PayPal decisions as a service


Published on

Describes patterns for decisioning in the cloud

Published in: Technology
  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide
  • Mr. Pradeep Ballal works as a Senior Architect in the Core Service Product Development with specific focus on Compliance and Risk products with PayPal Singapore. Mr. Ballal is a software generalist with 13 years of technology experience and has special interest in decision management, business rules, enterprise software and architectures. He has helped various large corporates world-wide adopt decision management technologies to increase business agility.Mr. Staale Nerboe works as a Senior Architect in the Core Service Product Development organization withPayPal Singapore. Mr. Nerboe has 15+ years of Technology Consulting and Software Architecture experience for large global companies world-wide.
  • A pattern for organizing design time database parts A pattern for organizing decision configuration parts A pattern for organizing decision runtime parts A pattern for organizing decision runtime data parts
  • PayPal decisions as a service

    1. 1. June 2013 Decisions as a Service Risk & Compliance Engineering, PayPal Pradeep Ballal Staale Nerboe This deck contains generic architecture information, and does not reflect the exact details of current or planned systems.
    2. 2. Confidential and Proprietary2 • Encompasses processes to improve, streamline and automate operational decision making within organizations. • Use all available organizational resources to increase precision, consistency and agility of decisions. • Treat decisions as reusable assets and leverage technology at key decision points to automate the process. • Uses tools such as Predictive Analytics Business Intelligence Business Rules Adaptive Control Artificial Intelligence DECISION MANAGEMENT
    3. 3. Confidential and Proprietary3 IMPORTANCE OF OPERATIONAL DECISION MANAGEMENT Low High LowHigh Strategic Decisions e.g. New markets, M&A Tactical Decisions e.g. New products, pricing, customer segmentation Operational Decisions e.g. Loan approvals, insurance application approvals, customer upgrades, cross-sell/up- sell, marketing offers Decision Volume DecisionValue
    4. 4. Confidential and Proprietary4 Decision Simulation Decision Optimi- zation Business Intelligence Business Rules Optimiz- ation Predictive models DECISION MANAGEMENT CYCLE Insights into Improvement Operational Decision Data Business Data Operational Strategic
    5. 5. Confidential and Proprietary5 AN ARCHITECTURE FOR DECISION MANAGEMENT Data warehouse External Data Industry Data Analytic Workbench Models Rules Operational Data stores Rules Management Rules Policy Documents Code Data Business Experts Adaptive Control Business Decisions Insights Feedback Loop Operational Systems Channels – web, mobile, contact center etc. Decision Service Analytic process Decision Modeling
    6. 6. Confidential and Proprietary6 COMPONENTS OF A DECISION ENGINE Designer Configuration center Repository Deployment Container rules rules rules rules Decision Service Decision Service Decision Service Client applications Design time Run time • Distinct components targeted to distinct roles • Design time − Define frameworks within which operational decisions are managed − Configure models and rules that make up the decision − E.g. setting up data models, rule structures, invocation models etc. • Run time − Managed execution of business rules to output decisions − Consumed by client applications via “Decision Services” Developers Business users End users System Admins
    7. 7. Confidential and Proprietary7 • Clients - internal cloud or external cloud? • Self service – all components need to be provisioned on a self service basis. Provide flexibility to cherry pick from various available components. • Multi-tenancy – for internal cloud, each team/domain can be a tenant within the cloud decision management infrastructure. Each tenant is isolated and gets all the services in the cloud based decisioning infrastructure. • Web based rules & models management – Web based interface to manage policies that lead up to the decision as well perform verification & validation. • Managed APIs – Provide REST APIs to interact with both design time aspects (repository, rule definitions, data models etc.) and run time (execute rules, rule analytics etc.). • Simulations – Invoke decisions against a sample set of input data to determine impact and optimize decisions. • Data Mining – Capture decisions for adaptive controls or corrections DECISION ENGINE IN THE CLOUD - CONSIDERATIONS
    8. 8. Confidential and Proprietary8 MULTI-TENANCY PaaS IaaS Tenant1 Tenant2 Tenant3 • One instance of the software system serves one tenant. • Tenant data fully isolated and not visible to each other. • Configuration center should have much of the functionality to enable self-service. • No technical development effort is required. • Rules can be configured immediately. Designer Automation Interface Configuration center Repository Deployment Container rules rules rules rules Decision Service Decision Service Decision Service Client applications Design time Run time Developers Business users End users System Admins Tenant4
    9. 9. Confidential and Proprietary9 Decision Server INDIVIDUAL TENANT ORGANIZATION Rules Repository User & Preferences Store Simulation (In/out data) Decision Management Portal User Management Rules Management Simulation Controller Decision Warehouse Decision Svc 1 Decision Svc 2 Decision Svc 3 Deployment Manager Server Monitor Rules Source Decision Server Decision Warehouse Decision Svc 1 Decision Svc 2 Decision Svc 3 Deployment Manager Server Monitor Rules Source deploy deploy  Design data pattern  Decision configuration pattern Decision server pattern  Decision server data pattern  Decision server pattern  Decision server data pattern StageLive Dev JSON JSON JSON JSON Client applications Model Management CEP Service Framework
    10. 10. Confidential and Proprietary10 • Each tenant should be configurable by adding parts • Built with parts − A database part (for user, preferences, rules, simulation data etc.) − A simulation application part for running simulations on eligible decision services − A rules maintenance part for managing decisions and creating new. • Group parts into patterns − A pattern for design time authoring. Some patterns may omit parts (for e.g. simulation not required all the time) − Another pattern for executing decisions (runtime). • Group patterns into virtual systems deployed in virtual environments − The design time data pattern and app pattern assembled together to form a virtual system for decision maintenance. INDIVIDUAL TENANT ORGANIZATION
    11. 11. Confidential and Proprietary11 Hadoop DATA PROCESSING FOR DECISIONINGData CacheEvent Data Rollup Offline Variables Clients • Transparently merges real time event data with offline data • Combined data blends the reliability of offline with the low latency of online data • Heavy calculations and large rollups are all done offline. • All data stored in highly available cache for fast access Data Warehouse Data Events DS DS DS CEP Filter Aggreg ate Data Window Pattern Join Variables PaaS
    12. 12. Confidential and Proprietary12 DECISION SERVICE DEVELOPMENT WORKFLOW Development workflow CloudliveCloudstaging Analysis& Design Cloud environment Select pattern and provision Is data model available? Design a data model Import data model into environment Create new decision service using the data model no yes Configure decisions & test Test REST end point from application Ready to deploy no yes Ready to deploy no Deploy decision service yes Deploy decision service Design a decision model, identify decision points All environment settings are preconfigured in the pattern. Development process starts here early! One click deploy process reduce admin overhead Operationalize strategies, models and business rules quickly and scale them to meet market demands.
    13. 13. THANK YOU