How MTV Networks leverages MongoDB for CMS

A presentation by Jeff Yemin of MTV Networks.

MTV Networks Online Division is leveraging MongoDB as the database platform for its next-generation CMS. The talk will focus on these points: A brief outline of the particular challenges of building a CMS for a large media-focused enterprise company; the selection criteria for the database platform (and how MongoDB fit them); the way the CMS leverages the various features of MongoDB.

Transcript

1. How MTV Networks Leverages MongoDB for CMS MongoDB @ MTV
2. What is MTV Networks Online? •  Several  hundred  websites,  including   –  www.comedycentral.com   –  www.thedailyshow.com   –  www.spike.com   –  www.nickelodeon.com   –  www.nickjr.com   –  www.mtv.com   –  www.vh1.com   –  …  
3. How are web sites constructed @ MTV? •  A  wide  variety  of  language  and  frameworks,  including   –  Java/JSP/Tiles   –  Java/Grails   –  Java/ATG  Dynamo/JHTML   –  PHP   –  Ruby  on  Rails   •  BUT,  almost  all  are  built  on  top  of  one  or  more  content   management  systems.  
4. A Brief History of CMS @ MTV •  Two  successful  CMS  built  in  house   –  TradiNonal  RDBMS  as  repository   –  RDBMS-­‐based  XML  CLOB  Store  as  repository  
5. Looking for a better solution •  Scalability,  Stability,  Speed   •  Client-­‐Defined  Content  Model   •  Hierarchical,  Linkable  Content  Model   •  Query-­‐ability   •  OperaNonal  Efficiency  
6. MongoDB •  Scalability,  Stability,  Speed   –  CollecNons,  Databases,  Replica  Sets   –  Proven  stability   –  Legendary  speed  
7. MongoDB •  Client-­‐Defined  Content  Model   –  MongoDB  is  schema-­‐less,  so  model  is  by  definiNon  a  client   concern  
8. MongoDB •  Hierarchical,  Linkable  Content  Model   –  MongoDB  is  a  document  store   –  DBRef  or  embedded  idenNfiers  
9. MongoDB •  Query-­‐ability   –  Field  level,  arrays,  embedded   –  Sort   –  Skip,  Limit   –  Operators   –  Regex   –  elemMatch  
10. MongoDB •  OperaNonal  Efficiency   –  Separate  process,  communicaNon  via  wire  protocol   –  Simple  clustering   –  No  clustered  storage  requirement   –  Explain   –  Slow  query  logging   –  Indexing  
11. How we use MongoDB for CMS •  As  a  repository   •  CMS  built  in-­‐house  as  a  Java  applicaNon  on  top   •  CMS  hides  repository  implementaNon  from  client  as   much  as  possible  
12. Content Modeling •  Content  Models  defined  in  XML   –  Field  and  Link  definiNons   –  MulNple  Inheritance   –  AnnotaNons   •  Content  Models,  managed  via  REST  endpoints,  are   stored  in  MongoDB  
13. Content Model Example
14. Tools Support
15. Content Publishing •  Three  Stage  Publishing  for  each  Site   •  Authoring  (shared),  Staging,  Live   •  Implemented  as  three  separate  MongoDB  databases   •  Copy  documents  between  collecNons  in  different  databases   •  OpNon  to  have  each  database  in  separate  replica  set  
18. Tools Support
19. Collections and Databases •  One  database  for  global  configuraNon   –  Schema  definiNons   –  Stored  query  definiNons   –  Site  and  stage  definiNons   •  One  database  for  each  stage  of  each  site   –  One  collecNon  for  all  content   –  Intra-­‐collecNon  document  references   –  Lots  of  indices,  mostly  on  link  structure  
21. Tools Support
22. Data Access •  Goals   –  To  abstract  underlying  repository   –  To  allow  clients  to  specify  at  a  very  fine  level  the  shape  of  the   data  access  response  object   –  To  minimize  the  size  of  the  data  access  response  object   –  To  protect  server  from  dangerous  queries  
23. JsonPath
24. Data Access Example
25. Data Access Example
26. Site Launches
27. Metrics