ࡱ>   !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~RdO) PicturesPowerPoint Document(5SummaryInformation(`'  T5$5Gh @]>`xVKAhŢ h-Ҥ P?T -XQɹ9x(CEBH{($4}3;Ynxͼo&;2h~҆_􌸶PAobf쑁Q4`~g8TD_mA: Sjgo` Ckt"h#q5JCӠ9#+zAzoD; }(+(hvA7#,ٞ1wCB +IM`sԋD,YX5sGf6|aO ;ز [KT&(Mx"7Ijl}=s'&P_,3]߽F(j(E}ɩ/0;++8s.ȇ"#z5=y}]ߣ{:4u٧q1lq 8y8iBz',3eeވ֧k=85 NuS~GSGST6=S >~]_uw)0}q "OE#rFf{j%Nzylz5I/.٫"_ȾԹ)!ϊ5-IV76ekb%M/eRw;vW3znkm-\Fmp6A(۪V<WmyF]ube$udWw,_College of ComputingE\mhyV5)3((   d# Chart MSGraph.Chart.800Microsoft Graph 98 Chart/0(  0;[0 0 000$([\{b00 000000000  0=] 0 0 0000 2 3 !A0C0E0G0I0c00000000000000000!%),.:;?]}acdeghijklmnop0DTimesBlack6i6i6e u6eue,mDHelveticak6i6i6e u6eue,m DArialicak6i6i6e u6eue,m0DArial Black6i6i6e u6eue,m@  @@``   @n?" dd@  @@`` h`P)R U m   ,1X.)'/X$dhB$5$5Gdh `1? AQ@ g4BdBdudPf6e6uP0ppp@ ?`%September 1998 (Advisory Board VisitO =< Information`Spaces and `Spheres : Interacting with Useful InformationGGKarsten Schwan with Calton Pu Mustaq Ahamad, Richard Fujimoto Ling Liu, Ken Mackenzie Kishore Ramachandran Sudha Yalamanchili (Ann Chervenak, Raja Das) 7001 Talk - Oct. 1999 schwan@cc.gatech.edu http://www.cc.gatech.edu/fac/karsten.schwan D@+     %         High Performance and scalability in terms of: number of sensor/output clients selected transformations, ie.computations for data interpretation and display data rates, real-time behavior like end-to-end delay Increased volatility of hardware infrastructure (e.g., consider wireless devices) Quality of Service: focused on real-time behavior: Sensor/output devices dyn. vary in rate/data needs, modes Different data characteristics/modes imply comp. changes heterogeneity is an assumption mobility and service variations are problems l..S!,g    Faculty: Karsten Schwan (Info Spaces, Middleware,QoS/Adaptation, ASAN) Calton Pu (InfoSphere, InfoPipes, Information Flows, Synthetix/Microfeedback) Kishore Ramachandran (Stampede, Stream-based Computing), (Ann Chervenak- Storage) Mustaq Ahamad (Scalable Services, Security) Ling Liu (Continual Queries, Xwrap, Transactional Workflow) Richard Fujimoto (Simulation Technologies) Ken MacKenzie (ASAN, Hardware Stream Switching, & ) Sudha Yalamanchili (ASAN) Research Staff: Greg Eisenhauer, Beth Plale, Rajamar Krishnamurthy, & Graduate Students, PostDocs, Visitors, Other Institutions: too many to list individually 6Y6;L       *  <I &  > Research Program (Some Detail): Constructing and Managing `Spaces CB`Construction and Management: Support Middleware: Objects and Events --> OS enhancements: CPU and Packet Scheduling --> Network-level support: NI Extension --> Hardware implications: Configurable hardware Management techniques and algorithms Domain- and application-specific: Exploiting consistency information for interactive and perceptual applications (object caching, optimization of communications using temporal/spatial info. - Ahamad) Exploiting application semantics (database-derived techniques for query optimization, filter creation and movement) DWCS to schedule CPU actions and info. transfers &%"M *& %" ,&   "  '!1Event/Messaging Middleware - Online Configuration22 I/O + event streams: creation, control, load migration Packet Scheduling Dealing with heterogeneity: Interpreted data conversion vs. conversion by dynamic code generation: $Ib*("  &Research Program (cont.)4 Management techniques and algorithms Exploiting consistency information for interactive and perceptual applications (object caching, optimization of communications using temporal/spatial info.) (West) Exploiting application semantics (database-derived techniques for query optimization, filter creation and movement) (Plale) DWCS schedules CPU and Comm. Device (West) `Pushing events, qual. mgt., policies into the OS Kernel (Linux), the wireless infrastructure (Yamacraw), and the network (ASAN) (Kravets, Poellabauer, Krishnamurthy) &   *$ /@>    ) Research Program (cont.) `Pushing Functionality -- An Example from the ASAN (Active System Area Networks) Project: Reduce communication latency and increase bandwidth by specializing communication functionality and by selectively assigning it to host CPUs vs. network interfaces@\4s[4s   *#  P ` ̙33` ` ff3333f` 333MMM` f` f` 3>?" dd@v? " dd@  " @ `"  n?" dd@   @@``PR    @ ` ` p>> |(     `wawa1 ?P i  T Click to edit Master title style! !@  Zwawa1 ? i  RClick to edit Master text styles Second level Third level Fourth level Fifth level!     SB  s *޽h ? a( &Microsoft Office 98 ZR@(   @  Ztwawa1 ? @ i  RClick to edit Master text styles Second level Third level Fourth level Fifth level!     Sp  01 ?   i B  s *޽h ? a( P 0(   B  s *޽h ? a(  `X0( a    ߬wawa1 ?$\  i    # l4wawa1 ?@@@  i  H  0޽h ? a( 4 D<`( X\PUT   T4wawa1?, B Motivation    Twawa1?0 JEvolution of Systems Research: Efficient programs on uni/multi/cluster processor machines Distributed -> Internet Services Interactive and collaborative services Future Technology Needs: Key fact: The focus is not the computational tool, or the specific information source, or the computer. Instead, it is tto make information available to users, in the right form, at the right time, in the appropriate place. Information `Spaces and  Spheres : Composed of Information Sources, Computational Tools, and User Interfaces `Spaces are stream/graph-based computations linking sources to tools, tools to users, users to users        $! w  9 *    $w9,6:H  0޽h ? a(h y pT(  T T TTwawa1?, NMotivation - Continued~ T Twawa1?0 Sample `Spaces Old: Server-based media delivery and shared manipulation File systems, juke boxes Future: Scientific (Collaboration among scientists --workbenches) Focus on info. generation, coupled with I/O streams Remote sensors, visualization, and collaboration High performance is often critical Future: Perceptual (Interactions among human users -- smart home, enhanced humans) Driven by information, captured via sensors, stored and accessed, generated/discovered Computations/Tools are entirely hidden Service quality plays key role, in part due to merge of physical and virtual worlds :   *D   *S   *:/ o@H T 0޽h ? a(j  (  $)p   Twawa1?$, NMotivation - Continued  TTwawa1?P0p Sample `Spheres Common Users (Driving to a friend) `Civilizing the information jungle: focus on finding the `right information (e.g., directions, road conditions), deliver it with suitable `freshness (find out about construction before getting stuck) Highly diverse and dynamically changing information contents and sources Continuous information access, processing, integration, and delivery Quality is typically critical: freshness, assurance, security Interest Groups (Playing a distributed game) Rapid information update, delivery, access Continuous and dynamic change Highly diverse devices ( $   *.  a *" ! NaH  0޽h ? a(?U  TTil`T( ^0 `v ` N1? 0 ` T1? 0  B< v2 ` N1?@v ` N1?0@v ` N1?0`v  ` N1?`Pv  ` N1?`v  ` N1?p`v  ` N1?@Pv  ` N1?@v ` N1?p@|B ` TD1?`@`|B ` TD1? @` |B ` TD1?@`|B ` TD1?@|B ` TD1?`@` |B ` TD1? @`|B ` TD1?@`|B ` TD1?@`|B ` TD1?`@` |B ` TD1? @`|B ` TD1?`@`|B ` TD1? @`|B ` TD1? @`|B ` TD1?`@`|B ` TD1?`@`p|B ` TD1?`@`|B ` TD1?|B  ` TD1? !` Nt1?6   Ytransport model. "` NԿ1?@  ;...< #` N1?   [chemical model0v2 $` N1?p v2 %` N1? p | &` T1?P '` T1?` @  B  v (` N1?| )` T1? @@ | *` T1?P P  +` c 0e0e    BCwDEdF. @ 1 8c8c     ?1 d0u0@Ty2 NP'p<'pA)BCD|E||?2!cAsHX/wsccz,x d#X!7 @        ` @7  ,` c 0e0e    B!CDEXF* @ 1 8c8c     ?1 d0u0@Ty2 NP'p<'pA)BCD|E||7-*)E doy)z !80!lW7@       `   -` c |0e0e    BqCDE4F @ 1 8c8c     ?1 d0u0@Ty2 NP'p<'pA)BCD|E|| $EpXitq\iFb/ZR $ $ @    ` q  .` c 0e0e    BCDEdF. @ 1 8c8c     ?1 d0u0@Ty2 NP'p<'pA)BCD|E||o]H8 -,$",3 37i~8J_oywo{dN!!ho"o@        `   /` c |0e0e    BCCDE4F @ 1 8c8c     ?1 d0u0@Ty2 NP'p<'pA)BCD|E||  ".C CC." @    `   0` Z 1?   QVisAD 3d Active Interface |B 1` TD1?|B 2` TD1? 3` c 0e0e    BCDE@F  @  1 8c8c     ?1 d0u0@Ty2 NP'p<'pA)BCD|E||p @`pHH(P x@@p@@ @     x 4` Zt¼ 1?| O2d applet visualization v2 5` N1?0v2 6` N1?0`v2 7` N1?0Pv2 8` N1? Pv2 9` N1?p |B :` TD1?``|B ;`@ TD1? |B <` TD1?`@|B =` TD1?`  >` Z4ü 1?Y" I3d vis & steering v2 ?` N1? P ` v2 @` N1? p |B A`@ TD1?@  @ |B B`@ TD1?@  p |B C` TD1?  @  D` Zü 1? p M  E Active Filter B E` ZD1?P@ B F` ZD1?P@ B G` ZD1? P@ B H` ZD1? p  B I`@ ZD1?p p v2 J` N1?@ v2 K` N1? p @|B L` TD1?@ |B M` TD1?@p p  N` c Z0e0e    BC0DEF @  1 8c8c     ?1 d0u0@Ty2 NP'p<'pA)BCD|E||T0080@   p  O` Zļ 1?  LResidual Circulation B P` ZD1? r Q` c J0e0e    BCDE F @  1 8c8c     ?1 d0u0@Ty2 NP'p<'pA)BCD|E||@ p|B R` TD1? @B S` ZD1? v T` N1? v2 U` N1?P0 v2 V` N1?0 P W` Ztż 1?0 S  E Spectral2Grid  X` Z4Ƽ 1? S - E Spectral2Grid v Y` N1? v Z` N1?@ pv [` N1?p v \` N1? v ]` N1? @vr ^` N1?  |r _` T#1? |B `` TD1?@P@B a`  `D)?0   b` NƼ1?P0  m!historical chemical model results0"!B c` ZD1? ` B d` ZD1? ` B e` ZD1?` B f` ZD1?` |B g`@ TD1?P @B h`@ ZD1?@  i` NǼ1? 1 iheterogeneous client interfaces.  j` Ntȼ1? ` ( p&information transformation and routing.'& l` 64ɼ1? FScientific SpaceH ` 0޽h ? a(  h<( lq@Y@ h h Zɼwawa1 ? Online Visualization and Collaboration tasks require high performance for generation, transformation, and display: scalability in terms of data size and complexity input/output define additional streams Inability to control hardware infrastructure Quality of Service: Heterogeneity and dynamic user needs: Typical `workbench machines range from home PC, to low cost video wall, to Immersadesk, to display wall User behavior varies from supervision, to browsing, to focused inspection/manipulationlsXkJ X1( i h TDwawa1?X Z"Interactive Scientific Application##H h 0޽h ? a(    | P (  P  P Zwawa1 ? hX E`Interest Group` Sphere: From Sensor Processing Pipelines to Games/VEFE^2 P 61?0X ^ P 61?D,^ P 61?D|, ^ P 61?@ h(   P Tdwawa1?pU  B Shared Space  ^  P 61?`H^  P 61?Dl, ^  P 61?    P Zģwawa1? b,end users: strongly connected high fidelity-- P Z$wawa1?   a+end users: weakly connected lower fidelity,,p P HG/*H 1?0 hH p P@ HZG/*H 1? ( P Zwawa1?0   o-active objects: displays, sensors, robots (?)&.-^ P 61? ^2 P 61? 0 x  P ZDwawa1?  R  Cpassive objects P Nwawa1? p DQoS = Hard and Soft Deadlines and Transitory Performance Dynamic changes in user needs and behaviors, external and computing environment, and in program behavior: Image Tracking -> Game examplesL98wpB P HD1?p`PpB P HD1?pH P 0޽h ? a(  NFd(    dl d C Pp i  2 d Zdwawa1 ?\ TInfoSpheres with Human Participants: Parallel/Distributed/Net Real-time ApplicationsUT JH d 0޽h ? a(   +pY( a  pd p <1?@`d p <1??_d p <1?0`d p <1?hcd p <1?@d  p <1?0  p Z$wawa1?  N    p  `䧾wawap))?@x OOnline Quality Management  p  `Dwawap))?P h _)Cluster Computers, OS and Network Enhance**>  p Twawa1?0 X`Domain-specific solutions provide functionality of immediate use to specific applications, while also providing loads and benchmarks to middleware/OS/network researchers.$  p Ndwawap?  ]'Applications: Sample Spaces and Spheres((R p c *BCDEF1?"..j.P.6......r.X.>. ....9{PaPGE:.  i 5"....q.W.="""""""z"`"F"""..EEhEN\'\ s ''*4AAXAoNh9Sz #=d5Ov-T{1r)Cj :a%?f  D k     ! |     3 M g       * Q l     < c }      4NhuoAsugN\ E E E E 9} .c .< " " "   l D   t Z 3 H .    " " .x .Q .7 . " s Y?% |aG - "@` p T$wawa1?Gde U!Overview of Research and Approach""XB p@ 0?XB p 0?@0 p  `䪾wawap))?x \&Information Capture, Discovery, Access'' p  `wawap))?   rEcho, JECho Event Channels p  `dwawap))?pH  sContinual Queries and XWrap p  `$wawap))?pH  TDistributed and Mirror Objects   p  `wawap))?pH  }%Monitoring,Steering,Adaptation Mechs.&& !p  `wawap))?   MStampede Infrastructure "p  `twawap))?   UASAN: Active Sys. Area Networks  XB #p@ 0?PpXB $p 0? pXB %p 0?`  XB &p 0?0pXB 'p 0? XB (p@ 0?0 pXB )p@ 0?P P  XB *p@ 0?P P  +p  `4wawap))?@@  KSimulation TechnologyH p 0޽h ? a(     O ( i 5    wawa1 ?p  i  R  c *BCqDEF1?)66~6d6I6/666666i6O66666D_o_TQG6-t?)6666y6_6D))))))))d)J)"))66QQjQOm(m *(F(a5|BBOj"=W*Dl,,,?,Z,:GUUU:UabbbbBppppp"p<pWp~bbbbb7UQUyUGGGG?GYGGUUb b9 ba b b b b b bA b b b U U/ GV Gq G G : : : : :7 GQ Gy G U b b b> bf b b b U UGG9Ga,{|S8z{m9QQQ Q D 6 6f )> )$ )    k D )     ~ V /     i N 4   ) ) 6 6n 6T 6, )tY?$ y_D*)@```  TTwawa1?!4 MProject Participants (GT)H  0޽h ? a(  `X$( )s)r)q $ $  wawa1 ?  i   $ # l4wawa1 ?PP0  i  H $ 0޽h ? a(.  --QT\N-(  \F t X  \ t X fB \ 61?3/ X  \ Zwawa1?t I = Data Events F p K  \  _ fB \ 61?8p +   \ Zwawa1? K  @Control EventsR  \ c *BsCDEF1?ss mT;!4Pc}cd}S9* 6Jcx 0J_!y*1CS[t$@EZg;umzEv- 6Wpz`E,/ AAHOYalaJi0irrrrrixacOPA=8(pyciP6;4.'!!g6 fE@ sk   \ N4wawa1? B  DMiddleware: Mirror Objects Event Channels Derived Events Java EventsEE ;F T|  \   hf2  \ 61?T| \ NvB;vB;1?s [Display     \ TvB;vB;1?`  ^       \ TT vB;vB;1? ^      MF $ < \  f2 \ 61?$ < \ N wawa1?m ~  =Manager \ Tt wawa1?   LMonitoring Information  \ T4 wawa1?R   RDetection and Decision Model F 8  \   fB \ 61?% fB \ 61?_ ;  \ TTNwawa1? 8E  HData/Monitoring EventsF  ' \   %CfB \B 61?; IfB \ 61?,  l \ <1? 'F d \ df2 \ 61?df  \ 61?z r" !\ BG 1?% l" "\ <G 1?c ]  #\ TOwawa1?b   MComputational Instrument J N [  $\ [ N   %\  2N {  &\ { f '\ 61?f (\ 61?A { 2N    )\   f *\ 61?  f +\ 61?I  N %[  ,\ %[ 2N %[{  -\ %[{ f .\ 61?%[f /\ 61?%A [{ 2N % [  0\ % [ f 1\ 61?% [ f 2\ 61?%I [ `B 3\ 01?#9 `B 4\ 01?#  `B 5\ 01?# A `B 6\ 01?B`B 7\B 01?!A `B 8\ 01?#`B 9\ 01? ` 6` `B :\ 01?  6 `B ;\ 01? V 4V `B <\ 01? 6 `B =\ 01?. `B >\ 01?d 6g `B ?\ 01? Bb `B @\ 01? i = `B A\ 01? I` `B B\ 01? Bg `B C\ 01?B `B D\ 01?"b BY l2 E\ <1? f F\ 61?dA { f G\ 61?  f H\ 61?cA { r" I\ BG 1?% cf J\ 61? f K\ 61?$c]f L\ 61?  Q M\ TtOwawa1? CMiddleware (Eisenhauer): Using Objects/Events for Control/TransportBD  -$  . N\ T4Pwawa1?\M RQuality-Controlled Experimentsd O\ <G/*1? P\ TPwawa1?`  X QoS Contract   Q\ TQwawa1? 3M  GAncillary Instruments  R\ TtRwawa1?(c" JMultiple User Interfaces  S\ Z4Swawa1? ~  @Control EventsH \ 0޽h ? a(   0`(     TSwawa1? HMiddleware: Events: `Under the Hood %% d  <? ^  6? ` j  B))?  . y  HTT3f))?c :   # ljJ L" ?N  # ljJ L" ?0D  3 rjJ L" ?8%dB  <DocQjB  BDo @  <U@; J Event Source"   dB  <Do7Px dB  <Do7 UdB  <DoV`   HU1?pD hEvent Channell$    ! H4V1?p  GHandler$  " HV3f))?0   :  # HTW1? B[  UDerived Event Channel$  $ HX1?0 N  K Transformer$   ^ % 6? p j & B))?ppdB ' <Dop0`dB ( <DoP@ * HX1?0E QActive Event Sink$  + # ljJ L" ? ZE jB ,@ BD1?0 p jB - BD1?   . HY1? 0) [  NControl Events$ l / 0A ??0 p  0 <Y1? c - @Addressing Heterogeneity: Optimization using Spatial Consistency4A&H  0޽h ? a(:   &(z( i@d@    wawa1 ?I  i     wawa1 ?  i  XB  01?,,XB  01?0(  ZDwawa1?  Jmsec XB  01?0XB  01?%0%XB  01?e 0e XB  01? 0 XB  01? 0 XB  01?% 0% XB  01?e 0e XB  01? 0 XB  01?0  Twawa1?B 50  Zwawa1?[ : 0.05  Zdwawa1? v 70.1  Zĵwawa1?A o  50.2  Z$wawa1? o  50.3  Zwawa1?A o  50.4XB  01?||XB  01?\ \ XB  01?<<XB  01?\ \   ZDwawa1?O  Gmessage size in bytes  Zwawa1? 460  Zwawa1? 420XB  01?ExuXB  01?X EXB ! 01?` 8XB " 01?@XB # 01?e xXB $ 01? X e XB % 01? 0  XB & 01?e   ' Zĸwawa1?Q ?+ @DCG Conversion ( Zwawa1? o  ^Interp. conversion H  0޽h ? a(  0$&:(  d  <o? P X2  0oP0    <o0 O   J Dispatcher$  d  <o? Pp  HGAHo? 0X2  0o0 0`2  6 0   8gate2  6d  8sel12  6Ļ ` 8sel42  6$P p  8sel22  6  Pjn2  6 @ Ppj^B  6D  ^B  6D P ^B  6D ^B  6D` ^B  6D ^B  6D` ^B  6Dp0 `   < X Cnet library    6d P  M Event Handler$ dB  <Do 0 dB  <Do P P 0 dB  <Do P P dB  <Do   <ľD? 7Addressing Dynamics: Filtering based on data semantics.8"2  6$ p 8sel3^B  6D@ XB  0D jB ! BDo0P PjB " BDo  # < ` 8 ; operation  $ <   W cnet object V % Twawa1? RDQUOB (Plale): Event `Query OptimizationB* & BdP  *-- instantiates operation and cnet objects++,  H  0޽h ? a(  @,( (X@%q@    wawa1 ?  i     $wawa1 ? i  H  0޽h ? a(*  P(-2( X@dW    wawa1 ?  i     Dwawa1 ?@ i  XB  01?XB  01?d  <1?(pXB  01?IXB  01?2I2XB  01?r Ir XB  01? I XB  01? I XB  01?2 I2 XB  01?r Ir XB  01? I XB  01?I  Twawa1?O- ! 50  Zwawa1?   620  Zdwawa1?N   440  Zwawa1?   460  Z$wawa1?N   480XB  01?==XB  01?  XB  01?XB  01?    Zwawa1? Q FMessage Size (bytes)  ZDwawa1? 710000  Twawa1?p 64000d  <1?uI d ! <1? d " <1?   #@  `dwawa1?u rReceived bandwidth (MBits/sec)XB $ 01?-- % Zwawa1?w 62000XB & 01?} } 2 '  ENGHZJQ1? `T`T`T`T`T`T`T`T`T`T L2 (  ENGHZJQ1? `T`T`T`T`T`T`T`T`T`T|2 )  ENGHZJQ1? `T`T`T`T`T`T`T`T`T`T@| * Zwawa1?p B  Qthree senders - no flow control  + TDwawa1?; s  Nthree senders - flow control - Zwawa1?   /Experimentation with ATM Fore Interfaces (Rosu)$0/*H  0޽h ??0'() a( >D 0(`@(  @ @ Tdwawa1? HProject Status @ Twawa1?0P  0JOngoing: Construct sample`Spaces , with existing middleware infrastructure Focus on  hiding tools from end users Experiment with heterogeneous/diverse devices Map selected computations into network interfaces and develop extensible NI infrastructure Extend from Information Spaces to Information Spheres (include. Integration with `sphere software and approach) Future: Improve middleware: experiment with objects, extend to Java Broaden domains: wireless infrastructure, wide-area and high performance interactions (access grid) Many other topics  c  !    c,0 A H @ 0޽h ? a(  SK (    Ttwawa1? TSponsors and Collaborators  Twawa1?0P  GFunded Projects: Middleware (NSF, NCSA) Computational Resources (Intel, NSF RI) Quality Management, Wireless (DARPA, Yamacraw, and with Pu) ASAN (NSF, DOE) Future: Improve middleware: CORBA and Java objects Broaden domains: wireless infrastructure, wide-area and high performance interactions (access grid) Many other topics Collaborators: Many, incl. locally: Ahamad, Yalamanchili, Pu, Essa, Zegura   !   ! <  < Y "   H  0޽h ? a(xXKpE{f&!4$&@I*a-f5`aB ʺGV @yвr"UzUѣ2~&EO?&7[{k~4( H"m;㑝?U. |`s  E^lˏ[_{karS(E_JRK4N/8E/CdOn8Os~YmXOW5֐:hh1]< l6G@P46@l`=؋"1^RD9,93}<~iҩd:nǒSLR`V d4*_ь}hrx"a٭C>En?̴VzGڦ=r9dI+k]SV~tj LW}`3rn 766v <95[[b|DP (Qjmڎ%.*>I4m6=B-!cQ_TLZԙԟ 粻hbV`ϩ@Lc;9q.L`P~0- ׅaJ??$7NJeKNf jUL$kz'GYƝtƄݵȽO v{+FӏVGn]$7fSi. ^Y~DE3ݸq.]hyeE"B&0`(:9(2P$Ҍ0 9|2A"9"r E  QZ2@jPdzMMIˠuz$dL@.jTQ}TQl*H ǔ(~̱h@z  &(1=7=);v6͢O;;դvqP>w5e# up%T&1B ) o:??n{ifjcՒO*X &}ƽ($+uCյӭQl}H>]; Q!$?McF`,(X^yfaz0)Kmx\Ҿ(%)s(JoTMjl6N^u׷kp h-D22p]4)?ab^Ƀ#w~_SjW4 kqmr0.10;(g275 Jo =@N " |%`[F$S(">Vv"?*ET&&{00020420-0000-0000-C000-000000000046}r rTypeLibVersion2.0&&{0D452EE1-E08F-101A-852E-02608C4D0BB4}b b&&{47FF8FE1-6198 Oh+'00'$0< \h    'RComputational, Information, and Perceptual Spaces: Attaining Suitable PerformanceaompKarsten SchwanIarsarsarsCollege of Computingati24lMicrosoft PowerPoint 4.0n, @y C@sZ@zc<@jG\%PICT%T @@2 2HH I2݁݁݁݁݁݁**<âۢƢߢ,ޢ   <̢ۢ***آ      G!!!;      0Ǣ3www**݁݁݁݁݁݁݁݁݁݁݁݁݁݁Bݾݾݾݾ   ݁݁ݧݧ3? w w w3 ݬ ݬ ݁݁݁݁݁݁݁݁3cc ݵ ݵ Q ݁݁0݁݁݁݁݁݁Z   $݁݁N   ݁݁݁݁݁݁]    ݁݁i      <݁݁݁݁݁݁݁݁݁݁݁݁݁݁{         f݁݁݁݁ȹϵ*Hv) ?Z=[Үsze,.ѐC&'3&:(n\_FWI(Uŭxtr k5p`7DY__c ?T0b٦[ >#`ِǚ[&2 DocumentSummaryInformation8Current User4 ՜.+,D՜.+,d     'On-screen ShowtCollege of ComputingNoφ: Times HelveticaArial Arial BlackMicrosoft Office 98Microsoft Graph 98 ChartGInformation`Spaces and `Spheres: Interacting with Useful InformationNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleCResearch Program (Some Detail): Constructing and Managing `SpacesNo Slide TitleNo Slide Title2Event/Messaging Middleware - Online ConfigurationNo Slide TitleResearch Program (cont.)Research Program (cont.)No Slide TitleNo Slide Title  Fonts UsedDesign TemplateEmbedded OLE Servers Slide Titles 6> _PID_GUID'AN{CF2E8800-7F2F-11D3-8BC6-0050E4C02F13}*,~eMd[m X?Ҙ.u \%~wƲwpCKK`SǪݩH)wa_A2D^vuzgEtdܯ Z6&ng#a:{!431 f2V?LF' `)sa&,|x#Mād%dOɦ[FzT"\ǣR8J] w[`jc|{gےlTBo9E?80+aye^.św`K:g7{hmEz4jNƻ'X\8dZ\Zi&3U~)zXsx/F,'Ֆ8k~ yu32/.Leކᖰh UBz)V*!,G"dW#Ei_s&ʹ E+''ZZ  Jl]{_ S9;a~*sVZq=gx%G As%aͿö1L 1 8 2Ioc\½ j]!RV`yq!ʧ\-&ߩg/>er;)!wsH IRQ3f._Cuu1:HxE;m5V;q- U 4@Cz÷FPWu*3l' =2u;CdhOէA ԩz{[6[ W odCl̸4/|9THN=QixeLt%آbW"5&JY PY)<kҖ l/p !k6d80ZYπ-ѳE`6>BɺE4~:,u3@vw |+l\G+]'H4%]^c/ F?'G,;T}fBTh.C㳽 <[rl6$ aµ={l Ů$.:Xes`:m8edFP;k6)Վ>*+f2ضZUaBuÁ,@IX. }Q S PѺWXQ}t ɡ#/J@yhוcV`:r #TOw:v7Gn(80_bST-[Asz?~{F52~f$ja&VT F=]DMrT:σ$1tl, p~%r/l`MP|D鄺nJxF[(D@JFLt5|7zt.ccH]£{U L$Q _k_+ۂdi<# $A_K#嬱n_d0ͫWLqSq &t*>;di0]9QU0}KAEj× Haw(wyfMpڜrqˤU>s>wbj&F50Ă}Ц21OX~9Е)x /*Vzd>j/I }V+)ӫ$|-13HJr.ֲ6cȯ~YLUi#EV mBH]HоT3}`thRFyZ6;v'%}2`5R_\}UYx-mZnxt+r@5ig $o&[SW5mcԧe<"\@)DZ!i5xec{.Jz [y UCјP(_SFl(+}nzqS0;2^]hQ@uDa#z{,|pi^*Tw%`G{- 'o)*c\l}5` dX]S0^h