Enterprise Analytics Accelerating Your Path to Value with an Open Analytics Platform Federico Pozzi @fedealbpozzi Mathias Coopmans @macoopma
Characteristics of a badly managed platform No clear data ownership Lack of board level support for fact based decisioning Low trust from end users in data Security leaks Copies of copies of copies of the same data Unclear process for accessing data Multitude of end user tools No clear alignment between business and IT Lack of process to guarantee privacy of data.
Why now? Access to Technology Cheap / Free Digitalisation Innovation Cloud adoption Big Data Revolution Regulation (GDPR) Cost Ethics
The seven traits of a modern analytical platform
Process supports the analytical lifecycle 1
Visual Interfaces in SAS Viya
SAS Enterprise Decision Management at a Global Financial Services Firm: Enabling More Rapid Implementation of Decision Models into Production
Engines for processing data in-memory, in-database and real-time streaming SOURCE DATA EVENT STREAM PROCESSING Real Time Processing Data/ Low Latency Alerting MODEL DEVELOPMENT & MANAGEMENT 2 Sensor Data/ Smart Device Telemetry ERP / IS App Web Data Databases Routers, switches Native Connectivity & WebServices Native Connectivity & WebServices In-Stream Sensor Data Management Data Cleansing Thresholds Model & Pattern Aggregation Queries BATCH DATA Batch or Mini-Batch Data Management Data Access Data Cleansing Data Preparation Data Aggregation Batch Processing IN-MEMORY ANALYTICS SERVER Machine Learning Optimization Data Mining Forecasting Statistical Analysis Reliability Modeling Simulation Model building and deployment STORAGE/ PROCESSING HADOOP Dashboard/Alerts Mobile Dashboard/ Alerts Visualization & Modelling Workflows/ Decision Management
New Landscape - New Needs Bigger Data Volume Velocity Variety Act Faster Reduced time to decision and action Immediate low latency answers Continuously evaluate opportunities and risks More agile, more responsive
Real Time Marketing Communications Fraud Detection Cyber Security IT Operations Cross-industry applicability and value Streaming Analytics Industry, Energy Health Care Capital Markets Manufacturing Enterprise Decisions Supply Chain
Smart Cities and Homes Connected Customer Communications Surveillance Connected Car Transportation I Things OF nternet Building Management Energy Agriculture Insurance Manufacturing Healthcare Retail
Analytics Lifecycle Traditional Analytics Lifecycle Data Data Storage ETL Deploy Alerts - Reports Decisioning Streaming Data Access - Store - Analyze
Enrich Store Deploy Analytics Lifecycle Stream Understand Act Data Data Storage ETL Deploy Alerts - Reports Decisioning Streaming Data Streaming Model Execution
Enrich Store Deploy SAS Event Stream Processing Engineered For Fast And Adaptive Action Data Data Storage ETL Deploy Alerts - Reports Decisioning Streaming Data SAS Event Stream Processing Apply high end analytics on event streams Offline, data at rest identifies emerging trends Pattern detection at event stream source Dynamically update models into live stream
The platform is scalable and can grow 3
The SAS platform architecture-stripped down In-Memory Engine Cloud Analytic Services (CAS) UAA UAA UAA Viya Core Microservices Data Source Mgmt Folders etc BI GUIs CAS Mgmt Log Analytics GUIs Data Mgmt GUIs Query Gen Env Mgr Model Mgmt Audit
SAS Viya Single Machine Deployment Single Machine Deployment SAS Viya 3.2 CAS Controller CAS Server Monitor SAS Object Spawner SAS Workspace Server SAS Studio Embedded Web Application Server Web Applications Microservices SAS Configuration Server Apache HTTP Server SAS Infrastructure Data Server SAS Data Connector SAS/CONNECT Server & Spawner SAS Message Broker System Services Linux x64 only
SAS Viya Multi-Machine Deployment Distributed CAS Distributed SAS Cloud Analytic Services Deployment SAS Cloud Analytic Services SAS Viya 3.2 CAS Controller CAS Worker SAS Object Spawner SAS Studio CAS Controller CAS Worker SAS Workspace Server Embedded Web Application Server CAS Server SAS Data SAS/CONNECT Monitor Connector Server & Spawner SAS Data Connector Web Applications SAS Configuration Server Apache HTTP Linux x64 only Linux x64 only Server SAS Infrastructure Microservices Data Server CAS Worker CAS Worker SAS Message Broker System Services CAS Worker CAS Worker SAS Data SAS Data Linux x64 only Connector Connector Linux x64 only Linux x64 only
The SAS platform architecture Parallel & Serial, Pub / Sub, Web Services, MQs Source-based Engines In-Stream In-Hadoop In-Database In-Memory Engine Cloud Analytic Services (CAS) UAA UAA UAA Viya Core Microservices Data Source Mgmt Folders etc BI GUIs CAS Mgmt Log Analytics GUIs Data Mgmt GUIs Query Gen Env Mgr Model Mgmt Audit Data Platforms Cloud & Containers Platform Google Cloud Platform
Access to a variety of data sources 4
Sas & hadoop User Interface SAS Data Management SAS Enterprise Miner SAS Studio SAS Visual Analytics SAS Visual Statistics SAS econometrics SAS Visual Forecasting SAS Visual Data Mining and machine learning SAS User Metadata SAS Metadata Data Access Base SAS & SAS/ACCESS to Hadoop & Impala & Hawk In-Memory In-Memory Data Access Data Access Data Processing Pig Impala Hawk Hive Hive Pig Sqoop Map Reduce SAS Embedded Process SAS Cloud Analytics Services File System HDFS YARN
A variety of analytical tools available to support a range of end user needs 5
Parallel & Serial, Pub / Sub, Web Services, MQs The SAS platform architecture Source-based Engines Customer Intelligence Analytics In-Stream In-Hadoop In-Database Data Platforms Cloud & Containers In-Memory Engine Cloud Analytic Services (CAS) Platform UAA UAA UAA Viya Core Microservices Data Source Mgmt Folders etc BI GUIs CAS Mgmt Log Analytics GUIs Data Mgmt GUIs Query Gen Env Mgr Model Mgmt Audit Solutions APIs Solutions Developers Business Visualization Risk Management! Fraud and Security Intelligence Data Management Google Cloud Platform
Open-source connectivity CAS SAS Viya Run R & Python from IDE eg Jupyter GIT sassoftware/python-swat package (SAS Scripting Wrapper for Analytics Transfer (SWAT)) GIT sassoftware/r-swat
Language versus actions An Example proc print data = hmeq (obs = 10); run; APIs Controller Workers df = s.castable( hmeq ) df.head(10) Translated Action [table.fetch] table.name = hmeq from = 1 to = 10 df <- defcastable(s, hmeq ) head(df, 10)
A shared services operating model 6 Decentralisation Higher costs Variable Standards Different Control Environments Duplications of Efforts Business Units retain Control Recognition of Local Priorities Shared Lean, Flat Organisation Independent of Businesses Identification of Efficiencies between Business Units Understanding of Group Functions and Missions Dissemination of Best Practices Common Systems & Support Efficient Knowledge Transfer, Standards & Tools Economies of Scale Centralised Unresponsive No Business Unit control of Central Overhead Costs Inflexible to Business Unit Needs Disconnect from Business Units
A governed environment, with room for experimentation 7 Analytics IT governance Data governance Analytic governance Big Data Lab Factory
The seven traits of a modern analytical platform 1. Process supports the analytical lifecycle 2. Engines for processing data in-memory, in-database and real-time streaming 3. The platform is scalable and can grow 4. Access to a variety of data sources 5. A variety of analytical tools available to support a range of end user needs 6. A shared services operating model 7. A governed environment, with room for experimentation
Thank you for joining the the data science jam sessions. Enjoy the networking drink & music! Going to and& festival or technology expedition? Download the and& summit app!
Thanks for your attention
MATHIAS COOPMANS Email: mathias.coopmans@sas.com Mobile: +32 475 571 284 @macoopma