ࡱ> :<9}[@ 25bjbj44 t`ViVi*JJJJ$...P/l|/< bl/T1j1j1j1HHHaaaaaaa$wcReaIH^HIIaJJj1j1aLLLIJj1j1aLIaLLL6]_ j1/ .K.mJ^aa0 b^fKf8_JJJJf_HHL I IHHHaad<-L4<-The Design of General Matchmaker and its application to Access grid Han Gao Chuang Liu May 7th, 2003 Abstract Heterogeneous characteristics of network systems, operating platforms, hardware devices and other specific requirements need a nice matchmaker to select computational resources and network services for large scale distributed computing and advanced collaborative environment. In this paper, we propose a model of general matchmaker for resources selection and capability negotiation. We also discuss a general approach and algorithms for implementing this theoretical model. Finally, we exemplify it into the current popular collaborative tools, Access Grid, by a constraint description language Redline. 1. Introduction Accelerated by the development of network technologies, the current network applications always require various kinds of resources from distributed location. A matchmaker is needed to coordinate all of the resources and the accurate matching of these objects should be the first step to guarantee the whole procedure running smoothly. For advanced collaborative environment tools, it needs to be an open effort between individuals, unlimited by the technology that brings them together. Therefore, network services are introduced for a seamless integration. One open issue is that how to select network services for specific applications. The general matchmaker discussed in this paper provides some heuristic ideas to solve these problems: matching network resources and negotiating capabilities with or without network service. Consider the following scenario: A large NSF grant is shared by five universities distributed across the country. Some of the larger universities have multiple individuals working as part of the team, while the smaller universities have individual PIs. The team members have agreed that regular discussions are much more productive than yearly meetings and have decided to adopt the Access Grid as a platform to support this; each site have the funds available to purchase and construct such an environment. It turns out during the first meeting that the wide-area network connection of one of the smaller universities is unable to handle the amount of network traffic required to sustain an AG session. This scenario demonstrates the problem we are addressing in this paper: a collaboration partner who is unable to participate fully in an AG session due to technical limitations. We provide a strategy for this situation: The matchmaker looks for the least common resolution capability of AG. If there is no common resolution, the matchmaker introduces one or one set of network services into active group, which can produce a set of common capabilities and make group members interact with each other under AG. 2. General Matchmaker Architecture For efficiency and portability, we propose a general matcher for various kinds of capabilities, network services and resources (Figure 1). The schemas become a key component for each description file, since for different capabilities there are different parameters and conditions.  Figure 1: General Matchmaker: each red block represent a schema for each capability description files and rule set files. According the above discussion, the general matchmaker includes the following nice properties: Reusable It can match whatever the resources are: stream capability (audio, video), network services and resources. Portable It can be embedded into any resource matching or selecting architecture. Reconfigurable The schemas and core matching algorithms in general matchmaker can be rewritten without impact of other components. To illustrate the overall operations with other resources and services, we integrate these modules as Figure 2 showing. For example, in AG, at the start of a session, user end applications and information services configure the set of capabilities that represent the current resource, service and environment. When the use logs into AG virtual venue, those description files are provided to matchmaker. When network service is needed, register module and discover module can provide the information about the related services. The matchmaker examines the session bundle, iterating through the capabilities and offerings, searching for conflicts, then finally resolve the conflicts and find the best capability.  SHAPE \* MERGEFORMAT  Figure 2: Processing Architecture: each yellow line represents a set of APIs or UIs. 3. Mathematical Description and General Approach We propose a heuristic algorithm for resources matching. We believe the concepts of linear algebra can help us describe them clearly. A resolution vector of a resource object is represented as  EMBED Equation.3  where  EMBED Equation.3 are n independent parameters and  EMBED Equation.3  is an n dimensional vector space. For each capability set, we can compose at least one segment a set of vectors in this space. Suppose there are two vectors  EMBED Equation.3  need to match with each other via network service. The network service can be represented as a  EMBED Equation.3  transformation matrix  EMBED Equation.3  related with EMBED Equation.3 and EMBED Equation.3 :  EMBED Equation.3  Suppose we have two sets of resolutions:  EMBED Equation.3 , EMBED Equation.3 . Our goal is obtaining the best common resolution for each set and, if network services are needed, the corresponding transformation matrix:  EMBED Equation.3  Where  EMBED Equation.3 is the common resolution set,  EMBED Equation.3 is the common resolution set after transformation and  EMBED Equation.3  Consider multiple sets situation. Suppose we have  EMBED Equation.3  sets of capabilities,  EMBED Equation.3  We obtain independent capability sets first:  EMBED Equation.3  As above discussion, the goal is the best common solution vector and the transformation matrices:  EMBED Equation.3  where  EMBED Equation.3 . We need define the function argmax details under specific conditions. 4. Implementation Details for AG We use above scenario as an example for audio stream matchmaking: There are five users. The capability descriptions of audio stream are described as following: The first four users are the same: A=B=C=D=[(32KHz, 256KB, Mono), (16KHz, 128KB, Mono), (8KHz, 64KB, Stereo)]; while the fifth user has the only difference as E=[(16KHz, 128KB, Mono), (8KHz, 64KB, Stereo)]. The solution should be the best common resolution and the network services (if necessary). 4.1 Schema Design As we discussed above, we should define schemas for our capability description files. The goal is that we need sufficient but not redundant information to describe a capability. We need find independent parameters to compose each resolution vectors and their resolution space. For example, we find that the sample rate (frequency), bit rate (bandwidth) and mode (channel) are the most important and independent parameters for audio stream. Also, we add one item preference in the description files, since we should consider the personal choice for each user. Currently, we just manually configure the capabilities. In future, this process may involve loading a previously created configuration, using a capability discovery tool to automatically probe the local machine and network environment or some combination of all of theses. 4.2 Redline Constraint Language Chuang, could you add some features about your language. 4.2 Implementation without Network Service In general approach, the simple case is that the common resolution set is not empty. We define the specific argmax function for each parameter and find the best capability for all members in the same session. The algorithm should look like the following: Algorithm without network service Prior the parameters and build a prior list. Define the argmax function for each parameter. Screen out the common capability set according to the priory parameter list. Find the best common capability according to the argmax function for each parameter. For example, we priory three audio parameters like this: Sample Rate, Bit Rate, Mode. We define the argmax function of both Sample Rate and Mode: the higher the better; Bit Rate: the lower the better. Following these rules and priority, we can find the best resolution of this session is (16KHz, 128KB, Mono). Due to the advantage of Redline, the logic of this algorithm is already embedded in it. 4.4 Implementation with Network Service We make a small change for the audio stream example: A=B=C=D=[(32KHz, 256KB, Mono), (16KHz, 128KB, Stereo)]; E=[(8KHz, 64KB, Mono)]; NetworkService1=[(32KHz, 256KB, Mono)((8KHz, 64KB, Mono)]; NetworkService2=[(16KHz, 128KB, Stereo) ((8KHz, 64KB, Mono)]. It comes out that the common capability set is empty. In this case, network services are introduced into this session. The core algorithm is a table tree building procedure (Figure 3). In the parent table, we list all the possible capabilities in this group. For each capability tree node, the first cell records the minimum depth of this tree. If the minimum depth is not zero, it means there is at least one user has no current capability. The value of the depth is equal to the number of network services we need. The second cell points their children table node. After building this table tree, we begin to search the zero depth table branch or the minimum depth table branch. In this example, although the minimum depths of all the capabilities are the same, due to the priority and argmax functions of each parameter. We choose the highest sample rate table tree (32K, 256K, M). Since the depth records from (32K, 256K, M) to (8K, 64K, M) and vice versa, also, we have NetworkService1 match this situation. The final solution we obtain is the best possible capability (32KHz, 256KB, Mode) and the specific network service NetworkService1.  SHAPE \* MERGEFORMAT  Figure 3: Algorithm with Network Service: A table tree building procedure. 4.5 Reusability Reusability is one nice feature of our matchmaker. As we discussed above, we implement the matching algorithm in the matchmaker block (Figure 1). The matching criterion and conditions are embedded in the rule set input. Therefore, we can change rules and conditions by rule set input for various types of matching. For example, we can input the parameters and attributes of audio type, argmax functions and priority of capability parameters to make specific audio stream matching. For video stream match, we can change the rule set to video matching criterion without change any core algorithm in matchmaker block. PAGE v 4 Rule Set Matcher Input Set Result User End Information Matchmaker Register Discover (32K, 256K, M) (16K, 128K, S) (8K, 64K, M) 1 1 1 (16K, 128K, S) (8K, 64K, M) 1 0 ------ (8K, 64K, M) 0 ---- (32K, 256K, S) (8K, 64K, M) 1 0 ------ (16K, 128K, S) 0 ---- (32K, 256K, S) (16K, 128K, M) 0 0 ---- ----- CDE\]bdk{ R s _ i * 2 4 F c K y {  m {  89û÷hZlhhsRh`hte[h!G h$xhhhtMhzh{[h|V h5hh?h?6h?hhqhhT;hcHhV-hV-H*hV-h#hbBh#hbB5h#h#52DE]kl :t;t;Y t;t;$ !dha$gd dhgd/$a$gd#7gdV-8gd#8M$ @]^a$gd#2359CE[Zei:;<DFSVZk=@dƾƮƨ同ƐhKh!Uho<hS@hc5 h-5CJ h?@qvx/0CDExtlUHj%h9h"?EHU-j:B h"?@CJOJQJUVnHtHjh"?Uhj"h`h`EHU-jdB h`@CJOJQJUVnHtHjh`Uh"?h`jh h EHU-jDdB h @CJOJQJUVnHtHjh h EHU-jdB h @CJOJQJUVnHtHjh Uh EF]^qrstuF G b c v w x y 󣛣w`Sj0h"h1EHU-jdB h1@CJOJQJUVnHtHj-h h1EHU-jGdB h1@CJOJQJUVnHtHjh1Uh1j\*hzh"?EHU-j:B h"?@CJOJQJUVnHtHh`j'h9h"?EHU-j:B h"?@CJOJQJUVnHtHh"?jh"?U !!D!!!"A"o"p"q"""""6#N#Y#Z#l##$ $$$$$%%%%&)&0&D&E&F&~&&&&''(';'>'M'ȿȿȳhCvh6Ch6C6h8?hEhEhE5 hE56 hE5h6Ch;h9hMk h145h14hph8; h6C5 h8;5 h45 h|f-5h|f-h@rh6Ch4W[6h4W[h%6"$&%&E&F&&&&&''''(=((((*n*o***P,0 !% gd! & F !% gdCv !% h^hgdCv !% gd1M'|'''((((((((((())D)J)**.*5*@*H*l*m*n*o****+++B+C++++++++++ , ,#,%,/,B,M,N,O,P,--ܻǷǯǝǕh=Y ho hhlg hNWhNW jhNWhNWh:Eh! hO5 hphphO hp6h6Chp6hph6Ch! hCvhCvh6ChCv6 hCv5hCvhE9---R...000)0*0+0,0-0.0/090Y0y0z0{0|00000{1122222222222wmgmgmg hZl0JVjhZl0JVUhhaJ h6aJ h%aJ haJ h5aJ h5aJ ho aJ hpCJ ho 5CJ ho CJ ho ho h6Cho j\4h7 Ujh7 UmHnHtHuh7 jh7 Uh6Ch=Y 6hh=Y ho &0.0/0z0{0|00022333333333333)3*33343gd3J3K3U3V3W3_3`3a3i3j3k333333(4K4L4}4~444444/505|thbCJaJhga@hga@hga@CJaJh:Fh:Fhga@CJaJhga@CJaJhLh7 h7 CJaJh:FCJaJh'CJaJh7 CJaJ h"5CJh"h"h"5CJh3J3K3V3W3`3a3j3k3333344L4M4~4444444444gdbgdb(4440515@5A5L5N5[5\555555555gdo gdb0515Z5[5\555555555555hhaJh9h:FhbCJaJhNWCJaJhbhga@hbCJaJhbCJaJhga@800P/R :p#. A!@ "{#p$p%8 00P/R :p. A!@ "{#p$p%Dd |b  c $A? ?3"`?2M}u D`!M}u  ` &0xcdd``> @c112BYL%bpu?Yh 2! YPP& >@ArWpenR~CP` W7.CwsN}jo G  J:zzz:p50уa $5XAFA|C=\aZQRϥJW1H |h``E#RpeqIj.HA " 1 1 @c112BYL%bpudK F&&\"CD1,İr!v120eDLuDd hb  c $A? ?3"`?2X^r -N2w]3`!X^r -N2w]2`@ |axcdd`` @c112BYL%bpu  dQ%~Cl')'ZQR}-8a{M]\v>4m 0y{@iI)$5$ŠQ υv<^'f,?Dd b  c $A? ?3"`?2f#D p&B `!:#D p&@` !xcdd`` @c112BYL%bpuea;)Dd Tb  c $A? ?3"`?2sk Vt($aO `!Gk Vt($a XJxcdd`` @c112BYL%bpu4L=`dbR ,.Iq1(2tB ;/G3X?^)Dd Tb   c $  !"#$%&'()*+,-./02345678h;>d?@BACEDFHGIKJLNMOPQRSTUVWYXZ[\]^`_abcefg{|ijklmnopqrstuvwxyzRoot Entryi F, K=@&Data 14WordDocumentht`ObjectPoolk0K, K_1111004814F0K0KOle CompObjfObjInfo $).14789:;<=@CFIJKLMNORW\_`abcdglopqtwxy| FMicrosoft Equation 3.0 DS Equation Equation.39qA€0,i 2v =(v 1 ,v 2 ,...,v n )"R nEquation Native _1111042779' F0K0KOle CompObj f FMicrosoft Equation 3.0 DS Equation Equation.39q V v 1 ,v 2 ,...,v n FMicrosoft Equation 3.0 DS Equation Equation.39qObjInfo Equation Native  r_1111005341F0K0KOle  CompObjfObjInfoEquation Native 6_1111005700F0K.KA|l R n FMicrosoft Equation 3.0 DS Equation Equation.39qAN |l 2v 1 ,2v 2 "R nOle CompObjfObjInfoEquation Native j_1111144629F.K.KOle CompObjfObjInfo FMicrosoft Equation 3.0 DS Equation Equation.39qC4~ nn FMicrosoft Equation 3.0 DS Equation Equation.39qEquation Native 1_1111144507 TF.K.KOle CompObj fObjInfo! Equation Native !)_1111043570$F.K.KOle "C  T FMicrosoft Equation 3.0 DS Equation Equation.39q  2v, 1CompObj#%#fObjInfo&%Equation Native &:_1111043669",)F.K.KOle 'CompObj(*(fObjInfo+*Equation Native +: FMicrosoft Equation 3.0 DS Equation Equation.39q h | 2v 2 FMicrosoft Equation 3.0 DS Equation Equation.39q_11111443486.F.K.KOle ,CompObj-/-fObjInfo0/Equation Native 0X_1111145624O3F.K.KOle 2CompObj243fC<4h 2v 1 =T2v 2 FMicrosoft Equation 3.0 DS Equation Equation.39qC·@Tb S 1n =2v 1,1 ,2v 1,2ObjInfo55Equation Native 6_11110442568F.K.KOle > ,...,2v 1,i ,...{,2v 1,n },S 2n =2v 2,1 ,2v 2,2 ,...,2v 2,j ,...{,2v 2,m } FMicrosoft Equation 3.0 DS EqCompObj79?fObjInfo:AEquation Native Bl_1113830436Y=F.K.Kuation Equation.39q P $k S 1n ,S 2n "R n FMicrosoft Equation 3.0 DS Equation Equation.39qOle DCompObj<>EfObjInfo?GEquation Native HOH~ (2v * ,T * )=(argmax2v i 2v i "S * {},I),S * `"(argmax2v i 2v i "S ** {},T 12 ),S * =,S ** `"{_1113900295;JBF.K.KOle PCompObjACQfObjInfoDS FMicrosoft Equation 3.0 DS Equation Equation.39qB0lk S * FMicrosoft Equation 3.0 DS Equation Equation.39qEquation Native T6_1113900356GF.K.KOle UCompObjFHVfObjInfoIXEquation Native Y:_1113913011E^LF.K.KOle ZB S ** FMicrosoft Equation 3.0 DS Equation Equation.39qB†p,i T 1,2 =T(S 1n ,S 2n )CompObjKM[fObjInfoN]Equation Native ^_1111148214QF.K.K=T i,j  argmax{2v2v"S ** } ,      where 2v=2v 1,i =T i,j 2v 2,j FMicrosoft Equation 3.0 DS Equation Equation.39qOle eCompObjPRffObjInfoShEquation Native i)C 4h k FMicrosoft Equation 3.0 DS Equation Equation.39qC S 1n ,S 2n ,...,S kn _11111484201@VF.K.KOle jCompObjUWkfObjInfoXmEquation Native n_1111149547[F.K.KOle rCompObjZ\sf, S in "R n , i=1,...,k. FMicrosoft Equation 3.0 DS Equation Equation.39qCp 2S  1n "2S  2n "..."2S  ObjInfo]uEquation Native v_1113913415c`F.K.KOle zln ,                   ld"k FMicrosoft Equation 3.0 DS Equation Equation.39qBT| (2v * ,T * )=(argmaxCompObj_a{fObjInfob}Equation Native ~_1113913522eF.K.K2v i 2v i "S * {},),S * `"(argmax2v i 2v i "S ** {},T * ),S * =,S ** `"{ FMicrosoft Equation 3.0 DS EqOle CompObjdffObjInfogEquation Native uation Equation.39qBc( T *= {T 1,2 ,T 1,3 ,...,T i,j ,...,T l"1,l } , T i,j =T(2S  in ,2S  jn ),i,j=1,...,lA? ?3"`?2sKZl/rO`!GKZl/r XJxcdd`` @c112BYL%bpu tV6aqyi靋M޷T 9kR,4Pb~_4gӛKaDd hb   c $A ? ?3"`? 2emy!R=`!emy!R=@ |Mxcdd``.ed``baV d,FYzP1n:&v! KA?H1 ߁qC0&dT20ͤ `[YB2sSRs=^~*g`0``*F\_tŲ|cFb_?c.` 2!Ob|^P Ma(w ڇ˟{:8[~ nF0O xBwpTrAS8倍 v0o(u021)WĸADs!Dd Lb   c $A ? ?3"`? 2!)ɻuD%5]|U`!)ɻuD%5]|/L xڥkAߛݍͦiZKQF Tb!`Jid [tQhPD ^<SA T[Y<`̴Уx)yLijm*5r_Sy90Pџ[tqɩ(ǃr=U9k>+[6T󜕟khk8ٯᠧ1![j52Z[^KAǗY|ֽǻ˜%QoX>asg``%=VTT2Ǔ^מ<>0s7eI6~N՞$ҾJǮ{q)JJ> T;"c5Gx 95f9fIXbt7 79XP_ Dd @b  c $A ? ?3"`? 2T L~؛t~0,`!( L~؛t~H xcdd``> @c112BYL%bpu!.EMg-:Z4O"И0Sm̕kh8G(~nY&V- i m4M;By0Xhu<\V]'6v{8\EoٛW*c{XGzy=g\\ʛwAs ?p*2֪+[6?T+a ^)7n$H/  ^S K?Ol?ޜ#ut|auOZ?k6Gܐ=aglO/7*[=i({h y:>"x^ޕw8iV4*NZ<ڥUN A qBK˪Uł7zWϯNk$6˂gDKbé5ak'[/^τ 2WDd b  c $A? ?3"`?2>pv}:-Xջ%`!pv}:-Xջ:@Hxcdd`` @c112BYL%bpuC9'qrrr|V1{ I0{ ;Gl%}k,<=S.iGCy)գKllA>~/%{mC;M\aT3YoDaŪZ3xyalDd b  c $A? ?3"`?2^+_9p"O41`!^+_9p"O6 02XxڥkA&1٦vٶ2VZlm xB=D I֠Bx̓x?HЋ/^NRd3y|w dq _ zGsH!,]Tb08yok & rZ,aMw1jVvⱅz;]k@f8͟szXYkEp\wK[zCnV/iG*S'! [d,^;48(`{#|u$^Fq=7\Ό|QW?{vyA~}8@mqܖ|va;-B| I7T:uhk9z /o0a(W9r1=_J O+Q=b㲾̀o_'9N ߅Y}ӟOe5sjg\}D1~au ]{^6tZiVbLp^z9j9 %DdD  3 @@"?1TablefSummaryInformation(j~DocumentSummaryInformation8CompObjj՜.+,D՜.+,4 hp|  LCM*{ Thesis Title0lt|UseDefaultLanguageVersionLCID F7w   FMicrosoft Word Document MSWoOh+'0  , 8 DPX`hpThesishesHan Gaoan an ThesisHan Gao57 Microsoft Word 10.0@:ˍ@@>rKi$G Rtg  | h  ]a."Systema(r0P a-@Garamond- )2 'ahTHE DESIGN OF GENERAGXIPI5*KVQ@KIVIHD2 J ahL MATCHMAKER B^DGFX^DEIHha')2 'ahAND ITS APPLICATION DVP*G5D??B*FDG)QV 2  ahTO ACCESS GRIDGQDFFI55KH*P 2 5ah 8ha'@Garamond- 2 y ah ,ha'.2 ahHan Gao Chuang LiuK(2L(2?20(2,80 2 R ah -ha'-2 r| ahMay 7nR().@Garamond-2 He ahth-2 r ah, 2003.... 2 rw ah -ha' 2  y ah ,ha'-"2 ah 2 : ahABSTRACTAC2DEACD 2 A ah 8ha'-2  ahHeterogen K)) 2,)2 2 Bahe)g2 k=ahous characteristics of network systems, operating platforms, 21#C)2( ()) #)#C3C2)B2!.C#)$)M#C22) (2,C2(3 M#ha'42 n ahhardware devices and other r2( 1B( )1).))#(3122) 22 n ahspecific requirements need$2))) )01 )L)2#2))1,2 n ah a nice matchmaker to (2))L()2L(.) 3ha'2  ahselect#)))\2  6ah computational resources and network services for larg)2L20)22( *#21 ))#(212)B3 .#) .))#2 ) ,)2  ahe scale distributed )#)()1$ 20)2ha'2  ahcomputing and )2L202,?(21F2 'ah advanced collaborative environment. In?(1.(2)*1?)2(22 (.)?)2.!23L)2?"2#2 ah this paper, we ?2#?2(2* ?B(ha'"2 hahpropose a model 2 223#)(L21)h2 h/>ah of general matchmaker for resources selection and capability 2,)2) (L()2L(.) 2  *#21 ))##)))22(21)(2(2)ha'2  ahnegotiati 2),2(b2 :ahon. We also discuss a general approach and algorithms for 22IX)I(#2I1#)0$#I(I,)2) (I(22 2))2I(21I(,2 2M#I3 ha'C2 %ahimplementing this theoretical model. IL2)L)22,2#2)3 ))(L21)2 N ahFinally, w72()B;2  ahe exemplify it into the current ()-(L2)222))0 )3ha'-22 bahpopular collaborative tool2220( )2(22 ).)22[2 b5ahs, Access Grid, by a constraint description language g$C)))##L 12)()22#!(21)#) 222(2,0(,)ha'@Garamond-2 ahRedlineT>(*-2 ah. ha' 2 )ah , 2 ah , 2 zah , 2 #ah , 2 ah , 2 tah , 2 ah , 2 ah , 2 oah ,-hhaagg``ff__ee^^dd]]cc\\bb[[aaZZ``YY _ _X X  ^ ^W W  ] ]V V  \ \U U  [ [T T ZZSSYYRRXXQQWWPPVVOOWV`V Normal$ !a$ @CJOJQJ_HmH sH tH H@H Heading 1$$dh@&a$5H@H Heading 2$$dh@&a$5KHL@L Heading 3$$dh@&a$ 6@KHL@L Heading 4$$dh@&a$ 6@KHL@L Heading 5$$dh@&a$ 6@KHN@N Heading 6$$xP@&a$ :@KHH@H Heading 7$P<@& ;@KHD@D Heading 8$$dh@&a$KHD @D Heading 9 $$dh@&a$KHDA@D Default Paragraph FontViV  Table Normal :V 44 la (k@(No List .X@. Emphasis6]8B@8 Body Text dhL L Index 1 $ x 0^`0a$CJL "L Index 2 $ x 0^`0a$CJJ 2J Index 3 x 80^8`0@CJN BN Index 4 x 80^8`0 6@CJNRN Index 5 x 80^8`0 6@CJ0b0 TOC 1  ! 8r8 TOC 2 ! h^h66 TOC 3 ! ^66 TOC 4 ! 8^866 TOC 5 ! ^VV  Comment Text$ Ed$x^`E:@: Header$$ a$: @: Footer$$ a$F!F  Index Heading $CJKHF"F Caption0]^@CJT#T Table of Figures ! 0^`0Z+Z  Endnote Text$! Ed$x^`ECJ\,"\ Table of Authorities" ! h^h`8-28  Macro Text#OJQJJ."J  TOA Heading$$$$a$5KHF/@RF List$% dhP^`R1@QbR List Number"& ! ^`:2@Qr: List 2' 88^8:3@Q: List 3( ^:4@Q: List 4) ^:5@Q: List 5* pp^p>:@a> List Number 2 +h^h>;@a> List Number 3 ,^><@a> List Number 4 -8^8>=@a> List Number 5 .^nJ@an Subtitle)/$$d@& ]^5CJOJQJ\^JaJ\C@\ Body Text Indent"0$hd^h`a$BD@QB List Continue 1 BE@"B List Continue 2 28^8BF@2B List Continue 3 3^BG@BB List Continue 4 4^BH@RB List Continue 5 5p^pL>@qL Title"6$d]^a$5;.L@r. Date 7$0a$8O8 Author8$da$\O\ Block Quotation9$d]^6<O< Body Text Keep:$ROR Chapter Label;$$$0a$ 6@FCJTOT Chapter Subtitle<$$$a$@KHVOV Chapter Title=$$$00a$ ;@KH2O2 Footer Even><O< Footer First ? !8O8 Footer Odd @ XOX Footnote Base A Ed$^`ECJHO"H Glossary DefinitionBdDO2D Header BaseC$$ a$2OB2 Header EvenD<OR< Header First E !8Ob8 Header Odd F NON Heading BaseG$$$dha$5KHPOP Index BaseH x 0^`0@CJ*O* NameI$a$:O: PictureJ$$da$NOqN Section LabelK$$a$ 5;@ VOV Subtitle CoverL$$0]^a$`Oq` Title Cover&M$d ]^a$ 5;@4O4 TOC Base N ! @&@ Footnote ReferenceH*>'> Comment ReferenceCJ>*> Endnote ReferenceH*:O!: Glossary Entry5\@O1@ Lead-in Emphasis;@2OA2 SuperscriptH*:R:  Footnote TextUx.)@a. Page Number ",6COYcw)T-%,4?CM ",6COYcw)T     -`:`-DE]kl:    9 : <  e,-.JvGz$%EF = "n"o"""P$(.(/(z({(|(((**++++++++++++)+*+3+4+=+>+J+K+V+W+`+a+j+k+++++,,L,M,~,,,,,,,,,,,,0-1-@-A-L-N-[-\--------M080807080/0000000000000000 0 0000000000 0 0 000h000000000000000000000000000000000000 0 0 0 00000 00 0 0 000 0 0 0(00(@0@0@0@0@0@0@0@00@ 0H00000000H0000000000@0000@0@0@0@000@0@0@0@000@0@0@0@000@0@0@0@000@00@000"9 : < "n"o"""P$->{0 0>{0 0>{0 0@0P@0P>;00h<;00<;000`0`0` @0e 9,E M'-2055 "#$%&')*,/J"04345!(+-.5/GH+-EY[i} !57 )=?/CE]qsbvx()(,(-_::::::::::::::::::::_ !8PW@:VL(  $  R $%5 $#  s"*?` # c $X99?R $%5T , # , G+G/  T 4 # 4 15  T % # % $d)  TB ' C D *&d+&TB ( C D 'd'TB - C D ,G,TB 0 C D*G++/TB 1 C D$)TB 3 C D$)TB . C D 3.G4.TB 5 C D 22TB 7 C D 4 4ZB : S D> . 1T ? # ?3.("2  TB @ C D/("/TB A C D 3. 2TB B C D1(" 1T C # C[1x5 TB D C D[2x2TB E C D[ 4x 4ZB HB S Dx1P3T M # M'Y$`, TB N C Dt)Y$u)TB O C D<'=`,TB P C D*Y$*ZB T S D [(*G+ZB U S D[( 3.ZB V S D3[(*n  +2 # 3"n  c $X99?"` +2  0)X?"` ~.[/   0)X?"`~.3/   s *)X?"`+-   0)X?"`j1x2 f  s *)X? ~. /f   s *)X?+-f   s *)X?j12f2   s *)X?G-~.f   s *)X?[9//f2   s *)X?G/j1.b B+  #" n  c $X99?"`B+T  #   Tb  #    Z  3 #d% Tb  # )(GZ* Tb  # )(PZ*  `B  c $D>BI"J"`B  c $D>on'o'4b   Q"4b   Q"4b  Gn')4b  Q"x#4b   %xn'x2 ! <ZGHzeIo O)B S  ? !: .*(-vWN4| 4$t&),]-j-8*urn:schemas-microsoft-com:office:smarttagsdate 200357DayMonthYearILRXFLu}~~  D!J!&&***++++++---IL uwzF0!4!""&&''***++-333333333E\]{*  G%EF~o""**++++++++(+*+2+4+<+>+I+K+U+W+_+a+i+k+++K,M,},,,,/-1-Z-\---u}~~*++-Han Gao |.}:N-~,xz+ &^j"x0zfb&%48PRL_2MtJJbP\v[&^`.^`.88^8`.^`.hh^h`.h ^`hH.h ^`hH.h pLp^p`LhH.h @ @ ^@ `hH.h ^`hH.h L^`LhH.h ^`hH.h ^`hH.h PLP^P`LhH.h ^`hH.h ^`hH.h pLp^p`LhH.h @ @ ^@ `hH.h ^`hH.h L^`LhH.h ^`hH.h ^`hH.h PLP^P`LhH.h ^`hH.h ^`hH.h pLp^p`LhH.h @ @ ^@ `hH.h ^`hH.h L^`LhH.h ^`hH.h ^`hH.h PLP^P`LhH.h hh^h`hH.h 88^8`hH.h L^`LhH.h   ^ `hH.h   ^ `hH.h xLx^x`LhH.h HH^H`hH.h ^`hH.h L^`LhH.h^`OJQJo(hHh^`OJQJ^Jo(hHohpp^p`OJQJo(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJQJo(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJQJo(hHh 88^8`hH.h ^`hH.h  L ^ `LhH.h   ^ `hH.h xx^x`hH.h HLH^H`LhH.h ^`hH.h ^`hH.h L^`LhH. ~}|b&%L_Mt^x0P\v                                                        fe`=Y o lg|VDcH"!m%bB-c57 (%b(V-|f- 2149T;( =8?S@ga@6C:F!G&[K<QsRNWiUX}Y4W[te[MkwnKoQo@rCv*{Ki~tM w!qtE:E!U'O?;s8{[