© Synerscope 2013
Sept 2013
© Synerscope 2013
© Synerscope 2013
HOW WE SOLVE THE PROBLEM
AGENDA • Synerscope Background • Problems of classic data analytics with Big Data • Solution of SynerScope mobilizes domain-expertise • How SynerScope works • How SynerScope uses the GPU • Live demo of SynerScope visual analytics
© Synerscope 2013
© Synerscope 2013
SynerScope
SEE -
EXPLORE -
DISCOVER
the best way to turn Big Data to Insight fast
© Synerscope 2013
© Synerscope 2013
SynerScope was started as a result of a problem encountered by its founder: analyzing complex interaction in networks proved impossible. In 2007 Jan-Kees Buenen set out to fix this problem, when in 2009 he met Danny Holten SynerScope was born. Jan-Kees Buenen: Sales, Manufacturing and Quality Analytics founder & CEO • Worked in multinational enterprises since 1993, managed $120 million sales • Deep experience with analytics: sales, CRM, operations research, demand-
forecasting , six-sigma, etc.
Danny Holten: Information Visualization founder & CSO • Worked on the technology of Hierarchical Edge Bundling during his PhD and
postdoc at the visualization department of the Eindhoven university of technology
Jorik Blaas: Computer imaging technologies in medical and photography CTO • Worked on various medical imaging technologies during his PhD at the Delft
university of technology • Successfully lead a team that developed an HDR photography product and an
other that designed and built a X-ray imaging instrument for paintings
Our road to here How we got here
© Synerscope 2013
© Synerscope 2013
HOW WE SOLVE THE PROBLEM
SynerScope allows everyone obtain Insight and Knowledge directly from Big Data: intuitively, quickly, safely and at low cost, and thus shortens the “Time to Insight” 20 Years of academic discussions and collaboration on Knowledge Discovery packed in an instrument that WORKS.
© Synerscope 2013
© Synerscope 2013
Customers and eco-system growing rapidly
(nog aan te vullen)
CUSTOMERS Insurance SIU: we could identify organized ring-fraud in hours Insurance claims director: we found hidden information for underwriting in claims Insurance healthcare: complex billing fraud shows up in the patterns Banking: we could find deviant cash movements instantly Big Four accountant: we could find related risk within 3000 trade positions in hours Telecoms –governmental: cell-tower to cell-tower tracks revealed violent incidents
Eco-system partners: Nvidia: this brings GPU rendering into the broader market of Big Data analytics Yarcdata: SynerScope iss a perfect scalable front-end for our ultra-scale graph engine SAP HANA labs: the complex nested queries of SynerScope add new sales options Dell: we look to support SynerScope for its private cloud appliance
© Synerscope 2013
© Synerscope 2013
Business Manager
Data scientist
Report Request
Business analyst
How business analytics work today
Data Warehouse Queries
Advanced stats
© Synerscope 2013
© Synerscope 2013
Business Manager
Data scientist
Report Request
Business analyst
SynerScope business analytics
Direct Insight
Advanced stats
Events Correlations Free search RCA
Data Warehouse
© Synerscope 2013
© Synerscope 2013
The problem with analytics and Big data
EXPENSIVE ( “RESOURCE- AND ASSET-HEAVY) • Heavy resource involvement in “waterfall method” Business defines questions-
Analysts translate into hypotheses- data scientists prepare the data models - expensive data warehouse loaded up – analysts perform queries and build reports – business does sense-making.
• Data models are core, but they require constant redesign, for new data and new questions
• Aimed at reporting, and repetitive BI, not built for ad-hoc or many new data types • Relies on Brute Force hardware for performance, rigidity built in • Multi-million dollar data ware-houses required to keep system performance
SLOW/ACCESSIBILITY • No access to data without query writing skills • Very slow to change for new data and/or new questions
AGILE BUSINESS INSIGHT DEMANDED
© Synerscope 2013
Introduction to SynerScope
SynerScope is an instrument for : access, search and analysis
o Notably to look at network patterns , scalable, flexible and interactive
Key features
o Proprietary bundling view, for scalable node-link views
o Showing both geo-location and time dimension of relationships
o All data is shown and interactive thanks to GPU acceleration
o Two-way integration with other applications (R, Python, Elasticsearch)
o Rapid set-up, roll-out, either stand-alone and/or virtualized
Synerscope
Legato
NoSQL
MongoDB
elasticsearch
source
data
Data integration (3rd party) Legato – machine based Marcato - human based
© Synerscope 2013
© Synerscope 2013
Volume Variety
Velocity Veracity
Computer Human
Structured Unstructured
LEGATO MARCATO
© Synerscope 2013
© Synerscope 2013
SynerScope – Analysis
12
Extract Data
Automated conversion
Statistics Starting point for analysis
Suggestions and
warnings
Business input
A
X
Y
Input business
questions
HUMAN SynerScope Marcato - Visual Analysis
Phase 1 Phase 2 Phase 3
Identify links over columns
Build logical entities
Add attribute
details
Add external
data Analyze
MACHINE Synerscope Legato Reporting (ETL)
© Synerscope 2013
SynerScope Legato - example
© Synerscope 2013
© Synerscope 2013
2.5% Columns with (nearly) unique values are prime candidates for joinss of sources (Policy #, Social security #, ID)
<40% Columns with many empty fields limited use for joinss or as starting point for visual analysis. (“”, NULL, NA, NaN)
>20% Columns with outliers , often also containing obvious mistakes (NLD * 1000, NL * 2, typo?)
>60% Tables with no direct primary key
>50% Tables with more than three types of data high potential for information
>40% Tables with many columns are mostly joins between sources. Often the relations in data were flattened as a result.
2.5% Columns with a field containing a date/time value important for patterned time views and link identification.
A data technical assessment is made of all data received
Information Quality – (sample 10,000 columns)
© Synerscope 2013
© Synerscope 2013
STEPS:
Example – Phase 2
Calculate candidates for nodes and links
Calculate candidates for joins
Visualize potential joins in SynerScope Marcato
Build a “life-story” around an entity as input to business questions
Determine “white spots” in the data
Suggestions and warnings
Realtime reporting – Synerscope Legato
Suggestions for joins
Construct logical entities
Select from available nodes
and links
Input towards the business
questions
Expand through external sources)
Entity
New services
Client focus
Profit / savings
© Synerscope 2013
SynerScope Legato - example
© Synerscope 2013
SynerScope Legato - example
© Synerscope 2013
© Synerscope 2013
Data profile – table – Data Quality
BIJLAGEN
SynerScope Marcato
19
© Synerscope 2013
© Synerscope 2013
SEE
EXPLORE
DISCOVER
BIJLAGEN
Live demo SynerScope Marcato
26
© Synerscope 2013
© Synerscope 2013
SynerScope
SEE -
EXPLORE -
DISCOVER
the best way to turn Big Data to Insight fast
domain-experts to work side-by-side with data experts
Upcoming GTC Express Webinars
Register at www.gputechconf.com/gtcexpress
September 10 - Virtualizing Tough 3D Workloads with VMware
Horizon View and NVIDIA Technologies
September 12 - Guided Performance Analysis with NVIDIA
Visual Profiler
September 17 - ArrayFire: A Productive GPU Software Library
for Defense and Intelligence Applications
September 19 - Learn How to Debug OpenGL 4.2 with NVIDIA®
Nsight™ Visual Studio Edition 3.1
September 25 - An Introduction to GPU Programming
GTC 2014 Call for Submissions
Looking for submissions in the fields of
Science and research
Professional graphics
Mobile computing
Automotive applications
Game development
Cloud computing
Submit by September 27 at www.gputechconf.com