBWC(DEHFRWMD :/%  !*4@!J(W&(@`"b b@ 0# $"G 0 xBSCuDEFSuS @`"b 0 xBCbDEFbv1b @`"~@ 0 xBC1DEF1 @`"5b b@ 0# bS"z 0 xBSCuDEFSuS @`"b 0 xBCbDEFbv1b @`"~@ 0 xBC1DEF1 @`"5\ b@ 0 "/o 0 xBSCuDEFSuS @`"b 0 xBCbDEFbv1b @`"~@ 0 xBC1DEF1 @`"5V 0  BCDEFY+ "'*33:tXRoJ=%   (Fen(<Pik{ Bm#)4|9|@wDpI_L_LEZwH3 ~wrmold^T={(cW H=1%  X\@                                          `"W^Rd  0 & BmCDEdFn% :N,`Gklmic]WQJB:5~5o8`;P8>0+48@`"L v $ !0  B6C(DEDFN  &.652!,&%(' $(@`" L3$ "0 BvCDEDFN!&36HC^OwW^[vuhwRE/$(@`" | #0 >BCDEpFzg}~oy\oDg)^Z]H FB&B3CAFNNQJQ:R%RPMLKLO Yg:<@`" $0 B/CVDE,F6 %%;VRI!?*3/$.%@`"m/ %0 BCDEF-- )eC2~ZGJl;0{MRf%}Z(e@kRAl8D7 =F \`@`"  &0 fgBCDEF!!2K bw&3CVjpaTH=}4k+X!E/ DH@`"l '0 NgB|CnDExFM|lxkkiYeDc-ab fnbYRL$H/G;GHJIGF8C&A?=;99;BM>@@`"gd (0 &gBmCDEdFn% :N,`Gklmic]WQJB:5~5o8`;P8>0+48@`"> )0 gB.C^DE4F> &> O^ XP E(9-,.* @`"*2$ *0 gB6C(DEDFN  &.652!,&%(' $(@`"I +0 BhCDEF&&  '):5S?lIMPLC6;WroE# )tR+'L}m4VO1`hdUX7'NP@`"vh ,0 fgBTCDEFAAtr hU>! -tETe7 :| !D q*AOT[SI6 >ghiG c1p]UK"A=?GTyee{TE90(!1St@`"8T\ -0 S # "A`} # T Click to edit Master title style! !  .0 C P# " ` # RClick to edit Master text styles Second level Third level Fourth level Fifth level!     S /0 C # "] `} # F*0 00 C # "`  # H*0 10 C # "] `} # H*0B 0 s *޽h ? fff̙80___PPT10.(02 Balloonsg1  00004m0(  4<,T  4 " b  3 4# i"j 4 fBTCDEFAAtr hU>! -tETe7 :| !D q*AOT[SI6 >ghiG c1p]UK"A=?GTyee{TE90(!1St@`"  4 fBCDEF!!2K bw&3CVjpaTH=}4k+X!E/ DH@`"   4 NB|CnDExFM|lxkkiYeDc-ab fnbYRL$H/G;GHJIGF8C&A?=;99;BM>@@`"7wLp d 4 &BmCDEdFn% :N,`Gklmic]WQJB:5~5o8`;P8>0+48@`", 4 B.C^DE4F> &> O^ XP E(9-,.* @`"lp $  4 B6C(DEDFN  &.652!,&%(' $(@`"7-  4 BCDEF**)%:5IFXZdqpx|~xm.\VC' 6] )'$!n#?):V9mUQ,,VX@`" ( 3d  4B &BCDEdFn^i s{2Not^<1& .?Op[Sa.^48@`"Y.   4 BCDEF//{:[It&m~An,$:OgR)LP5j+ W3\^]{`d@`"   4 NB!CDExF%EDn:4Pmm}=}j!OyR^#du<[&>@@`"2 ! t 4B 6BuCDElFvK$./B>TPfgtuxsklZcJY:N-C$: EOYbimqrK8<@`"'   4B BCMDE4F>  ,< M5% @`" v \ 4 BTCDE`Fj_]^ 3zEAemz&uN4l!T24@`" b b@ 4# 5"%  M 4 xBSCuDEFSuS @`"b 4 xBCbDEFbv1b @`"~@ 4 xBC1DEF1 @`"5b b@ 4# z" f 4 xBSCuDEFSuS @`"b 4 xBCbDEFbv1b @`"~@ 4 xBC1DEF1 @`"5b b@ 4# &"  4 xBSCuDEFSuS @`"b 4 xBCbDEFbv1b @`"~@ 4 xBC1DEF1 @`"5b b@ 4C# &".a& 4 xBSCuDEFSuS @`"b 4 xBCbDEFbv1b @`"~@  4 xBC1DEF1 @`"5b b@ !4C# Y"N   "4 xBSCuDEFSuS @`"b #4 xBCbDEFbv1b @`"~@ $4 xBC1DEF1 @`"5J %4 B^CvDEXF`k%S"RFCov^SXCm-NSi-0@"x &4 fBCDEF!!2K bw&3CVjpaTH=}4k+X!E/ DH@`"nf  '4 NB|CnDExFM|lxkkiYeDc-ab fnbYRL$H/G;GHJIGF8C&A?=;99;BM>@@`"Z p (4 B.C^DE4F> &> O^ XP E(9-,.* @`"(+  )4 B|CyDE@FJ|q cQ)?6)BJK `q)y<ySoeXt5|"$@`"Cy *4 BCDEF**)%:5IFXZdqpx|~xm.\VC' 6] )'$!n#?):V9mUQ,,VX@`"}  +4 B$CDEF$ @`"8$ ,4 C Hm "` ` m F*0 -4 C Mm "`  m H*0 .4 C Qm "] `} m H*0 /4 BUm3 3fԔ8c?"x HH  m T Click to edit Master title style! ! 04 C Wm " @0 m W#Click to edit Master subtitle style$ $B 4 s *޽h ? fff̙80___PPT10.(02   0(  x  c $cm/4P m x  c $pdm04 @0 m H  0޽h ? 33___PPT10i. 7FM+D=' = @B +}  <$(  <r < S ` #-0A`}  # r < S  #.0 ` # H < 0޽h ? fff̙___PPT10i. 7 ~t+D=' = @B +}   H$(  Hr H S -0A`}   r H S .0 `  H H 0޽h ? fff̙___PPT10i."7F +D=' = @B +}  @$(  @r @ S -0A`}   r @ S .0 `  H @ 0޽h ? fff̙___PPT10i. 7py+D=' = @B +}  D$(  Dr D S  -0A`}   r D S .0 `  H D 0޽h ? fff̙___PPT10i."7)v+D=' = @B +}  0L$(  Lr L S -0A`}   r L S p.0 `  H L 0޽h ? fff̙___PPT10i.#7H+D=' = @B + 2`+(  ` ` <XܱP ` YGlidein  ` < p 0` >TG/OSG 8  `TB ` c $DppTB ` c $DTB ` c $DppTB ` c $D`TB ` c $D@@TB ` c $D` ` 0P` ` 0P` ` 0P` ` 0` ` 0` ` 0` ` 00` ` 00` ` 00dB ` <D1@ p @X ` 00p ` 0@  ' <Condor-G 2  ` <vm"`  c+Execution stack for SCEC, Montage workflows, ,`" ` s *"`  ` 0`` <Grid 2*8  0 `` p ` <X@   hManager DAGMan &N  `  P TB ` c $DppTB ` c $DTB ` c $DppTB ` c $D`TB ` c $D@@TB ` c $D` ` 0P` ` 0P` ` 0P` ` 0` ` 0` ` 0` ` 00` ` 00` ` 00` ` 0 h" ` s *"` 0 ` 0p j NProvisioned Resource( 2dB ` <D1@@ ` 0  :AP 2 ` 0,ؗp j  9S 2dB ` <D1 @ H ` 0޽h ? fff̙___PPT10i.80+D=' = @B + K0 @P$(  Pr P S  }m-0A`}  m r P S }m.0 ` m H P 0޽h ? fff̙___PPT10i.$7 +D=' = @B + K0 PT(  Tr T S -m-0A`}  m r T S .m.0 ` m j T s *A .000W mH T 0޽h ? fff̙___PPT10i.<7M+D=' = @B +}  p\$(  \r \ S m-0A`}  m r \ S hm.0 ` m H \ 0޽h ? fff̙___PPT10i.=7M"+D=' = @B +r8UQĂU_ڇiz&?  X( b/ 0DArialax#0z[ 0"DVerdanax#0z[ 0" A .  @n?" dd@  @@`` ldd  Oh+'0` hp   $ 0<D1Resource Allocation and Scheduling for Workflowsicrgurmeet BalloonsAllGurmeet Singhat17mMicrosoft Office PowerPoint@W@Q 7@lP9CGg    Y-- @ !Y--'--$A.024|7x:q<j>b?Z?Q=H:?66/.%& "+6BP]kyynbWLB:2,)'( *,0$4*9.?0E2L3S4[3c2j1r0z.--'--F$!*(%"   !%*--'--@$--'--6$1111259@KKKLLLLLKIEB>:741--'--$ --'--&$STTTTTTTSUWXYZYWS--'--X$*'5FUbny{}|n^J4--'--6$      !! --'--b$/n"t}~uncT?7/)$ '.4:AFLQW\`fjov{||yvrn"--'--@$ #',0 2&5,748;8C7K5S2L7E:=<7>1?*?$?>=;8 64.--'--:$m)445z7t;m?gE`M[W[S\N_HcCf=j9n6q5n3k3h2e1c0a/`/`/m)--'--$ T5V6X7Z9[=\A\GZLWRWIVCU>T5--'--4$vj~_|Tk|--'-- $?V<I$R?V--'-- $.mOvJ.m--'-- $v+!v--'-- $w=uFAw=--'-- $.t&|.--'-- $**--'-- $t{xit--'-- $asYPas--'-- $Z`Cs@kZ`--'-- $vHSfgvH--'-- $PL^"N!PL--'-- $2`%8B2`--'-- $NZ:N--'-- $(+<.-(--'-- $--'--0$~ *9EOW[\XYWRLC:/# --'--F$! !!!! '-27<A EHKNOLHD@ < 83/*& --'--@$[zyxusqooqnlkjjjkmlhdaaa`_^]\[--'--$ U]cgjihfda]YU--'--$$ --'--X$*,9 F#R%\(gdaYM>,--'-- $--'@"Verdana-. %2 :Resource Allocation "!.. ff%2 9Resource Allocation "!."System-@"Verdana-. $2 Gand Scheduling for .. ff$2 Fand Scheduling for .-@"Verdana-. 2 Workflows1*.. ff2 Workflows1*.-@"Verdana-. %2 7:Gurmeet Singh, Carl      .. f%2 69Gurmeet Singh, Carl      .-@"Verdana-. 2 7q Kesselman.. f2 6p Kesselman.-@"Verdana-.  2 7, .. f 2 6, .-@"Verdana-.  2 WEwa.. f 2 VEwa.-@"Verdana-. 2 W Deelman .. f2 V Deelman .--. 2 W Deelman .. f2 Root EntrydO)R9PicturesCurrent User GSummaryInformation(db/ 0DAriala}0z[ 0DVerdana}0z[ 0" A .  @n?" dd@  @@`` ldd    2 1,"$ī|A ')0g5 0AA@83ʚ;ʚ;g4BdBd z[ 0(ppp@ <4ddddL 0} 80___PPT10 ?  %F0Resource Allocation and Scheduling for Workflows*Gurmeet Singh, Carl Kesselman, Ewa Deelman> ProblemGiven a workflow and a Grid, how to allocate resources and schedule the workflow in order to optimize performance. Which resources should be provisioned When they should be provisioned How much capacity should be provisioned Scheduling on provisioned resources*sZZs Previous WorkNo allocation in case of time sharing systems. Combined allocation and scheduling in case of queuing based systems. Scheduling of data is coupled with scheduling of computations. Operations Research (inventory control, MRP)  Formal Model An allocation plan consisting of allocation requests and cost AP = {ar1, ar2, , arn} ar = CA(AP) = f(ar1, ar2, arn) e.g. utilization, wait time A policy, S, for scheduling partial workflows and associated cost, CSCH costs can be makespan, lateness, reliability etc. The goal is to minimize the total cost of allocation and scheduling. CT = g(CA(AP), CSCH)>rI3F> JJ JJ JJJDJ3BFB B  J  B  J  B  J  B  J tV  ti Relation to previous workThe AP defines the Virtual Grid. Can be used to explain previous scheduling approaches based on the appropriate cost formulations. Dedicated resources imply constant allocation cost Others imply the cost of allocation is the wait time associated with it. &}} Approaches to AllocationDeterministic (using reservations) Stochastic (using predictions) A combination of both The space of possible allocation plans is likely to be exponential even with reservations. Search based heuristics might be useful.    Preliminary WorkJava based Grid Simulator A parametric task graph generator 4 Grid sites with background load Backfill based FCFS scheduling policy for Grid resources. Resources can be queried for earliest start time of a request.  Preliminary Work_Preliminary results show good results with very simple allocation strategy and small workflows.`` Current ResearchModeling resource availability in Grids Heuristics for creating an allocation plan Greedy approach Search based metaheuristics Experiments to evaluate the performance. Implementation to be integrated with current planning and execution systems such as Pegasus. <S,S,p 2`5(  ` ` <P ` YGlidein  ` <l p 0` >TG/OSG 8  `TB ` c $DppTB ` c $DTB ` c $DppTB ` c $D`TB ` c $D@@TB ` c $D` ` 0P` ` 0P` ` 0P` ` 0` ` 0` ` 0` ` 00` ` 00` ` 00dB ` <D1@ p @X ` 00p ` 0`"@  ' <Condor-G 2  ` <&"`  c+Execution stack for SCEC, Montage workflows, ,`" ` s *"`  ` 0*`` <Grid 248  0 `` p ` <.@   rPegasus DAGMan $&N  `  P TB ` c $D  !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcefghijklmnopqruvwxyz{|}~tRoot EntrydO)}P9PicturesCurrent User GSummaryInformation(dPowerPoint Document( DocumentSummaryInformation8  2 1,"$ī|A ')0g5 0AA@83ʚ;ʚ;g4BdBd z[ 0(ppp@ <4ddddL 0x# 80___PPT10 ?  %P0Resource Allocation and Scheduling for Workflows*Gurmeet Singh, Carl Kesselman, Ewa Deelman> ProblemGiven a workflow and a Grid, how to allocate resources and schedule the workflow in order to optimize performance. Which resources should be provisioned When they should be provisioned How much capacity should be provisioned Scheduling on provisioned resources*sZZs Previous WorkNo allocation in case of time sharing systems. Combined allocation and scheduling in case of queuing based systems. Scheduling of data is coupled with scheduling of computations. Operations Research (inventory control, MRP)  Formal Model An allocation plan consisting of allocation requests and cost AP = {ar1, ar2, , arn} ar = CA(AP) = f(ar1, ar2, arn) e.g. utilization, wait time A policy, S, for scheduling partial workflows and associated cost, CSCH costs can be makespan, lateness, reliability etc. The goal is to minimize the total cost of allocation and scheduling. CT = g(CA(AP), CSCH)>rI3F> JJ JJ JJJDJ3BFB B  J  B  J  B  J  B  J ~V  tiRelation to previous workThe AP defines the Virtual Grid. Can be used to explain previous scheduling approaches based on the appropriate cost formulations. Dedicated resources imply constant allocation cost Others imply the cost of allocation is the wait time associated with it. &}} Approaches to AllocationDeterministic (using reservations) Stochastic (using predictions) A combination of both The space of possible allocation plans is likely to be exponential even with reservations. Search based heuristics might be useful.    Preliminary WorkJava based Grid Simulator A parametric task graph generator 4 Grid sites with background load Backfill based FCFS scheduling policy for Grid resources. Resources can be queried for earliest start time of a request.  Preliminary Work_Preliminary results show good results with very simple allocation strategy and small workflows.`` Current ResearchModeling resource availability in Grids Heuristics for creating an allocation plan Greedy approach Search based metaheuristics Experiments to evaluate the performance. Implementation to be integrated with current planning and execution systems such as Pegasus. <S,S,p}  @$(  @r @ S -0A`}   r @ S .0 `  H @ 0޽h ? fff̙___PPT10i. 7py+D=' = @B +r?  X(  ՜.+,0  $ , BOn-screen Show$USC, Information Sciences Institute9 A  ArialVerdana Balloons1Resource Allocation and Scheduling for WorkflowsProblemPrevious WorkFormal Model Relation to previous workApproaches to AllocationSlide 7Preliminary WorkPreliminary WorkCurrent Research  Fonts UsedDesign Template Slide Titles %_~ Gurmeet SinghGurmeet SinghppTB ` c $DTB ` c $DppTB ` c $D`TB ` c $D@@TB ` c $D` ` 0P` ` 0P` ` 0P` ` 0` ` 0` ` 0` ` 00` ` 00` ` 00` ` 0 h" ` s *"` 0 ` 0p8p j NProvisioned Resource( 2dB ` <D1@@ ` 04=  :AP 2 ` 0@p j  9S 2dB ` <D1 @ H ` 0޽h ? fff̙___PPT10i.80+D=' = @B +r  ? f   !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcefghijklmnopqruvwxyz{|}~t  PowerPoint Document( DocumentSummaryInformation8՜.+,D՜.+,  $ , BOn-screen Show$USC, Information Sciences Institute9 A  ArialVerdana Balloons1Resource Allocation and Scheduling for WorkflowsProblemPrevious WorkFormal Model Relation to previous workApproaches to AllocationSlide 7Preliminary WorkPreliminary WorkCurrent Research  Fonts UsedDesign Template Slide Titles 4 $%_~ Gurmeet SinghGurmeet Singh,