QCon London 2012: Opening keynote: The Data Panorama

Changing nature of data

Data is growing outside of our control

  • Wallmart : 1 million transaction/day
  • Facebook : 40 millions photo upload/day

It’s a problem for everyone, not just for Google er Amazon

Data is distributed

98% ot Internet acces in Africa is made on a mobile device

Data is valuable

US Health $300 millions/year

Data is urgent

need to respond in real time

Data is connected

within its ecosystem

Response to data change

(before -> after)

1 huge server -> lots of small boxes

SQL doesn’t work in a cluster -> NoSQL

  • Google : BigTable
  • Amazon : Dynamo

NoSQL, Aggregates (from E. Evans’ DDD)

  • non relational
  • open source
  • cluster frientdly
  • 21st century web

store aggregates directly, no need to shard -> polyglot persistance

moving from “Integration DB” to “Application DB”

one DB/application different than one DB for all applications

Event sourcing

2 levels of storage :

  • in app state
  • log

You can rebuild your state with the log.

Datasources has changed

Now it’s text, images, videos, connections. Analysis will be :

  • pattern recognition
  • data mining
  • chasing connection (via social networks)

This needs different algorithms :

  • map-reduce is a very (hard and) different thinking
  • visualization data is necessary to understand data

How do we use data now ?

  • data “scientist” -> data journalist
  • personalization (see iGoogle)
  • data warehouse -> Agile Analitics (Ken Coller)
  • crowd sourcing (ushahidi.com)
  • govs open data

Beware of privacy : protection VS privacy

Developpers have :

  • to be careful not to distord information
  • the responsability to make accurate connections

Réactions