Docstoc

Title Arial pt

Document Sample
Title Arial pt Powered By Docstoc
					Integrating SAS & Teradata
Past – Present - Future




Scott Overby
Vice President, Decision Support

Amanda Kreutziger
Director, Decisioning and Business Intelligence Applications
Discover Financial Services
• Leading Credit Card Issuer and
  Electronic Payment Services
  Company
    – Issuer of the Discover Card
      • Pioneer of Cash Rewards


• Payment Services – Credit, Debit
  & Prepaid
    –   Discover Network
    –   PULSE ATM/Debit Network
    –   Diners Club International
    –   Alliances
        • China Union Pay
        • JCB
        • BC card

• Discover Bank
    –   Savings/CD’s
    –   Money Market
    –   Personal Loans
    –   Student Loans
1
SAS @ Discover
• SAS Installed Since The Early
  1990’s

• Primary Tool of Choice for Analytic
  Community

• More than 650 Users of Varying
  Skills and Roles

• User Differentiation:
    – Power Users: Bias Toward Base
      SAS / STAT, Macros, Proc SQL (75)
    – Mid-Level: Base SAS / STAT and EG
      (300+)
    – Low–Level: EG, Enterprise Miner
      (250)

2
Teradata @ Discover
• Teradata is Discover’s 3rd Generation Data Warehouse

• First Teradata Implementation in 2005
    – Now 5500/5400/2500 Across Data Centers

• Created Foundations of a True EDW
    –   Centralized
    –   Scalable
    –   Stable
    –   Reusable Data
    –   Data as a Service

• Empowered Business Users and Reduced IT Dependency

• Emergence of Real Time Use of Data
    – “Operational” Analytics Phenomenon
3
   Common SAS Implementation

   • Many SAS/Data Deployments
         – Lines of Business, Business Function, Data Sources and Geography
           Drive Deployment Model
                                      SAS Users                                 SAS Users




                                        Line Of Business/
                                                                                      Line Of Business/
       SAS Users                        Business Function
                                                                                      Business Function




Line Of Business/
Business Function
                                                            Line Of Business/
                                                            Business Function

                                                                                                     Data
                       Departmental                                                                Warehouse
                          Marts
                                                                    SPDS




   4
 Achieving Symbiotic Relationships
• 2006 Implementation Leveraging Best Practices
     – Server Consolidation on IBM SMP Power Architecture
     – Data Consolidation Through the Teradata EDW

• Eliminated SPDS & Migrated Sourcing to EDW



                                        SAS
                        SAS Users




                                                    SAS Users
                                    Teradata Data
           Other Data
                                     Warehouse




 5
In Retrospect Our Key Decisions
• Establishment of the Enterprise Data Warehouse
    – Common Data Sources for Model Development and Analysis
    – Emphasis on Data Quality / Governance

• Consolidation of SAS Server Infrastructure and Ownership
    – Infrastructure Decisions are Made Consistently
    – More Users But Larger More Stable Platform

• Eliminating SPDS and Sourcing Data From the Enterprise Data
  Warehouse
    – Reduced Resourcing of Data, Data Movement, Cost
    – Migrate SQL Steps into Database for Performance

• 2010 GRID Deployment

6
Legacy SAS Infrastructure

• Disadvantages
  – Scalability
  – Hardware Failure Risk
  – Expensive DR                                      SAS




                                      SAS Users




                                                                  SAS Users
  – Disruptive Upgrades

• 2010 presented opportunity to
  build 3rd Generation SAS                        Teradata Data
                                  Other Data
  Environment                                      Warehouse




• Grid implemented in April
  2010



7
   Grid Architecture
                    SAS Users
                                                                         Teradata
       SAS GRID                                                       Data Warehouse
  Master

                                                        • New Hardware/Storage Infrastructure
                                                          – Typically low-cost servers
       Node 1        Node 3     Node 5     Node 7         – Ability to scale out as
                                                            environment/user base grows
Failover

                                                        •   SAS Grid Computing
                                                            – Management/Governance of SAS
        Node 2        Node 4     Node 6     Node 8            Workloads
4 processors per Node optimized for heavy computation       – Prioritization (Queuing) capability
                   XIV/GPFS Storage                         – High Availability Configuration

                                                        • SAS Version Upgrade – SAS 9.1.3 to SAS
                                                          9.2

   8
Benefits
• Hot fixes & rolling server reboots with no impact to users
• Workload Balancing with Rules-Based Prioritization
• Horizontal scalability
• Co-Location of SAS Grid environment with Teradata
• Run Time Improvements:
  – Key Segmentation Model: 4 hour 50 minutes to 1 hour 47
    minutes
  – Critical Marketing Model: 9 hours 7 minutes to 3 hours 35
    minutes




9
Grid Lessons Learned
              • Chart existing utilization patterns

              • Strong Unix Operations/Administration
                support

              • Plan for adjustment of jobs to take
                advantage of Grid parallelism

              • Establish Power User Group




10
 What’s Next?
• Tune & Optimize Grid and Expand as
  needed

• Next Generation of Training and internal
  SAS/Teradata User Group

• DR Capability

• Implement SAS 9.2 M3
  – Grid Performance Charting/Optimization
  – Explore SAS In-Database Procedures on
    Teradata v13


 11
Future State

• SAS Analytics in The Warehouse – DFS is Ready
     – Reduces Server Infrastructure
     – Eliminates Data Movement
     – Increased ROI on Model Implementation
                   SAS Users




                                               SAS Users
                                  Data
                                  Data
                                Warehouse
                                Warehouse




12
Primary Challenge to the Future State

• DFS “Command Prompt” Bias Toward Base SAS/STAT
     – SAS / Teradata Partnership Focus on Applications (e.g. Enterprise Miner)
     – Limited Deployment of SAS PROCS in Teradata
                     SAS Users




                                                      SAS Users
                                     Data
                                     Data
                                   Warehouse
                                   Warehouse




13
 Questions




14

				
DOCUMENT INFO