Transcript Formation HUPI-INTERACTIF et HUPI-LINK

A Big Data Platform for Geoscience
June, 6th 2016 – Vincent Moreno, CEO
BIG DATA PLATFORM = BIG DATA FACTORY
A COMPLETE PROCESS
1 Raw data acquisition
2 Analytics processing
3 Visualize & Integrate Data
Experts Information
Systems
ALL data sources
openData
ARCHITECTURE
HUPI Factory
API
RestFul
Sources
Destination
Vizualisation
6
Service client
Logs
CatchBox
(collect)
Data Mining
Treatments
Calls/ Restitutions
Objets connectés
4
7
5
Fichiers
2
Storag
Data
Qualification
Web Services
Applications
mobile
Kafka
MongoDB
HDFS/DFS
3
Hive
1
8
API/RestFul
PORTAL DESCRIPTION
DISTRIBUTED COMPUTING : SPARK
IMAGINE IF YOU COULD ON LARGE DATA VOLUMES, ITERATE IN
CONTINUOUS
Benefits
•
Support for complex analytics
•
Treat in parallel large data volumes
•
Reduct price cost for compute (less expensive to have multiple servers, than one
BIG server)
•
SPARK 100x faster than HADOOP (MapReduce)
•
Support multi-langages:
SPARK OVERVIEW
FRAMEWORK
WHY USE A BIG DATA PLATEFORM?
#1. Do not worry about the underlying technology, focus on the analytics
and the applications
#2. Reduce the time from raw data to results and insights; accelerate
experiments
#3. Share a common plateform to facilitate collaboration
#4. Centralize and reuse components
MERCI DE
VOTRE
ATTENTION
HUPI