0
H T 2012	  	  Technologies	  
HOST:	  Eric	  Kavanagh	  
 	  	  THIS	  YEAR	  WAS…	  
ANALYTIC	  PLATFORMS	  ž  Analytic	  Platforms	  represent	  the	  next	  major	      phase	  in	  the	  evolution	  of	 ...
ANALYST:	                          Mark	  Madsen	  THE	  LINE	  UP	     CEO,	  Third	  Nature	  Inc.	                     ...
INTRODUCING	     Mark	  Madsen	  
Philosophical	  ques.on	                        When	  modeling	  a	  data	                        warehouse,	  is	  it	  ...
I need that          It would be logical                             data now.            to keep all the                 ...
Analy.cs	  embiggens	  the	  data	  volume	  problem	     Many	  of	  the	  processing	  problems	  are	  O(n2)	  or	  wor...
Big	  changes	  for	  data	  warehousing	  workloads	                                                 Much	  of	  the	  an...
What	  do	  we	  mean	  by	  analy.cs	  pla?orm?	    AnalyHcs<>	  BI,	  different	    usage	  model	  and	  workload	    De...
Analy.c	  pla?orm	  design	  goals                                                            	       1.  Decouple	  the	 ...
Be	  suspicious	  of	  anyone	  who	  says	  Hadoop	  is	  the	  only	  answer                                            ...
IT	  reality	  is	  mul.ple	  data	  stores,	  distributed	  pla?orm	    Separate, purpose-built databases and processing ...
INTRODUCING	     John	  O’Brien	  
16                                     ROLE OF ANALYTIC DBMS IN                                     MODERN BI ARCHITECTURE...
17      Modern BI Architectures      ROLE OF ANALYTIC DATABASES      Data persistence for optimized BI workloads      •  2...
18      Modern BI Architectures      MIXED WORKLOAD CAPABILITIES                                                          ...
19      Modern BI Architectures      MIXED WORKLOAD CAPABILITIES                                            Hadoop        ...
20      Modern BI Architectures      SEMANTIC INTEGRATION AT DATA     Know when pulling data                              ...
21      Modern BI Architectures      SEMANTIC INTEGRATION ABOVE DATA    Where should    semantic knowledge                ...
22      Modern BI Architectures      THINGS TO KEEP AN EYE ON      1.  Expect modern BI architectures to evolve in coming ...
INTRODUCING	     Walter	  Maguire	  
Enabling	  Big	  Data	  ApplicaHons	          Walter	  Maguire,	  Director	  of	  AnalyHcs	  Copyright 2012 ParAccel, Inc....
ParAccel	  Analy.c	  Pla?orm	  is…	     …built	  for	  high	  performance,	  interac.ve	  analy.cs.	     On	  Demand	  Int...
ParAccel	  technology	  is	  the	  first	  to	  deliver	  on	     Coopera.ve	  Analy.c	  Processing	                   SQL-...
ParAccel	  ODI	  Services	  makes	  our	  pla?orm	  the	  analy.c	     engine	  for	  en.re	  ecosystems.	                ...
One	  Size	  Does	  Not	  Fit	  All:	  Why	  an	  Ecosystem?	                  ReporHng	              AnalyHcs	           ...
The	  Best	  Way	  to	  Do	  Analy.cs	  on	  Hadoop	     Data	           Create	  a	  high-­‐performance,	  node-­‐to-­‐no...
Read	  from	  Hadoop:	  	                       INSERT	  INTO	  mytable	  SELECT	  *	  FROM	                       	  Hado...
What s	  Next	  for	  the	  Hadoop	  ODI?	     HCatalog	  Integra.on	           •        Apache	  HCatalog	  is	  a	  tabl...
ODI	  Services	  Architecture	  Overview	                                               Leader	  Node	                    ...
ODI	  Services	  Architecture	  Overview	                                               Leader	  Node	                    ...
ODI	  Services	  Architecture	  Overview	                                                 Leader	  Node	                  ...
ODI	  Services	  Architecture	  Overview	                                                 Leader	  Node	                  ...
Developing	  and	  Deploying	  ODIs	           Write	  command	  line	  executable	  or	  interpreted	  script	           ...
Developing	  and	  Deploying	  ODIs	           o	  	  	  Enables	  a	  spectrum	  of	  use	  cases	  from	  fast	  prototy...
ODI	  services:	  examples	           Event	  Capture	           Smart	  Meter	  Logging	           RFID	  Tag	  Capture	 ...
Coopera.ve	  Analy.c	  Processing	  is	  the	  Future	                   SQL-­‐Based	  Business	                          ...
THANK	  YOU!	  The	  Archive	  Trifecta:	      •  Inside	  Analysis	  	  www.insideanalysis.com	      •  SlideShare	  	  w...
Hot Technologies of 2012
Upcoming SlideShare
Loading in...5
×

Hot Technologies of 2012

545

Published on

Analytic Platforms with Mark Madsen, John O'Brien and ParAccel
Live Webcast Dec. 5, 2012

There's a good reason why so many people are talking about analytic platforms these days. The surge in popularity of Big Data, coupled with the need to reconcile this new source of insights with Business Intelligence and Data Warehousing, has fueled a wave of innovation for synthesizing analytical capabilities. What are the latest innovations in analytic platforms? Check out this episode of Hot Technologies to find out!

Veteran Analysts Mark Madsen of Third Nature and John O'Brien of Radiant Advisors will offer their insights on what to look for in a robust analytic platform. They'll then take a briefing from Walter Maguire of ParAccel, who will provide details about his company's platform offering, which includes a high-performance analytic database, Hadoop integration, and innovative extensions that allow companies to embed analytics in business process, create big data apps, and create on demand access to 100s of new data sources.

Visit: http://insideanalysis.com

Published in: Technology
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
545
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
8
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Transcript of "Hot Technologies of 2012"

  1. 1. H T 2012    Technologies  
  2. 2. HOST:  Eric  Kavanagh  
  3. 3.      THIS  YEAR  WAS…  
  4. 4. ANALYTIC  PLATFORMS  ž  Analytic  Platforms  represent  the  next  major   phase  in  the  evolution  of  Business   Intelligence  and  Analytics  ž  These  platforms  should  foster  collaboration   and  transparency  ž  Users  should  be  enabled  to  access  and   analyze  the  data  they  want,  quickly  and   effectively  
  5. 5. ANALYST:   Mark  Madsen  THE  LINE  UP   CEO,  Third  Nature  Inc.   ANALYST:   John  O’Brien   Principal  &  CEO,  Radiant  Advisors   GUEST:   Walter  Maguire   Director  of  Analytics,  ParAccel  
  6. 6. INTRODUCING   Mark  Madsen  
  7. 7. Philosophical  ques.on   When  modeling  a  data   warehouse,  is  it  best  to:     A.  Choose  each  data   element  in  your  schema   based  on  usefulness  /usage   or   B.  Keep  every  element  in   the  source  data?    © Third Nature Inc.
  8. 8. I need that It would be logical data now. to keep all the It will take.   data in one place. 6 months       The  common  situa.on  for  analysts  © Third Nature Inc.
  9. 9. Analy.cs  embiggens  the  data  volume  problem   Many  of  the  processing  problems  are  O(n2)  or  worse,  so  even   moderate  data  can  be  a  problem  for  DW  models  &  architectures  © Third Nature Inc.
  10. 10. Big  changes  for  data  warehousing  workloads   Much  of  the  analyHcs  data  is   being  read,  wriIen  and   processed  interacHvely  with   people  waiHng,  or  in  real  Hme   machine  to  machine  contexts.   The  results  of  analyHc   processing  can  –  oMen  do  –   feed  back  into  the  system  from   which  they  originate.   Our  DW  design  point  was  not   changing  tables,  ephemeral   paIerns,  large  data  movement   –  it  was  a  pub-­‐sub  model.  © Third Nature Inc.
  11. 11. What  do  we  mean  by  analy.cs  pla?orm?   AnalyHcs<>  BI,  different   usage  model  and  workload   Deployment  environment?   What  sort?   ▪  Batch   ▪  Real  Hme   Development  or  exploraHon   environment?  For  what?   ▪  The  process  of  model   building   ▪  Exploratory  analysis   ▪  AnalyHc  data  management   A real analytics production workflow Hatch, CIKM 2011  © Third Nature Inc.
  12. 12. Analy.c  pla?orm  design  goals   1.  Decouple  the  analyHc  plaXorm  from  the  data   warehouse:  it  can  be  a  part  of  the  delivery  layer,  or   the  integraHon  layer,  or  both.   2.  Support  the  analyHc  development  and   maintenance  processes,  preferably  without   unsupported  data  copying.   3.  Support  the  producHon  deployment  processes.   Don’t  try  to  force-­‐fit  “offload”  and  “merge”  paIerns.   To  the  extent  you  can  do  all  of  this  without  moving  data   around,  it’s  a  big  win.    © Third Nature Inc.
  13. 13. Be  suspicious  of  anyone  who  says  Hadoop  is  the  only  answer  © Third Nature Inc.
  14. 14. IT  reality  is  mul.ple  data  stores,  distributed  pla?orm   Separate, purpose-built databases and processing systems for different types of data and query / computing workloads is the new norm for information delivery. Delivery must be separated.     Informa.on     delivery  layer     Platform layer     1 Marge  Inovera $150,000 Statistician 1 Marge  Inovera $150,000 Statistician 1 Marge  Inovera $150,000 Statistician 1 Marge  Inovera $150,000 Statistician   1 Marge  Inovera $150,000 Statistician 1 Marge  Inovera $150,000 Statistician 1 Marge  Inovera $150,000 Statistician 1 Marge  Inovera $150,000 Statistician 1 Marge  Inovera $150,000 Statistician Data 2 Anita  Bath $120,000 Sewer  inspector 2 Anita  Bath $120,000 Sewer  inspector 2 Anita  Bath $120,000 Sewer  inspector 2 Anita  Bath $120,000 Sewer  inspector 2 Anita  Bath $120,000 Sewer  inspector 2 Anita  Bath $120,000 Sewer  inspector 2 Anita  Bath $120,000 Sewer  inspector 2 Anita  Bath $120,000 Sewer  inspector 2 Anita  Bath $120,000 Sewer  inspector 3 Ivan  Awfulitch $160,000 Dermatologist 3 Ivan  Awfulitch $160,000 Dermatologist 3 Ivan  Awfulitch $160,000 Dermatologist 3 Ivan  Awfulitch $160,000 Dermatologist 3 Ivan  Awfulitch $160,000 Dermatologist 3 Ivan  Awfulitch $160,000 Dermatologist 3 Ivan  Awfulitch $160,000 Dermatologist 3 Ivan  Awfulitch $160,000 Dermatologist 3 Ivan  Awfulitch $160,000 Dermatologist 4 Nadia  Geddit $36,000 DBA 4 Nadia  Geddit $36,000 DBA 4 Nadia  Geddit $36,000 DBA 4 Nadia  Geddit $36,000 DBA 4 Nadia  Geddit $36,000 DBA 4 Nadia  Geddit $36,000 DBA 4 Nadia  Geddit $36,000 DBA 4 Nadia  Geddit $36,000 DBA 4 Nadia  Geddit $36,000 DBA   Warehouse 1 Marge  Inovera $150,000 Statistician 1 Marge  Inovera $150,000 Statistician 1 Marge  Inovera $150,000 Statistician 1 Marge  Inovera $150,000 Statistician 1 Marge  Inovera $150,000 Statistician 1 Marge  Inovera $150,000 Statistician 1 Marge  Inovera $150,000 Statistician 1 Marge  Inovera $150,000 Statistician 1 Marge  Inovera $150,000 Statistician 2 Anita  Bath $120,000 Sewer  inspector 2 Anita  Bath $120,000 Sewer  inspector 2 Anita  Bath $120,000 Sewer  inspector 2 Anita  Bath $120,000 Sewer  inspector 2 Anita  Bath $120,000 Sewer  inspector 2 Anita  Bath $120,000 Sewer  inspector 2 Anita  Bath $120,000 Sewer  inspector 2 Anita  Bath $120,000 Sewer  inspector 2 Anita  Bath $120,000 Sewer  inspector 3 Ivan  Awfulitch $160,000 Dermatologist 3 Ivan  Awfulitch $160,000 Dermatologist 3 Ivan  Awfulitch $160,000 Dermatologist 3 Ivan  Awfulitch $160,000 Dermatologist 3 Ivan  Awfulitch $160,000 Dermatologist 3 Ivan  Awfulitch $160,000 Dermatologist 3 Ivan  Awfulitch $160,000 Dermatologist 3 Ivan  Awfulitch $160,000 Dermatologist 3 Ivan  Awfulitch $160,000 Dermatologist 4 Nadia  Geddit $36,000 DBA 4 Nadia  Geddit $36,000 DBA 4 Nadia  Geddit $36,000 DBA 4 Nadia  Geddit $36,000 DBA 4 Nadia  Geddit $36,000 DBA 4 Nadia  Geddit $36,000 DBA 4 Nadia  Geddit $36,000 DBA 4 Nadia  Geddit $36,000 DBA 4 Nadia  Geddit $36,000 DBA Databases Documents Flat Files XML Queues ERP Applications Source  Environments  © Third Nature Inc.
  15. 15. INTRODUCING   John  O’Brien  
  16. 16. 16 ROLE OF ANALYTIC DBMS IN MODERN BI ARCHITECTURES Hot Technologies – December 5, 2012 John O’Brien, Radiant Advisors john.obrien@radiantadvisors.com© Copyright 2012 Radiant Advisors. All Rights Reserved v1.00.000
  17. 17. 17 Modern BI Architectures ROLE OF ANALYTIC DATABASES Data persistence for optimized BI workloads •  2-tier versus 3-tier debate •  Why 3-tier will be next generation Integrating semantics “in” or “above” data •  Cross database versus data virtualization debate •  Why a evolving combination will be next generation Predictions for 2013 and 2014© Copyright 2012 Radiant Advisors. All Rights Reserved v1.00.000
  18. 18. 18 Modern BI Architectures MIXED WORKLOAD CAPABILITIES 3-Tier BI Architecture Key Value Store (Hadoop) Discovery Oriented Analytic Database EDW Highest Scalability Technologies RDBMS Lowest Cost Schema-less Without Context Accessibility: Programming SQL, MDX, UDF SQL Access Workload: Flexible, Scalable Analytic Optimized Reference Data Mgmt Maturity: Emerging Accepted Mature© Copyright 2012 Radiant Advisors. All Rights Reserved v1.00.000
  19. 19. 19 Modern BI Architectures MIXED WORKLOAD CAPABILITIES Hadoop 2-Tier BI Architecture Programming Batch Oriented Highest Scalability Lowest Cost Analytic EDW Flexibility Workloads RDBMS Schema-less Without Context Broad SQL Accessibility by Users What’s not to like about this? •  While possible, analytic execution will be slower performing and more time consuming to develop and manage in Hadoop stores for BI teams •  Broad accessibility of BI tools will be a limitation© Copyright 2012 Radiant Advisors. All Rights Reserved v1.00.000
  20. 20. 20 Modern BI Architectures SEMANTIC INTEGRATION AT DATA Know when pulling data BI tools into ADBMS is ok (today) SQL HCatalog / Hive-QL MapReduce Semantic Projections Integration Links Gateways text Analytic DBMS EDW Columnar storage (RDBMS) Hadoop In-memory access Document stores Text Analysis Semantic Discovery Graph Analysis ROLAP/MOLAP© Copyright 2012 Radiant Advisors. All Rights Reserved v1.00.000
  21. 21. 21 Modern BI Architectures SEMANTIC INTEGRATION ABOVE DATA Where should semantic knowledge Future BI tools live in the architecture? HCatalog Services SQL / Data Virtualization MapReduce text In-memory Semantic Discovery Hadoop Analytic DBMS EDW© Copyright 2012 Radiant Advisors. All Rights Reserved v1.00.000
  22. 22. 22 Modern BI Architectures THINGS TO KEEP AN EYE ON 1.  Expect modern BI architectures to evolve in coming years as technologies pave the way 2.  Adoption of R and PMML for analytic models to become portable across platforms 3.  How vendors push-down execution code in Hadoop or pull-through data into analytic databases 4.  Polyglot persistence will optimize on multiple storage engines with service layer access© Copyright 2012 Radiant Advisors. All Rights Reserved v1.00.000
  23. 23. INTRODUCING   Walter  Maguire  
  24. 24. Enabling  Big  Data  ApplicaHons   Walter  Maguire,  Director  of  AnalyHcs  Copyright 2012 ParAccel, Inc. 24
  25. 25. ParAccel  Analy.c  Pla?orm  is…   …built  for  high  performance,  interac.ve  analy.cs.   On  Demand  Integra.on   Integrated  Analy.cs   Database   ParAccel  Analy.c  Pla?orm   Basic  AnalyHcs   Teradata   Advanced  AnalyHcs   Hadoop   Analy.c  Engine   Streaming  Data   Columnar   ApplicaHons   Compression   Compiled   Parallel  Processing   SQL  OpHmizaHon   Data  Scale   In-­‐Memory  Op.on  Available   Plan  OpHmizaHon   AnalyHc  Scale   ExecuHon  OpHmizaHon   User  Scale   Comms  OpHmizaHon   InteracHve  Scale   I/O  OpHmizaHon  Copyright 2012 ParAccel, Inc. 25
  26. 26. ParAccel  technology  is  the  first  to  deliver  on   Coopera.ve  Analy.c  Processing   SQL-­‐Based  Business   Advanced     Analy.c     Intelligence  and   Analy.cs   Applica.ons   Repor.ng  Tools   ParAccel  Analy.c  Pla?orm   Enterprise   Hadoop   Data  Warehouse   On  Demand  Integra.on   Embedded   3rd  Party   Big  Data   Machine   Opera.onal   Streaming   Analy.cs   Info   Logs   Apps   Data   Data   Data   Provider  Copyright 2012 ParAccel, Inc. 26
  27. 27. ParAccel  ODI  Services  makes  our  pla?orm  the  analy.c   engine  for  en.re  ecosystems.   ParAccel  Analy.c  Pla?orm   Enterprise   Hadoop   Data  Warehouse   1.  Share  both  data  and  processes  in  both  direcHons   2.  Transform  incoming  data  for  analyHc  performance   3.  Interact  with  many  programming  languages  (Java,  Python,  more)   4.  Persist  or  stream  data  through  analyHc  processing   5.  Rapidly  build  new  On  Demand  IntegraHon  modules   On  Demand  Integra.on  Services   Embedded   3rd  Party   Big  Data   Machine   Opera.onal   Streaming   Analy.cs   Info   Logs   Apps   Data   Data   Data   Provider  Copyright 2012 ParAccel, Inc. 27
  28. 28. One  Size  Does  Not  Fit  All:  Why  an  Ecosystem?   ReporHng   AnalyHcs   Archiving   Dashboards   Data  Mining   Filtering   StaHc  Analysis   Dynamic  Analysis   Text  Search   OLAP   Complexity   Text  AnalyHcs       TransformaHon        Copyright 2011 ParAccel, Inc. 28
  29. 29. The  Best  Way  to  Do  Analy.cs  on  Hadoop   Data   Create  a  high-­‐performance,  node-­‐to-­‐node,  bi-­‐ direcHonal,  connecHon  between  Hadoop  and   an  analyHc  plaXorm  that  is  capable  of  sharing   both  data  and  processes  so  that  the  analyHc   plaXorm  becomes  an  extension  of  the  Hadoop   cluster  and  you  can  uHlize  the  lingua  franca  of   analyHcs,  SQL.  Copyright 2012 ParAccel, Inc. 29
  30. 30. Read  from  Hadoop:     INSERT  INTO  mytable  SELECT  *  FROM    HadoopIn(with  hfs_name( hadoopfile )                                                          mr_job( xyz )                                                        pa_schema( mytable ));   Write  to  Hadoop:     SELECT  num_rows  FROM   HadoopOut(on  (select  *  from  mytable)                                              WITH  hdfs_name(   hadoopfile ));  Copyright 2012 ParAccel, Inc. 30
  31. 31. What s  Next  for  the  Hadoop  ODI?   HCatalog  Integra.on   •  Apache  HCatalog  is  a  table  and  storage  management  layer   for  Hadoop   Provides  table  abstracHon  for  HDFS  file  for  various  data  processing  tools     •  ODI  Scan  filters   UDF  Filters  from  the  SQL  will  be  pushed  down  to  Hadoop  as  parHHon   filters     Greatly  simplify  invesHgaHve  workflow  on  large  volumes  of   data  in  Hadoop  before  bringing  it  into  ParAccel   Simplify  development  of  Hadoop  to  ParAccel  integraHons    Copyright 2012 ParAccel, Inc. 31
  32. 32. ODI  Services  Architecture  Overview   Leader  Node   ODI  Services   Service  Mgmt.   Service  Context   Compute  Node   Compute  Node   Perl   Perl   Python   Python   Services   Services   Services   Java   Java   ODI   ODI   Bash   Bash   R   R   Etc.   Etc.   Compute  Node   Perl   Python   Services   Java   ODI   Bash   R   Etc.  
  33. 33. ODI  Services  Architecture  Overview   Leader  Node   •  Job Progress & Status •  Installation ODI  Services   •  Logging Service  Mgmt.   •  Balancing Service  Context   •  Optimization Compute  Node   Compute  Node   Perl   Perl   Python   Python   Services   Services   Services   Java   Java   ODI   ODI   Bash   Bash   R   R   Etc.   Etc.   Compute  Node   Perl   Python   Services   Java   ODI   Bash   R   Etc.  
  34. 34. ODI  Services  Architecture  Overview   Leader  Node   •  Job Progress & Status •  Installation ODI  Services   •  Logging Service  Mgmt.   •  Balancing Service  Context   •  Optimization Compute  Node   Compute  Node   Perl   Perl   Python   Python   Services   Services   Services   Java   Java   ODI   ODI   Bash   Bash   R   R   Etc.   Etc.   STDIN STDOUT STDERR Compute  Node   Metadata Perl   Mgmt Framework Python   Services   Java   ODI   Bash   R   Etc.  
  35. 35. ODI  Services  Architecture  Overview   Leader  Node   •  Job Progress & Status •  Installation ODI  Services   •  Logging Service  Mgmt.   •  Balancing Service  Context   •  Optimization Compute  Node   Compute  Node   Perl   Perl   Python   Python   Services   Services   Services   Java   Java   ODI   ODI   Bash   Bash   R   R   Etc.   Etc.   STDIN STDOUT STDERR Compute  Node   Metadata Perl   •  Command line Mgmt Framework Python   executable Services   Java   •  3rd party ODI   Bash   interpreter (e.g. R   Perl, Python, Etc.   Java VM)
  36. 36. Developing  and  Deploying  ODIs   Write  command  line  executable  or  interpreted  script   Test  with  ODI  Services  test  harness   Load  to  lead  node   Lead  node  distributes  ODI  across  the  compute  nodes    Copyright 2011 ParAccel, Inc. 36
  37. 37. Developing  and  Deploying  ODIs   o      Enables  a  spectrum  of  use  cases  from  fast  prototyping  to   one-­‐off  and  producHon  data  loads/unloads   o      No  need  to  code  to  C++  APIs  or  be  exposed  to  any   complexity   o      Fast  development   o      Handles  parallelism  for  you   o      Simple  protocol   o      Logging   o      Monitoring  progress  Copyright 2011 ParAccel, Inc. 37
  38. 38. ODI  services:  examples   Event  Capture   Smart  Meter  Logging   RFID  Tag  Capture   Tweets,  Facebook,  consolidated  social  streams   Web  services  (Salesforce,  Eloqua,  Omniture,  etc.)   Enterprise  Semi-­‐Structured  sources  (Outlook,  Gmail,  Zendesk,   etc.)   Embedded  business  processes  (ex:  call  center,  distribuHon   rouHng)  Copyright 2011 ParAccel, Inc. 38
  39. 39. Coopera.ve  Analy.c  Processing  is  the  Future   SQL-­‐Based  Business   Advanced     Analy.c     Intelligence  and   Analy.cs   Applica.ons   Repor.ng  Tools   ParAccel  Analy.c  Pla?orm   Enterprise   Hadoop   Data  Warehouse   On  Demand  Integra.on   Embedded   3rd  Party   Big  Data   Machine   Opera.onal   Streaming   Analy.cs   Info   Logs   Apps   Data   Data   Data   Provider  Copyright 2012 ParAccel, Inc. 39
  40. 40. THANK  YOU!  The  Archive  Trifecta:   •  Inside  Analysis    www.insideanalysis.com   •  SlideShare    www.slideshare.net/InsideAnalysis   •  YouTube    www.youtube.com/user/BloorGroup  
  1. A particular slide catching your eye?

    Clipping is a handy way to collect important slides you want to go back to later.

×