ETL is No Longer King, Long Live SDD

Transcription

1 ETL is No Longer King, Long Live SDD How to Close the Loop from Discovery to Information () to Insights (Analytics) to Outcomes (Business Processes) A presentation by Brian McCalley of DXC Technology, Glenn Field of SiriusIQ and Gavin Robertson of WhamTech, Inc.

2 No secret that most organizations face major datarelated hurdles Location System Access Security Container Format Age Dirty Typo/Transposition Missing Meaning Duplication Obfuscation Governance

3 and ANALYTICS is the prime driver to lower costs and increase revenue which, in turn, drives the need for applications* to have clean and understood data in specific formats *Reporting, BI, analytics, CDI-MDM, CRM, SCM, fraud detection, antimoney laundering, ERP, etc.

4 Goals for an optimal data architecture 1. Complete, clean, transformed, standardized and secure data, and master data, for multiple applications 2. Near real-time minimal update and query latency 3. Automation, including workflow and event processing 4. Support reporting, BI and analytics, including graph database 5. Minimize copies of data 6. discovery, metadata repository and data governance 7. Write back to data sources

5 Managing data in, or from, multiple disparate systems requires a new approach CONVENTIONAL APPROACHES ETL: Copy and transform schemas and data to a one-size-fits-all data warehouse Copy: To a single Lake/Big repository Federate: Submit queries through adapters to source systems Search: E.g., Solr /Elasticsearch - copy, read, parse and index data, process queries to provide data Big options for storing data CULTURAL HURDLES TECHNICAL HURDLES SECURITY & PRIVACY HURDLES The new approach leverages the advantages of each conventional approach

6 Typical Warehouse/Big / Lake/Search Application(s) Source Source Source Extract Extract Extract and Schema Transform and Schema Transform and Schema Transform Load Load Load Warehouse/ Big / Lake/ Search Queries resolved in the Warehouse/Big / Lake/ Search Expensive in terms of time and cost to implement and maintain

7 Typical federated data access with conventional adapters Source Connector Adapter Source Connector Adapter Middleware Application(s) Source Connector Adapter Queries resolved mainly in the data source Expensive to implement and limits capabilities

8 ETL has been the only option to come close to meeting the goals for an optimal data architecture, until now Introducing Software Defined (SDD) consisting of unconventional federated adapters that Read, Transform (process) and Index (RTI) source data and process queries against these indexes

9 SDD initial discovery, index and adapter configuration, index build and Standard View (SDV) mapping Discovery and raw indexbased Profiling Metadata Discovery and Semantic Mapping Classification and Security Develop and test Transforms using profiles Distributed Metadata Repository, incl. Governance Source Discovery Network Asset and Device Discovery Source Read, Transform/ clean-up (and Index) Indexes SDD Adapter w/sdv Twelve ways to build and maintain indexes Index schema and names same as data source Indexes do not store data only queryable representations Indexes mapped to Standard View (SDV)

10 SDD index update, query processing and results retrieval Distributed Metadata Repository, incl. Governance Queries resolved in the adapter and indexes Result-set pointers to data in source Source Read, Transform/ clean-up (and Index) User-level access Continuous EIQ Indexes updates Indexes SDD Adapter w/sdv Raw results data transformed/cleaned-up from source SDD Federation Server (submiddleware) w/sdv Applications/middleware connect with standard drivers, APIs, Web/data services and SQL Middleware Application(s) Results provided in almost any format SDD Federation Server Multiple other data sources

11 F I R E W A L L F I R E W A L L SDD adapters can co-exist with other types of adapters to System of Records RDBMS Indexes SDD Adapter Social Media Feed Indexes SDD Adapter SDD Federation Server Hadoop Indexes SDD Adapter Mainframe ERP System Indexes SDD Conventional Adapter SDD Adapter SDD Federation Server SDD Federation Server TCP / IP ODBC/ JDBC Driver REST API Etc. Application(s) Salesforce 3 rd Party Adapter

12 SDD indexes and adapters can be deployed and accessed anywhere, and at any level in multiple combinations Indexes are 100% contiguous, regardless of where or how deployed

13 How does Software Defined compare with other approaches?

14 Goal #1: Complete, clean, transformed, standardized and secure data, and master data, for multiple applications Goal #1 Complete, clean, transformed, standardized and secure data, and master data, for multiple applications SDD ETL to a Warehouse Big Lake Conventional Federated Adapters Solr/ Elasticsearch ( )

15 Goal Goal #2: Near real-time minimize update and query latency #1 Complete, clean, transformed, standardized and secure data, and master data, for multiple applications SDD ETL to a Warehouse Big Lake Conventional Federated Adapters Solr/ Elasticsearch ( ) #2 Near real-time minimize update and query latency ( ) ( ) ( ) ( )

16 Goal Goal #3: Automation, including workflow and event processing #1 Complete, clean, transformed, standardized and secure data, and master data, for multiple applications SDD ETL to a Warehouse Big Lake Conventional Federated Adapters Solr/ Elasticsearch ( ) #2 Near real-time minimize update and query latency ( ) ( ) ( ) ( ) #3 Automation, including workflow and event processing ( )

17 Goal Goal #4: Support reporting, BI and analytics, including graph database #1 Complete, clean, transformed, standardized and secure data, and master data, for multiple applications SDD ETL to a Warehouse Big Lake Conventional Federated Adapters Solr/ Elasticsearch ( ) #2 Near real-time minimize update and query latency ( ) ( ) ( ) ( ) #3 Automation, including workflow and event processing ( ) #4 Support reporting, BI and analytics, including graph database ( ) ( ) ( ) ( )

18 Goal #5: Minimize copies of data Goal #1 Complete, clean, transformed, standardized and secure data, and master data, for multiple applications SDD ETL to a Warehouse Big Lake Conventional Federated Adapters Solr/ Elasticsearch ( ) #2 Near real-time minimize update and query latency ( ) ( ) ( ) ( ) #3 Automation, including workflow and event processing ( ) #4 Support reporting, BI and analytics, including graph database ( ) ( ) ( ) ( ) #5 Minimize copies of data

19 Goal Goal #6: discovery, metadata repository and data governance #1 Complete, clean, transformed, standardized and secure data, and master data, for multiple applications SDD ETL to a Warehouse Big Lake Conventional Federated Adapters Solr/ Elasticsearch ( ) #2 Near real-time minimize update and query latency ( ) ( ) ( ) ( ) #3 Automation, including workflow and event processing ( ) #4 Support reporting, BI and analytics, including graph database ( ) ( ) ( ) ( ) #5 Minimize copies of data #6 discovery, metadata repository and data governance ( ) ( ) ( )

20 Goal #7: Write back to data sources Goal #1 Complete, clean, transformed, standardized and secure data, and master data, for multiple applications SDD ETL to a Warehouse Big Lake Conventional Federated Adapters Solr/ Elasticsearch ( ) #2 Near real-time minimize update and query latency ( ) ( ) ( ) ( ) #3 Automation, including workflow and event processing ( ) #4 Support reporting, BI and analytics, including graph database ( ) ( ) ( ) ( ) #5 Minimize copies of data #6 discovery, metadata repository and data governance ( ) ( ) ( ) #7 Write back to data sources

21 How SDD meets goals for an optimal data architecture Goal #1 Complete, clean, transformed, standardized and secure data, and master data, for multiple applications Software Defined Process source data as building and maintaining indexes and master data, and as reading raw results data Multiple indexes, views, means of access and result formats #2 Near real-time minimize update and query latency Changed data capture High performance, parallel distributed processing almost no load on data sources #3 Automation, including workflow and event processing Index monitoring, REST APIs and workflow integration #4 Support reporting, BI and analytics, including graph database Indexed views, provision highly curated data to analytics, run analytics, and built-in virtual graph database and link analysis #5 Minimize copies of data Can leave and secure data in sources, a Lake or indexes #6 discovery, metadata repository and data governance Use raw indexes for discovery, metadata and combining with IAM and RBAC for data governance from edge/bottom up #7 Write back to data sources Can read as well as insert, delete and update

22 Software Defined (SDD) Implementation and alignment of use-cases is the key to driving Enterprise IP. Technology prohibits this due to binding of data elements within applications Freeing data to create workflows will dramatically reduce time to market Incrementally developing enterprise use-cases through SDD drives innovation to next-gen path allowing a reinvention of the enterprise Agnostic de-coupling of silo solutions drives speed to market Use-case consumption for any user, on any device, anywhere securely enhances collaboration and productivity

23 Software Defined (SDD) Process to eliminate upwards of 95% of today s regression issues translates into 50-75% calendar time savings Business logic and code base functions grow incrementally as business dictates Cumulative code growth ensures reuse, optimal performance and agnostic access. You also benefit from globally available logic Zero impact deployments eliminates downtime and simplifies the SDLC Dynamic, intelligent workflows consume new features when live

24 AI and NLP are the new UI for many applications Leveraging next-gen cognitive rapidly delivers functionality and results Millennial workforce alignment Dramatic decrease in training requirements Dramatic decrease in time to market on features and results Disconnect 3 rd party backend solutions from natural language UI Allow seamless app upgrades by using AI UI, which will interact with both old and new systems

25 Abstracted 3-tier architecture connected through a Smart Fabric PLATFORM MANAGEMENT SERVICES Ecosystem Solutions SYSTEM OF ENGAGEMENT - APPLICATIONS Line-of-Business Users SYSTEM OF INSIGHTS ANALYTICS WAREHOUSE SYSTEM OF RECORDS INTERNAL / EXTERNAL DATA SOURCES SDD Fabric API Catalog

26 SDD is the First Paradigm Shift in how data and analytics are managed in a common meta-object framework *Indexes per data source on structured, unstructured and semistructured data that can either store or not store (default) data SYSTEM OF ENGAGEMENT - APPLICATIONS SYSTEM OF INSIGHTS ANALYTICS WAREHOUSE SYSTEM OF RECORDS INTERNAL / EXTERNAL DATA SOURCES Identity Management Access Control Other Corporate Governance Application and Other Corporate Security Standard drivers, APIs, Web/data services, SQL, Spark SQL and other query languages Distributed Management, Including Source Results Read and Write Back INDEXES* Metadata Repository Governance Role-based Access Control API Management Business Rules Discovery Profiling Security Analytics Quality Transformation Standardization Relationships BPM Software Master Management Audit Logs Virtual Graph base Optional Lake Real-time, Incremental or Batch Updated, Centralized or Distributed

27 The Second Paradigm Shift is the concept of the Analytics Warehouse

28 The Second Paradigm Shift is the concept of the Analytics Warehouse

29 An example of an Analytics Warehouse architecture data ingestion/warehouse model

30 An example of an Analytics Warehouse architecture spanning enterprise systems federated model

31 Examples of applying SDD and an Analytics Warehouse to healthcare analytics

32 3rd Party Partner Ecosystem Healthcare Cloud SDD Fabric Healthcare Clinical Network Management 3-Tier Architecture End-End Integrated Application Ecosystem Environment across Continuum of Care SoE Analytical Insights Catalog and Ecosystem Clinical, Financial, Administrative, Operational SoI Enterprise/ Ecosystem-wide Master Management Platform Longitudinal Patient Record SoR Development and IT Operations Support Environment Human Capital Management, Finance, Sales & Marketing AI Driven Operations Automation End-to-End Security

33 SDD Fabric enables a Longitudinal Patient Record (LPR) view across multiple System of Records, across multiple enterprises Transparent distributed data management layer that plugs-and-plays in existing IT infrastructures Complements and leverages existing IT systems, tools and applications Leave and guard data in sources, copies, e.g., Lake, or stored in indexes a hybrid approach Address upfront data discovery, security, quality, standards, MDM and other data-related processes

34 Use cases from healthcare that combine data and analytics management Use Cases Applications Clinical Applications Diabetes, Hypertension, Heart Failure, etc. Gaps in Care Predictive Readmissions Management Clinical Wellness Management Operational Management Operational Management Hospital Operational Management Physician Practices Physician Quality Reporting Scores Financial Performance Financial Management Hospital Financial Management Physician Practices Claims Analytics Regulatory Reporting Hospital Value Based Purchasing (HVBP) HEDIS Patient Centered Medical Home Scorecard MU 2 Clinical Quality Measures-Hospitals/Physicians MU2 Usage Scorecard - Physicians ACO Quality Reporting Hospital Outpatient Quality Reporting Patient Similarity Comparative Effectiveness Research Predictive models Chronic disease management Ability to create patient cohorts Cohort Manager/Chronic Condition Management Population Management Population Focus & Population Care

35 Conclusion of why ETL is no longer King, long live SDD SDD enables a data management paradigm shift SDD supports an analytics management paradigm shift ETL still has value, but not exclusively SDD can greatly enhance, complement and/or replace ETL in the future SDD is more suited than ETL to the new world of: everywhere API data services/catalog Parallel distributed processing Event and workflow processing Near real-time architectures

36 Thank you A presentation by Brian McCalley of DXC Technology, Glenn Field of SiriusIQ and Gavin Robertson of WhamTech, Inc.

37 Q&A